root / lib / cmdlib / __init__.py @ f380d53c
History | View | Annotate | Download (399.6 kB)
1 |
#
|
---|---|
2 |
#
|
3 |
|
4 |
# Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013 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=W0201,C0302
|
25 |
|
26 |
# W0201 since most LU attributes are defined in CheckPrereq or similar
|
27 |
# functions
|
28 |
|
29 |
# C0302: since we have waaaay too many lines in this module
|
30 |
|
31 |
import os |
32 |
import time |
33 |
import logging |
34 |
import copy |
35 |
import OpenSSL |
36 |
import itertools |
37 |
import operator |
38 |
|
39 |
from ganeti import utils |
40 |
from ganeti import errors |
41 |
from ganeti import hypervisor |
42 |
from ganeti import locking |
43 |
from ganeti import constants |
44 |
from ganeti import objects |
45 |
from ganeti import compat |
46 |
from ganeti import masterd |
47 |
from ganeti import netutils |
48 |
from ganeti import query |
49 |
from ganeti import qlang |
50 |
from ganeti import opcodes |
51 |
from ganeti import ht |
52 |
from ganeti import rpc |
53 |
from ganeti import pathutils |
54 |
from ganeti import network |
55 |
from ganeti.masterd import iallocator |
56 |
|
57 |
from ganeti.cmdlib.base import ResultWithJobs, LogicalUnit, NoHooksLU, \ |
58 |
Tasklet, _QueryBase |
59 |
from ganeti.cmdlib.common import _ExpandInstanceName, _ExpandItemName, \ |
60 |
_ExpandNodeName, _ShareAll, _CheckNodeGroupInstances, _GetWantedNodes, \ |
61 |
_GetWantedInstances, _RunPostHook, _RedistributeAncillaryFiles, \ |
62 |
_MergeAndVerifyHvState, _MergeAndVerifyDiskState, _GetUpdatedIPolicy, \ |
63 |
_ComputeNewInstanceViolations, _GetUpdatedParams, _CheckOSParams, \ |
64 |
_CheckHVParams, _AdjustCandidatePool, _CheckNodePVs, \ |
65 |
_ComputeIPolicyInstanceViolation, _AnnotateDiskParams, _SupportsOob, \ |
66 |
_ComputeIPolicySpecViolation, _GetDefaultIAllocator, \ |
67 |
_CheckInstancesNodeGroups, _LoadNodeEvacResult, _MapInstanceDisksToNodes, \ |
68 |
_CheckInstanceNodeGroups
|
69 |
|
70 |
from ganeti.cmdlib.cluster import LUClusterActivateMasterIp, \ |
71 |
LUClusterDeactivateMasterIp, LUClusterConfigQuery, LUClusterDestroy, \ |
72 |
LUClusterPostInit, _ClusterQuery, LUClusterQuery, LUClusterRedistConf, \ |
73 |
LUClusterRename, LUClusterRepairDiskSizes, LUClusterSetParams, \ |
74 |
LUClusterVerify, LUClusterVerifyConfig, LUClusterVerifyGroup, \ |
75 |
LUClusterVerifyDisks
|
76 |
from ganeti.cmdlib.group import LUGroupAdd, LUGroupAssignNodes, \ |
77 |
_GroupQuery, LUGroupQuery, LUGroupSetParams, LUGroupRemove, \ |
78 |
LUGroupRename, LUGroupEvacuate, LUGroupVerifyDisks |
79 |
from ganeti.cmdlib.tags import LUTagsGet, LUTagsSearch, LUTagsSet, LUTagsDel |
80 |
from ganeti.cmdlib.network import LUNetworkAdd, LUNetworkRemove, \ |
81 |
LUNetworkSetParams, _NetworkQuery, LUNetworkQuery, LUNetworkConnect, \ |
82 |
LUNetworkDisconnect
|
83 |
from ganeti.cmdlib.test import LUTestDelay, LUTestJqueue, LUTestAllocator |
84 |
|
85 |
import ganeti.masterd.instance # pylint: disable=W0611 |
86 |
|
87 |
|
88 |
# States of instance
|
89 |
INSTANCE_DOWN = [constants.ADMINST_DOWN] |
90 |
INSTANCE_ONLINE = [constants.ADMINST_DOWN, constants.ADMINST_UP] |
91 |
INSTANCE_NOT_RUNNING = [constants.ADMINST_DOWN, constants.ADMINST_OFFLINE] |
92 |
|
93 |
#: Instance status in which an instance can be marked as offline/online
|
94 |
CAN_CHANGE_INSTANCE_OFFLINE = (frozenset(INSTANCE_DOWN) | frozenset([ |
95 |
constants.ADMINST_OFFLINE, |
96 |
])) |
97 |
|
98 |
|
99 |
def _IsExclusiveStorageEnabledNode(cfg, node): |
100 |
"""Whether exclusive_storage is in effect for the given node.
|
101 |
|
102 |
@type cfg: L{config.ConfigWriter}
|
103 |
@param cfg: The cluster configuration
|
104 |
@type node: L{objects.Node}
|
105 |
@param node: The node
|
106 |
@rtype: bool
|
107 |
@return: The effective value of exclusive_storage
|
108 |
|
109 |
"""
|
110 |
return cfg.GetNdParams(node)[constants.ND_EXCLUSIVE_STORAGE]
|
111 |
|
112 |
|
113 |
def _IsExclusiveStorageEnabledNodeName(cfg, nodename): |
114 |
"""Whether exclusive_storage is in effect for the given node.
|
115 |
|
116 |
@type cfg: L{config.ConfigWriter}
|
117 |
@param cfg: The cluster configuration
|
118 |
@type nodename: string
|
119 |
@param nodename: The node
|
120 |
@rtype: bool
|
121 |
@return: The effective value of exclusive_storage
|
122 |
@raise errors.OpPrereqError: if no node exists with the given name
|
123 |
|
124 |
"""
|
125 |
ni = cfg.GetNodeInfo(nodename) |
126 |
if ni is None: |
127 |
raise errors.OpPrereqError("Invalid node name %s" % nodename, |
128 |
errors.ECODE_NOENT) |
129 |
return _IsExclusiveStorageEnabledNode(cfg, ni)
|
130 |
|
131 |
|
132 |
def _CopyLockList(names): |
133 |
"""Makes a copy of a list of lock names.
|
134 |
|
135 |
Handles L{locking.ALL_SET} correctly.
|
136 |
|
137 |
"""
|
138 |
if names == locking.ALL_SET:
|
139 |
return locking.ALL_SET
|
140 |
else:
|
141 |
return names[:]
|
142 |
|
143 |
|
144 |
def _ReleaseLocks(lu, level, names=None, keep=None): |
145 |
"""Releases locks owned by an LU.
|
146 |
|
147 |
@type lu: L{LogicalUnit}
|
148 |
@param level: Lock level
|
149 |
@type names: list or None
|
150 |
@param names: Names of locks to release
|
151 |
@type keep: list or None
|
152 |
@param keep: Names of locks to retain
|
153 |
|
154 |
"""
|
155 |
assert not (keep is not None and names is not None), \ |
156 |
"Only one of the 'names' and the 'keep' parameters can be given"
|
157 |
|
158 |
if names is not None: |
159 |
should_release = names.__contains__ |
160 |
elif keep:
|
161 |
should_release = lambda name: name not in keep |
162 |
else:
|
163 |
should_release = None
|
164 |
|
165 |
owned = lu.owned_locks(level) |
166 |
if not owned: |
167 |
# Not owning any lock at this level, do nothing
|
168 |
pass
|
169 |
|
170 |
elif should_release:
|
171 |
retain = [] |
172 |
release = [] |
173 |
|
174 |
# Determine which locks to release
|
175 |
for name in owned: |
176 |
if should_release(name):
|
177 |
release.append(name) |
178 |
else:
|
179 |
retain.append(name) |
180 |
|
181 |
assert len(lu.owned_locks(level)) == (len(retain) + len(release)) |
182 |
|
183 |
# Release just some locks
|
184 |
lu.glm.release(level, names=release) |
185 |
|
186 |
assert frozenset(lu.owned_locks(level)) == frozenset(retain) |
187 |
else:
|
188 |
# Release everything
|
189 |
lu.glm.release(level) |
190 |
|
191 |
assert not lu.glm.is_owned(level), "No locks should be owned" |
192 |
|
193 |
|
194 |
def _CheckOutputFields(static, dynamic, selected): |
195 |
"""Checks whether all selected fields are valid.
|
196 |
|
197 |
@type static: L{utils.FieldSet}
|
198 |
@param static: static fields set
|
199 |
@type dynamic: L{utils.FieldSet}
|
200 |
@param dynamic: dynamic fields set
|
201 |
|
202 |
"""
|
203 |
f = utils.FieldSet() |
204 |
f.Extend(static) |
205 |
f.Extend(dynamic) |
206 |
|
207 |
delta = f.NonMatching(selected) |
208 |
if delta:
|
209 |
raise errors.OpPrereqError("Unknown output fields selected: %s" |
210 |
% ",".join(delta), errors.ECODE_INVAL)
|
211 |
|
212 |
|
213 |
def _CheckParamsNotGlobal(params, glob_pars, kind, bad_levels, good_levels): |
214 |
"""Make sure that none of the given paramters is global.
|
215 |
|
216 |
If a global parameter is found, an L{errors.OpPrereqError} exception is
|
217 |
raised. This is used to avoid setting global parameters for individual nodes.
|
218 |
|
219 |
@type params: dictionary
|
220 |
@param params: Parameters to check
|
221 |
@type glob_pars: dictionary
|
222 |
@param glob_pars: Forbidden parameters
|
223 |
@type kind: string
|
224 |
@param kind: Kind of parameters (e.g. "node")
|
225 |
@type bad_levels: string
|
226 |
@param bad_levels: Level(s) at which the parameters are forbidden (e.g.
|
227 |
"instance")
|
228 |
@type good_levels: strings
|
229 |
@param good_levels: Level(s) at which the parameters are allowed (e.g.
|
230 |
"cluster or group")
|
231 |
|
232 |
"""
|
233 |
used_globals = glob_pars.intersection(params) |
234 |
if used_globals:
|
235 |
msg = ("The following %s parameters are global and cannot"
|
236 |
" be customized at %s level, please modify them at"
|
237 |
" %s level: %s" %
|
238 |
(kind, bad_levels, good_levels, utils.CommaJoin(used_globals))) |
239 |
raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
|
240 |
|
241 |
|
242 |
def _CheckNodeOnline(lu, node, msg=None): |
243 |
"""Ensure that a given node is online.
|
244 |
|
245 |
@param lu: the LU on behalf of which we make the check
|
246 |
@param node: the node to check
|
247 |
@param msg: if passed, should be a message to replace the default one
|
248 |
@raise errors.OpPrereqError: if the node is offline
|
249 |
|
250 |
"""
|
251 |
if msg is None: |
252 |
msg = "Can't use offline node"
|
253 |
if lu.cfg.GetNodeInfo(node).offline:
|
254 |
raise errors.OpPrereqError("%s: %s" % (msg, node), errors.ECODE_STATE) |
255 |
|
256 |
|
257 |
def _CheckNodeNotDrained(lu, node): |
258 |
"""Ensure that a given node is not drained.
|
259 |
|
260 |
@param lu: the LU on behalf of which we make the check
|
261 |
@param node: the node to check
|
262 |
@raise errors.OpPrereqError: if the node is drained
|
263 |
|
264 |
"""
|
265 |
if lu.cfg.GetNodeInfo(node).drained:
|
266 |
raise errors.OpPrereqError("Can't use drained node %s" % node, |
267 |
errors.ECODE_STATE) |
268 |
|
269 |
|
270 |
def _CheckNodeVmCapable(lu, node): |
271 |
"""Ensure that a given node is vm capable.
|
272 |
|
273 |
@param lu: the LU on behalf of which we make the check
|
274 |
@param node: the node to check
|
275 |
@raise errors.OpPrereqError: if the node is not vm capable
|
276 |
|
277 |
"""
|
278 |
if not lu.cfg.GetNodeInfo(node).vm_capable: |
279 |
raise errors.OpPrereqError("Can't use non-vm_capable node %s" % node, |
280 |
errors.ECODE_STATE) |
281 |
|
282 |
|
283 |
def _CheckNodeHasOS(lu, node, os_name, force_variant): |
284 |
"""Ensure that a node supports a given OS.
|
285 |
|
286 |
@param lu: the LU on behalf of which we make the check
|
287 |
@param node: the node to check
|
288 |
@param os_name: the OS to query about
|
289 |
@param force_variant: whether to ignore variant errors
|
290 |
@raise errors.OpPrereqError: if the node is not supporting the OS
|
291 |
|
292 |
"""
|
293 |
result = lu.rpc.call_os_get(node, os_name) |
294 |
result.Raise("OS '%s' not in supported OS list for node %s" %
|
295 |
(os_name, node), |
296 |
prereq=True, ecode=errors.ECODE_INVAL)
|
297 |
if not force_variant: |
298 |
_CheckOSVariant(result.payload, os_name) |
299 |
|
300 |
|
301 |
def _CheckNodeHasSecondaryIP(lu, node, secondary_ip, prereq): |
302 |
"""Ensure that a node has the given secondary ip.
|
303 |
|
304 |
@type lu: L{LogicalUnit}
|
305 |
@param lu: the LU on behalf of which we make the check
|
306 |
@type node: string
|
307 |
@param node: the node to check
|
308 |
@type secondary_ip: string
|
309 |
@param secondary_ip: the ip to check
|
310 |
@type prereq: boolean
|
311 |
@param prereq: whether to throw a prerequisite or an execute error
|
312 |
@raise errors.OpPrereqError: if the node doesn't have the ip, and prereq=True
|
313 |
@raise errors.OpExecError: if the node doesn't have the ip, and prereq=False
|
314 |
|
315 |
"""
|
316 |
result = lu.rpc.call_node_has_ip_address(node, secondary_ip) |
317 |
result.Raise("Failure checking secondary ip on node %s" % node,
|
318 |
prereq=prereq, ecode=errors.ECODE_ENVIRON) |
319 |
if not result.payload: |
320 |
msg = ("Node claims it doesn't have the secondary ip you gave (%s),"
|
321 |
" please fix and re-run this command" % secondary_ip)
|
322 |
if prereq:
|
323 |
raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
|
324 |
else:
|
325 |
raise errors.OpExecError(msg)
|
326 |
|
327 |
|
328 |
def _GetClusterDomainSecret(): |
329 |
"""Reads the cluster domain secret.
|
330 |
|
331 |
"""
|
332 |
return utils.ReadOneLineFile(pathutils.CLUSTER_DOMAIN_SECRET_FILE,
|
333 |
strict=True)
|
334 |
|
335 |
|
336 |
def _CheckInstanceState(lu, instance, req_states, msg=None): |
337 |
"""Ensure that an instance is in one of the required states.
|
338 |
|
339 |
@param lu: the LU on behalf of which we make the check
|
340 |
@param instance: the instance to check
|
341 |
@param msg: if passed, should be a message to replace the default one
|
342 |
@raise errors.OpPrereqError: if the instance is not in the required state
|
343 |
|
344 |
"""
|
345 |
if msg is None: |
346 |
msg = ("can't use instance from outside %s states" %
|
347 |
utils.CommaJoin(req_states)) |
348 |
if instance.admin_state not in req_states: |
349 |
raise errors.OpPrereqError("Instance '%s' is marked to be %s, %s" % |
350 |
(instance.name, instance.admin_state, msg), |
351 |
errors.ECODE_STATE) |
352 |
|
353 |
if constants.ADMINST_UP not in req_states: |
354 |
pnode = instance.primary_node |
355 |
if not lu.cfg.GetNodeInfo(pnode).offline: |
356 |
ins_l = lu.rpc.call_instance_list([pnode], [instance.hypervisor])[pnode] |
357 |
ins_l.Raise("Can't contact node %s for instance information" % pnode,
|
358 |
prereq=True, ecode=errors.ECODE_ENVIRON)
|
359 |
if instance.name in ins_l.payload: |
360 |
raise errors.OpPrereqError("Instance %s is running, %s" % |
361 |
(instance.name, msg), errors.ECODE_STATE) |
362 |
else:
|
363 |
lu.LogWarning("Primary node offline, ignoring check that instance"
|
364 |
" is down")
|
365 |
|
366 |
|
367 |
def _ComputeIPolicyInstanceSpecViolation( |
368 |
ipolicy, instance_spec, disk_template, |
369 |
_compute_fn=_ComputeIPolicySpecViolation): |
370 |
"""Compute if instance specs meets the specs of ipolicy.
|
371 |
|
372 |
@type ipolicy: dict
|
373 |
@param ipolicy: The ipolicy to verify against
|
374 |
@param instance_spec: dict
|
375 |
@param instance_spec: The instance spec to verify
|
376 |
@type disk_template: string
|
377 |
@param disk_template: the disk template of the instance
|
378 |
@param _compute_fn: The function to verify ipolicy (unittest only)
|
379 |
@see: L{_ComputeIPolicySpecViolation}
|
380 |
|
381 |
"""
|
382 |
mem_size = instance_spec.get(constants.ISPEC_MEM_SIZE, None)
|
383 |
cpu_count = instance_spec.get(constants.ISPEC_CPU_COUNT, None)
|
384 |
disk_count = instance_spec.get(constants.ISPEC_DISK_COUNT, 0)
|
385 |
disk_sizes = instance_spec.get(constants.ISPEC_DISK_SIZE, []) |
386 |
nic_count = instance_spec.get(constants.ISPEC_NIC_COUNT, 0)
|
387 |
spindle_use = instance_spec.get(constants.ISPEC_SPINDLE_USE, None)
|
388 |
|
389 |
return _compute_fn(ipolicy, mem_size, cpu_count, disk_count, nic_count,
|
390 |
disk_sizes, spindle_use, disk_template) |
391 |
|
392 |
|
393 |
def _ComputeIPolicyNodeViolation(ipolicy, instance, current_group, |
394 |
target_group, cfg, |
395 |
_compute_fn=_ComputeIPolicyInstanceViolation): |
396 |
"""Compute if instance meets the specs of the new target group.
|
397 |
|
398 |
@param ipolicy: The ipolicy to verify
|
399 |
@param instance: The instance object to verify
|
400 |
@param current_group: The current group of the instance
|
401 |
@param target_group: The new group of the instance
|
402 |
@type cfg: L{config.ConfigWriter}
|
403 |
@param cfg: Cluster configuration
|
404 |
@param _compute_fn: The function to verify ipolicy (unittest only)
|
405 |
@see: L{_ComputeIPolicySpecViolation}
|
406 |
|
407 |
"""
|
408 |
if current_group == target_group:
|
409 |
return []
|
410 |
else:
|
411 |
return _compute_fn(ipolicy, instance, cfg)
|
412 |
|
413 |
|
414 |
def _CheckTargetNodeIPolicy(lu, ipolicy, instance, node, cfg, ignore=False, |
415 |
_compute_fn=_ComputeIPolicyNodeViolation): |
416 |
"""Checks that the target node is correct in terms of instance policy.
|
417 |
|
418 |
@param ipolicy: The ipolicy to verify
|
419 |
@param instance: The instance object to verify
|
420 |
@param node: The new node to relocate
|
421 |
@type cfg: L{config.ConfigWriter}
|
422 |
@param cfg: Cluster configuration
|
423 |
@param ignore: Ignore violations of the ipolicy
|
424 |
@param _compute_fn: The function to verify ipolicy (unittest only)
|
425 |
@see: L{_ComputeIPolicySpecViolation}
|
426 |
|
427 |
"""
|
428 |
primary_node = lu.cfg.GetNodeInfo(instance.primary_node) |
429 |
res = _compute_fn(ipolicy, instance, primary_node.group, node.group, cfg) |
430 |
|
431 |
if res:
|
432 |
msg = ("Instance does not meet target node group's (%s) instance"
|
433 |
" policy: %s") % (node.group, utils.CommaJoin(res))
|
434 |
if ignore:
|
435 |
lu.LogWarning(msg) |
436 |
else:
|
437 |
raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
|
438 |
|
439 |
|
440 |
def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status, |
441 |
minmem, maxmem, vcpus, nics, disk_template, disks, |
442 |
bep, hvp, hypervisor_name, tags): |
443 |
"""Builds instance related env variables for hooks
|
444 |
|
445 |
This builds the hook environment from individual variables.
|
446 |
|
447 |
@type name: string
|
448 |
@param name: the name of the instance
|
449 |
@type primary_node: string
|
450 |
@param primary_node: the name of the instance's primary node
|
451 |
@type secondary_nodes: list
|
452 |
@param secondary_nodes: list of secondary nodes as strings
|
453 |
@type os_type: string
|
454 |
@param os_type: the name of the instance's OS
|
455 |
@type status: string
|
456 |
@param status: the desired status of the instance
|
457 |
@type minmem: string
|
458 |
@param minmem: the minimum memory size of the instance
|
459 |
@type maxmem: string
|
460 |
@param maxmem: the maximum memory size of the instance
|
461 |
@type vcpus: string
|
462 |
@param vcpus: the count of VCPUs the instance has
|
463 |
@type nics: list
|
464 |
@param nics: list of tuples (name, uuid, ip, mac, mode, link, net, netinfo)
|
465 |
representing the NICs the instance has
|
466 |
@type disk_template: string
|
467 |
@param disk_template: the disk template of the instance
|
468 |
@type disks: list
|
469 |
@param disks: list of tuples (name, uuid, size, mode)
|
470 |
@type bep: dict
|
471 |
@param bep: the backend parameters for the instance
|
472 |
@type hvp: dict
|
473 |
@param hvp: the hypervisor parameters for the instance
|
474 |
@type hypervisor_name: string
|
475 |
@param hypervisor_name: the hypervisor for the instance
|
476 |
@type tags: list
|
477 |
@param tags: list of instance tags as strings
|
478 |
@rtype: dict
|
479 |
@return: the hook environment for this instance
|
480 |
|
481 |
"""
|
482 |
env = { |
483 |
"OP_TARGET": name,
|
484 |
"INSTANCE_NAME": name,
|
485 |
"INSTANCE_PRIMARY": primary_node,
|
486 |
"INSTANCE_SECONDARIES": " ".join(secondary_nodes), |
487 |
"INSTANCE_OS_TYPE": os_type,
|
488 |
"INSTANCE_STATUS": status,
|
489 |
"INSTANCE_MINMEM": minmem,
|
490 |
"INSTANCE_MAXMEM": maxmem,
|
491 |
# TODO(2.9) remove deprecated "memory" value
|
492 |
"INSTANCE_MEMORY": maxmem,
|
493 |
"INSTANCE_VCPUS": vcpus,
|
494 |
"INSTANCE_DISK_TEMPLATE": disk_template,
|
495 |
"INSTANCE_HYPERVISOR": hypervisor_name,
|
496 |
} |
497 |
if nics:
|
498 |
nic_count = len(nics)
|
499 |
for idx, (name, _, ip, mac, mode, link, net, netinfo) in enumerate(nics): |
500 |
if ip is None: |
501 |
ip = ""
|
502 |
env["INSTANCE_NIC%d_NAME" % idx] = name
|
503 |
env["INSTANCE_NIC%d_IP" % idx] = ip
|
504 |
env["INSTANCE_NIC%d_MAC" % idx] = mac
|
505 |
env["INSTANCE_NIC%d_MODE" % idx] = mode
|
506 |
env["INSTANCE_NIC%d_LINK" % idx] = link
|
507 |
if netinfo:
|
508 |
nobj = objects.Network.FromDict(netinfo) |
509 |
env.update(nobj.HooksDict("INSTANCE_NIC%d_" % idx))
|
510 |
elif network:
|
511 |
# FIXME: broken network reference: the instance NIC specifies a
|
512 |
# network, but the relevant network entry was not in the config. This
|
513 |
# should be made impossible.
|
514 |
env["INSTANCE_NIC%d_NETWORK_NAME" % idx] = net
|
515 |
if mode == constants.NIC_MODE_BRIDGED:
|
516 |
env["INSTANCE_NIC%d_BRIDGE" % idx] = link
|
517 |
else:
|
518 |
nic_count = 0
|
519 |
|
520 |
env["INSTANCE_NIC_COUNT"] = nic_count
|
521 |
|
522 |
if disks:
|
523 |
disk_count = len(disks)
|
524 |
for idx, (name, size, mode) in enumerate(disks): |
525 |
env["INSTANCE_DISK%d_NAME" % idx] = name
|
526 |
env["INSTANCE_DISK%d_SIZE" % idx] = size
|
527 |
env["INSTANCE_DISK%d_MODE" % idx] = mode
|
528 |
else:
|
529 |
disk_count = 0
|
530 |
|
531 |
env["INSTANCE_DISK_COUNT"] = disk_count
|
532 |
|
533 |
if not tags: |
534 |
tags = [] |
535 |
|
536 |
env["INSTANCE_TAGS"] = " ".join(tags) |
537 |
|
538 |
for source, kind in [(bep, "BE"), (hvp, "HV")]: |
539 |
for key, value in source.items(): |
540 |
env["INSTANCE_%s_%s" % (kind, key)] = value
|
541 |
|
542 |
return env
|
543 |
|
544 |
|
545 |
def _NICToTuple(lu, nic): |
546 |
"""Build a tupple of nic information.
|
547 |
|
548 |
@type lu: L{LogicalUnit}
|
549 |
@param lu: the logical unit on whose behalf we execute
|
550 |
@type nic: L{objects.NIC}
|
551 |
@param nic: nic to convert to hooks tuple
|
552 |
|
553 |
"""
|
554 |
cluster = lu.cfg.GetClusterInfo() |
555 |
filled_params = cluster.SimpleFillNIC(nic.nicparams) |
556 |
mode = filled_params[constants.NIC_MODE] |
557 |
link = filled_params[constants.NIC_LINK] |
558 |
netinfo = None
|
559 |
if nic.network:
|
560 |
nobj = lu.cfg.GetNetwork(nic.network) |
561 |
netinfo = objects.Network.ToDict(nobj) |
562 |
return (nic.name, nic.uuid, nic.ip, nic.mac, mode, link, nic.network, netinfo)
|
563 |
|
564 |
|
565 |
def _NICListToTuple(lu, nics): |
566 |
"""Build a list of nic information tuples.
|
567 |
|
568 |
This list is suitable to be passed to _BuildInstanceHookEnv or as a return
|
569 |
value in LUInstanceQueryData.
|
570 |
|
571 |
@type lu: L{LogicalUnit}
|
572 |
@param lu: the logical unit on whose behalf we execute
|
573 |
@type nics: list of L{objects.NIC}
|
574 |
@param nics: list of nics to convert to hooks tuples
|
575 |
|
576 |
"""
|
577 |
hooks_nics = [] |
578 |
for nic in nics: |
579 |
hooks_nics.append(_NICToTuple(lu, nic)) |
580 |
return hooks_nics
|
581 |
|
582 |
|
583 |
def _BuildInstanceHookEnvByObject(lu, instance, override=None): |
584 |
"""Builds instance related env variables for hooks from an object.
|
585 |
|
586 |
@type lu: L{LogicalUnit}
|
587 |
@param lu: the logical unit on whose behalf we execute
|
588 |
@type instance: L{objects.Instance}
|
589 |
@param instance: the instance for which we should build the
|
590 |
environment
|
591 |
@type override: dict
|
592 |
@param override: dictionary with key/values that will override
|
593 |
our values
|
594 |
@rtype: dict
|
595 |
@return: the hook environment dictionary
|
596 |
|
597 |
"""
|
598 |
cluster = lu.cfg.GetClusterInfo() |
599 |
bep = cluster.FillBE(instance) |
600 |
hvp = cluster.FillHV(instance) |
601 |
args = { |
602 |
"name": instance.name,
|
603 |
"primary_node": instance.primary_node,
|
604 |
"secondary_nodes": instance.secondary_nodes,
|
605 |
"os_type": instance.os,
|
606 |
"status": instance.admin_state,
|
607 |
"maxmem": bep[constants.BE_MAXMEM],
|
608 |
"minmem": bep[constants.BE_MINMEM],
|
609 |
"vcpus": bep[constants.BE_VCPUS],
|
610 |
"nics": _NICListToTuple(lu, instance.nics),
|
611 |
"disk_template": instance.disk_template,
|
612 |
"disks": [(disk.name, disk.size, disk.mode)
|
613 |
for disk in instance.disks], |
614 |
"bep": bep,
|
615 |
"hvp": hvp,
|
616 |
"hypervisor_name": instance.hypervisor,
|
617 |
"tags": instance.tags,
|
618 |
} |
619 |
if override:
|
620 |
args.update(override) |
621 |
return _BuildInstanceHookEnv(**args) # pylint: disable=W0142 |
622 |
|
623 |
|
624 |
def _DecideSelfPromotion(lu, exceptions=None): |
625 |
"""Decide whether I should promote myself as a master candidate.
|
626 |
|
627 |
"""
|
628 |
cp_size = lu.cfg.GetClusterInfo().candidate_pool_size |
629 |
mc_now, mc_should, _ = lu.cfg.GetMasterCandidateStats(exceptions) |
630 |
# the new node will increase mc_max with one, so:
|
631 |
mc_should = min(mc_should + 1, cp_size) |
632 |
return mc_now < mc_should
|
633 |
|
634 |
|
635 |
def _CheckNicsBridgesExist(lu, target_nics, target_node): |
636 |
"""Check that the brigdes needed by a list of nics exist.
|
637 |
|
638 |
"""
|
639 |
cluster = lu.cfg.GetClusterInfo() |
640 |
paramslist = [cluster.SimpleFillNIC(nic.nicparams) for nic in target_nics] |
641 |
brlist = [params[constants.NIC_LINK] for params in paramslist |
642 |
if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED]
|
643 |
if brlist:
|
644 |
result = lu.rpc.call_bridges_exist(target_node, brlist) |
645 |
result.Raise("Error checking bridges on destination node '%s'" %
|
646 |
target_node, prereq=True, ecode=errors.ECODE_ENVIRON)
|
647 |
|
648 |
|
649 |
def _CheckInstanceBridgesExist(lu, instance, node=None): |
650 |
"""Check that the brigdes needed by an instance exist.
|
651 |
|
652 |
"""
|
653 |
if node is None: |
654 |
node = instance.primary_node |
655 |
_CheckNicsBridgesExist(lu, instance.nics, node) |
656 |
|
657 |
|
658 |
def _CheckOSVariant(os_obj, name): |
659 |
"""Check whether an OS name conforms to the os variants specification.
|
660 |
|
661 |
@type os_obj: L{objects.OS}
|
662 |
@param os_obj: OS object to check
|
663 |
@type name: string
|
664 |
@param name: OS name passed by the user, to check for validity
|
665 |
|
666 |
"""
|
667 |
variant = objects.OS.GetVariant(name) |
668 |
if not os_obj.supported_variants: |
669 |
if variant:
|
670 |
raise errors.OpPrereqError("OS '%s' doesn't support variants ('%s'" |
671 |
" passed)" % (os_obj.name, variant),
|
672 |
errors.ECODE_INVAL) |
673 |
return
|
674 |
if not variant: |
675 |
raise errors.OpPrereqError("OS name must include a variant", |
676 |
errors.ECODE_INVAL) |
677 |
|
678 |
if variant not in os_obj.supported_variants: |
679 |
raise errors.OpPrereqError("Unsupported OS variant", errors.ECODE_INVAL) |
680 |
|
681 |
|
682 |
def _GetNodeInstancesInner(cfg, fn): |
683 |
return [i for i in cfg.GetAllInstancesInfo().values() if fn(i)] |
684 |
|
685 |
|
686 |
def _GetNodeInstances(cfg, node_name): |
687 |
"""Returns a list of all primary and secondary instances on a node.
|
688 |
|
689 |
"""
|
690 |
|
691 |
return _GetNodeInstancesInner(cfg, lambda inst: node_name in inst.all_nodes) |
692 |
|
693 |
|
694 |
def _GetNodePrimaryInstances(cfg, node_name): |
695 |
"""Returns primary instances on a node.
|
696 |
|
697 |
"""
|
698 |
return _GetNodeInstancesInner(cfg,
|
699 |
lambda inst: node_name == inst.primary_node)
|
700 |
|
701 |
|
702 |
def _GetNodeSecondaryInstances(cfg, node_name): |
703 |
"""Returns secondary instances on a node.
|
704 |
|
705 |
"""
|
706 |
return _GetNodeInstancesInner(cfg,
|
707 |
lambda inst: node_name in inst.secondary_nodes) |
708 |
|
709 |
|
710 |
def _GetStorageTypeArgs(cfg, storage_type): |
711 |
"""Returns the arguments for a storage type.
|
712 |
|
713 |
"""
|
714 |
# Special case for file storage
|
715 |
if storage_type == constants.ST_FILE:
|
716 |
# storage.FileStorage wants a list of storage directories
|
717 |
return [[cfg.GetFileStorageDir(), cfg.GetSharedFileStorageDir()]]
|
718 |
|
719 |
return []
|
720 |
|
721 |
|
722 |
def _FindFaultyInstanceDisks(cfg, rpc_runner, instance, node_name, prereq): |
723 |
faulty = [] |
724 |
|
725 |
for dev in instance.disks: |
726 |
cfg.SetDiskID(dev, node_name) |
727 |
|
728 |
result = rpc_runner.call_blockdev_getmirrorstatus(node_name, (instance.disks, |
729 |
instance)) |
730 |
result.Raise("Failed to get disk status from node %s" % node_name,
|
731 |
prereq=prereq, ecode=errors.ECODE_ENVIRON) |
732 |
|
733 |
for idx, bdev_status in enumerate(result.payload): |
734 |
if bdev_status and bdev_status.ldisk_status == constants.LDS_FAULTY: |
735 |
faulty.append(idx) |
736 |
|
737 |
return faulty
|
738 |
|
739 |
|
740 |
def _CheckIAllocatorOrNode(lu, iallocator_slot, node_slot): |
741 |
"""Check the sanity of iallocator and node arguments and use the
|
742 |
cluster-wide iallocator if appropriate.
|
743 |
|
744 |
Check that at most one of (iallocator, node) is specified. If none is
|
745 |
specified, or the iallocator is L{constants.DEFAULT_IALLOCATOR_SHORTCUT},
|
746 |
then the LU's opcode's iallocator slot is filled with the cluster-wide
|
747 |
default iallocator.
|
748 |
|
749 |
@type iallocator_slot: string
|
750 |
@param iallocator_slot: the name of the opcode iallocator slot
|
751 |
@type node_slot: string
|
752 |
@param node_slot: the name of the opcode target node slot
|
753 |
|
754 |
"""
|
755 |
node = getattr(lu.op, node_slot, None) |
756 |
ialloc = getattr(lu.op, iallocator_slot, None) |
757 |
if node == []:
|
758 |
node = None
|
759 |
|
760 |
if node is not None and ialloc is not None: |
761 |
raise errors.OpPrereqError("Do not specify both, iallocator and node", |
762 |
errors.ECODE_INVAL) |
763 |
elif ((node is None and ialloc is None) or |
764 |
ialloc == constants.DEFAULT_IALLOCATOR_SHORTCUT): |
765 |
default_iallocator = lu.cfg.GetDefaultIAllocator() |
766 |
if default_iallocator:
|
767 |
setattr(lu.op, iallocator_slot, default_iallocator)
|
768 |
else:
|
769 |
raise errors.OpPrereqError("No iallocator or node given and no" |
770 |
" cluster-wide default iallocator found;"
|
771 |
" please specify either an iallocator or a"
|
772 |
" node, or set a cluster-wide default"
|
773 |
" iallocator", errors.ECODE_INVAL)
|
774 |
|
775 |
|
776 |
def _CheckHostnameSane(lu, name): |
777 |
"""Ensures that a given hostname resolves to a 'sane' name.
|
778 |
|
779 |
The given name is required to be a prefix of the resolved hostname,
|
780 |
to prevent accidental mismatches.
|
781 |
|
782 |
@param lu: the logical unit on behalf of which we're checking
|
783 |
@param name: the name we should resolve and check
|
784 |
@return: the resolved hostname object
|
785 |
|
786 |
"""
|
787 |
hostname = netutils.GetHostname(name=name) |
788 |
if hostname.name != name:
|
789 |
lu.LogInfo("Resolved given name '%s' to '%s'", name, hostname.name)
|
790 |
if not utils.MatchNameComponent(name, [hostname.name]): |
791 |
raise errors.OpPrereqError(("Resolved hostname '%s' does not look the" |
792 |
" same as given hostname '%s'") %
|
793 |
(hostname.name, name), errors.ECODE_INVAL) |
794 |
return hostname
|
795 |
|
796 |
|
797 |
def _WaitForSync(lu, instance, disks=None, oneshot=False): |
798 |
"""Sleep and poll for an instance's disk to sync.
|
799 |
|
800 |
"""
|
801 |
if not instance.disks or disks is not None and not disks: |
802 |
return True |
803 |
|
804 |
disks = _ExpandCheckDisks(instance, disks) |
805 |
|
806 |
if not oneshot: |
807 |
lu.LogInfo("Waiting for instance %s to sync disks", instance.name)
|
808 |
|
809 |
node = instance.primary_node |
810 |
|
811 |
for dev in disks: |
812 |
lu.cfg.SetDiskID(dev, node) |
813 |
|
814 |
# TODO: Convert to utils.Retry
|
815 |
|
816 |
retries = 0
|
817 |
degr_retries = 10 # in seconds, as we sleep 1 second each time |
818 |
while True: |
819 |
max_time = 0
|
820 |
done = True
|
821 |
cumul_degraded = False
|
822 |
rstats = lu.rpc.call_blockdev_getmirrorstatus(node, (disks, instance)) |
823 |
msg = rstats.fail_msg |
824 |
if msg:
|
825 |
lu.LogWarning("Can't get any data from node %s: %s", node, msg)
|
826 |
retries += 1
|
827 |
if retries >= 10: |
828 |
raise errors.RemoteError("Can't contact node %s for mirror data," |
829 |
" aborting." % node)
|
830 |
time.sleep(6)
|
831 |
continue
|
832 |
rstats = rstats.payload |
833 |
retries = 0
|
834 |
for i, mstat in enumerate(rstats): |
835 |
if mstat is None: |
836 |
lu.LogWarning("Can't compute data for node %s/%s",
|
837 |
node, disks[i].iv_name) |
838 |
continue
|
839 |
|
840 |
cumul_degraded = (cumul_degraded or
|
841 |
(mstat.is_degraded and mstat.sync_percent is None)) |
842 |
if mstat.sync_percent is not None: |
843 |
done = False
|
844 |
if mstat.estimated_time is not None: |
845 |
rem_time = ("%s remaining (estimated)" %
|
846 |
utils.FormatSeconds(mstat.estimated_time)) |
847 |
max_time = mstat.estimated_time |
848 |
else:
|
849 |
rem_time = "no time estimate"
|
850 |
lu.LogInfo("- device %s: %5.2f%% done, %s",
|
851 |
disks[i].iv_name, mstat.sync_percent, rem_time) |
852 |
|
853 |
# if we're done but degraded, let's do a few small retries, to
|
854 |
# make sure we see a stable and not transient situation; therefore
|
855 |
# we force restart of the loop
|
856 |
if (done or oneshot) and cumul_degraded and degr_retries > 0: |
857 |
logging.info("Degraded disks found, %d retries left", degr_retries)
|
858 |
degr_retries -= 1
|
859 |
time.sleep(1)
|
860 |
continue
|
861 |
|
862 |
if done or oneshot: |
863 |
break
|
864 |
|
865 |
time.sleep(min(60, max_time)) |
866 |
|
867 |
if done:
|
868 |
lu.LogInfo("Instance %s's disks are in sync", instance.name)
|
869 |
|
870 |
return not cumul_degraded |
871 |
|
872 |
|
873 |
def _BlockdevFind(lu, node, dev, instance): |
874 |
"""Wrapper around call_blockdev_find to annotate diskparams.
|
875 |
|
876 |
@param lu: A reference to the lu object
|
877 |
@param node: The node to call out
|
878 |
@param dev: The device to find
|
879 |
@param instance: The instance object the device belongs to
|
880 |
@returns The result of the rpc call
|
881 |
|
882 |
"""
|
883 |
(disk,) = _AnnotateDiskParams(instance, [dev], lu.cfg) |
884 |
return lu.rpc.call_blockdev_find(node, disk)
|
885 |
|
886 |
|
887 |
def _CheckDiskConsistency(lu, instance, dev, node, on_primary, ldisk=False): |
888 |
"""Wrapper around L{_CheckDiskConsistencyInner}.
|
889 |
|
890 |
"""
|
891 |
(disk,) = _AnnotateDiskParams(instance, [dev], lu.cfg) |
892 |
return _CheckDiskConsistencyInner(lu, instance, disk, node, on_primary,
|
893 |
ldisk=ldisk) |
894 |
|
895 |
|
896 |
def _CheckDiskConsistencyInner(lu, instance, dev, node, on_primary, |
897 |
ldisk=False):
|
898 |
"""Check that mirrors are not degraded.
|
899 |
|
900 |
@attention: The device has to be annotated already.
|
901 |
|
902 |
The ldisk parameter, if True, will change the test from the
|
903 |
is_degraded attribute (which represents overall non-ok status for
|
904 |
the device(s)) to the ldisk (representing the local storage status).
|
905 |
|
906 |
"""
|
907 |
lu.cfg.SetDiskID(dev, node) |
908 |
|
909 |
result = True
|
910 |
|
911 |
if on_primary or dev.AssembleOnSecondary(): |
912 |
rstats = lu.rpc.call_blockdev_find(node, dev) |
913 |
msg = rstats.fail_msg |
914 |
if msg:
|
915 |
lu.LogWarning("Can't find disk on node %s: %s", node, msg)
|
916 |
result = False
|
917 |
elif not rstats.payload: |
918 |
lu.LogWarning("Can't find disk on node %s", node)
|
919 |
result = False
|
920 |
else:
|
921 |
if ldisk:
|
922 |
result = result and rstats.payload.ldisk_status == constants.LDS_OKAY
|
923 |
else:
|
924 |
result = result and not rstats.payload.is_degraded |
925 |
|
926 |
if dev.children:
|
927 |
for child in dev.children: |
928 |
result = result and _CheckDiskConsistencyInner(lu, instance, child, node,
|
929 |
on_primary) |
930 |
|
931 |
return result
|
932 |
|
933 |
|
934 |
class LUOobCommand(NoHooksLU): |
935 |
"""Logical unit for OOB handling.
|
936 |
|
937 |
"""
|
938 |
REQ_BGL = False
|
939 |
_SKIP_MASTER = (constants.OOB_POWER_OFF, constants.OOB_POWER_CYCLE) |
940 |
|
941 |
def ExpandNames(self): |
942 |
"""Gather locks we need.
|
943 |
|
944 |
"""
|
945 |
if self.op.node_names: |
946 |
self.op.node_names = _GetWantedNodes(self, self.op.node_names) |
947 |
lock_names = self.op.node_names
|
948 |
else:
|
949 |
lock_names = locking.ALL_SET |
950 |
|
951 |
self.needed_locks = {
|
952 |
locking.LEVEL_NODE: lock_names, |
953 |
} |
954 |
|
955 |
self.share_locks[locking.LEVEL_NODE_ALLOC] = 1 |
956 |
|
957 |
if not self.op.node_names: |
958 |
# Acquire node allocation lock only if all nodes are affected
|
959 |
self.needed_locks[locking.LEVEL_NODE_ALLOC] = locking.ALL_SET
|
960 |
|
961 |
def CheckPrereq(self): |
962 |
"""Check prerequisites.
|
963 |
|
964 |
This checks:
|
965 |
- the node exists in the configuration
|
966 |
- OOB is supported
|
967 |
|
968 |
Any errors are signaled by raising errors.OpPrereqError.
|
969 |
|
970 |
"""
|
971 |
self.nodes = []
|
972 |
self.master_node = self.cfg.GetMasterNode() |
973 |
|
974 |
assert self.op.power_delay >= 0.0 |
975 |
|
976 |
if self.op.node_names: |
977 |
if (self.op.command in self._SKIP_MASTER and |
978 |
self.master_node in self.op.node_names): |
979 |
master_node_obj = self.cfg.GetNodeInfo(self.master_node) |
980 |
master_oob_handler = _SupportsOob(self.cfg, master_node_obj)
|
981 |
|
982 |
if master_oob_handler:
|
983 |
additional_text = ("run '%s %s %s' if you want to operate on the"
|
984 |
" master regardless") % (master_oob_handler,
|
985 |
self.op.command,
|
986 |
self.master_node)
|
987 |
else:
|
988 |
additional_text = "it does not support out-of-band operations"
|
989 |
|
990 |
raise errors.OpPrereqError(("Operating on the master node %s is not" |
991 |
" allowed for %s; %s") %
|
992 |
(self.master_node, self.op.command, |
993 |
additional_text), errors.ECODE_INVAL) |
994 |
else:
|
995 |
self.op.node_names = self.cfg.GetNodeList() |
996 |
if self.op.command in self._SKIP_MASTER: |
997 |
self.op.node_names.remove(self.master_node) |
998 |
|
999 |
if self.op.command in self._SKIP_MASTER: |
1000 |
assert self.master_node not in self.op.node_names |
1001 |
|
1002 |
for (node_name, node) in self.cfg.GetMultiNodeInfo(self.op.node_names): |
1003 |
if node is None: |
1004 |
raise errors.OpPrereqError("Node %s not found" % node_name, |
1005 |
errors.ECODE_NOENT) |
1006 |
else:
|
1007 |
self.nodes.append(node)
|
1008 |
|
1009 |
if (not self.op.ignore_status and |
1010 |
(self.op.command == constants.OOB_POWER_OFF and not node.offline)): |
1011 |
raise errors.OpPrereqError(("Cannot power off node %s because it is" |
1012 |
" not marked offline") % node_name,
|
1013 |
errors.ECODE_STATE) |
1014 |
|
1015 |
def Exec(self, feedback_fn): |
1016 |
"""Execute OOB and return result if we expect any.
|
1017 |
|
1018 |
"""
|
1019 |
master_node = self.master_node
|
1020 |
ret = [] |
1021 |
|
1022 |
for idx, node in enumerate(utils.NiceSort(self.nodes, |
1023 |
key=lambda node: node.name)):
|
1024 |
node_entry = [(constants.RS_NORMAL, node.name)] |
1025 |
ret.append(node_entry) |
1026 |
|
1027 |
oob_program = _SupportsOob(self.cfg, node)
|
1028 |
|
1029 |
if not oob_program: |
1030 |
node_entry.append((constants.RS_UNAVAIL, None))
|
1031 |
continue
|
1032 |
|
1033 |
logging.info("Executing out-of-band command '%s' using '%s' on %s",
|
1034 |
self.op.command, oob_program, node.name)
|
1035 |
result = self.rpc.call_run_oob(master_node, oob_program,
|
1036 |
self.op.command, node.name,
|
1037 |
self.op.timeout)
|
1038 |
|
1039 |
if result.fail_msg:
|
1040 |
self.LogWarning("Out-of-band RPC failed on node '%s': %s", |
1041 |
node.name, result.fail_msg) |
1042 |
node_entry.append((constants.RS_NODATA, None))
|
1043 |
else:
|
1044 |
try:
|
1045 |
self._CheckPayload(result)
|
1046 |
except errors.OpExecError, err:
|
1047 |
self.LogWarning("Payload returned by node '%s' is not valid: %s", |
1048 |
node.name, err) |
1049 |
node_entry.append((constants.RS_NODATA, None))
|
1050 |
else:
|
1051 |
if self.op.command == constants.OOB_HEALTH: |
1052 |
# For health we should log important events
|
1053 |
for item, status in result.payload: |
1054 |
if status in [constants.OOB_STATUS_WARNING, |
1055 |
constants.OOB_STATUS_CRITICAL]: |
1056 |
self.LogWarning("Item '%s' on node '%s' has status '%s'", |
1057 |
item, node.name, status) |
1058 |
|
1059 |
if self.op.command == constants.OOB_POWER_ON: |
1060 |
node.powered = True
|
1061 |
elif self.op.command == constants.OOB_POWER_OFF: |
1062 |
node.powered = False
|
1063 |
elif self.op.command == constants.OOB_POWER_STATUS: |
1064 |
powered = result.payload[constants.OOB_POWER_STATUS_POWERED] |
1065 |
if powered != node.powered:
|
1066 |
logging.warning(("Recorded power state (%s) of node '%s' does not"
|
1067 |
" match actual power state (%s)"), node.powered,
|
1068 |
node.name, powered) |
1069 |
|
1070 |
# For configuration changing commands we should update the node
|
1071 |
if self.op.command in (constants.OOB_POWER_ON, |
1072 |
constants.OOB_POWER_OFF): |
1073 |
self.cfg.Update(node, feedback_fn)
|
1074 |
|
1075 |
node_entry.append((constants.RS_NORMAL, result.payload)) |
1076 |
|
1077 |
if (self.op.command == constants.OOB_POWER_ON and |
1078 |
idx < len(self.nodes) - 1): |
1079 |
time.sleep(self.op.power_delay)
|
1080 |
|
1081 |
return ret
|
1082 |
|
1083 |
def _CheckPayload(self, result): |
1084 |
"""Checks if the payload is valid.
|
1085 |
|
1086 |
@param result: RPC result
|
1087 |
@raises errors.OpExecError: If payload is not valid
|
1088 |
|
1089 |
"""
|
1090 |
errs = [] |
1091 |
if self.op.command == constants.OOB_HEALTH: |
1092 |
if not isinstance(result.payload, list): |
1093 |
errs.append("command 'health' is expected to return a list but got %s" %
|
1094 |
type(result.payload))
|
1095 |
else:
|
1096 |
for item, status in result.payload: |
1097 |
if status not in constants.OOB_STATUSES: |
1098 |
errs.append("health item '%s' has invalid status '%s'" %
|
1099 |
(item, status)) |
1100 |
|
1101 |
if self.op.command == constants.OOB_POWER_STATUS: |
1102 |
if not isinstance(result.payload, dict): |
1103 |
errs.append("power-status is expected to return a dict but got %s" %
|
1104 |
type(result.payload))
|
1105 |
|
1106 |
if self.op.command in [ |
1107 |
constants.OOB_POWER_ON, |
1108 |
constants.OOB_POWER_OFF, |
1109 |
constants.OOB_POWER_CYCLE, |
1110 |
]: |
1111 |
if result.payload is not None: |
1112 |
errs.append("%s is expected to not return payload but got '%s'" %
|
1113 |
(self.op.command, result.payload))
|
1114 |
|
1115 |
if errs:
|
1116 |
raise errors.OpExecError("Check of out-of-band payload failed due to %s" % |
1117 |
utils.CommaJoin(errs)) |
1118 |
|
1119 |
|
1120 |
class _OsQuery(_QueryBase): |
1121 |
FIELDS = query.OS_FIELDS |
1122 |
|
1123 |
def ExpandNames(self, lu): |
1124 |
# Lock all nodes in shared mode
|
1125 |
# Temporary removal of locks, should be reverted later
|
1126 |
# TODO: reintroduce locks when they are lighter-weight
|
1127 |
lu.needed_locks = {} |
1128 |
#self.share_locks[locking.LEVEL_NODE] = 1
|
1129 |
#self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
|
1130 |
|
1131 |
# The following variables interact with _QueryBase._GetNames
|
1132 |
if self.names: |
1133 |
self.wanted = self.names |
1134 |
else:
|
1135 |
self.wanted = locking.ALL_SET
|
1136 |
|
1137 |
self.do_locking = self.use_locking |
1138 |
|
1139 |
def DeclareLocks(self, lu, level): |
1140 |
pass
|
1141 |
|
1142 |
@staticmethod
|
1143 |
def _DiagnoseByOS(rlist): |
1144 |
"""Remaps a per-node return list into an a per-os per-node dictionary
|
1145 |
|
1146 |
@param rlist: a map with node names as keys and OS objects as values
|
1147 |
|
1148 |
@rtype: dict
|
1149 |
@return: a dictionary with osnames as keys and as value another
|
1150 |
map, with nodes as keys and tuples of (path, status, diagnose,
|
1151 |
variants, parameters, api_versions) as values, eg::
|
1152 |
|
1153 |
{"debian-etch": {"node1": [(/usr/lib/..., True, "", [], []),
|
1154 |
(/srv/..., False, "invalid api")],
|
1155 |
"node2": [(/srv/..., True, "", [], [])]}
|
1156 |
}
|
1157 |
|
1158 |
"""
|
1159 |
all_os = {} |
1160 |
# we build here the list of nodes that didn't fail the RPC (at RPC
|
1161 |
# level), so that nodes with a non-responding node daemon don't
|
1162 |
# make all OSes invalid
|
1163 |
good_nodes = [node_name for node_name in rlist |
1164 |
if not rlist[node_name].fail_msg] |
1165 |
for node_name, nr in rlist.items(): |
1166 |
if nr.fail_msg or not nr.payload: |
1167 |
continue
|
1168 |
for (name, path, status, diagnose, variants,
|
1169 |
params, api_versions) in nr.payload:
|
1170 |
if name not in all_os: |
1171 |
# build a list of nodes for this os containing empty lists
|
1172 |
# for each node in node_list
|
1173 |
all_os[name] = {} |
1174 |
for nname in good_nodes: |
1175 |
all_os[name][nname] = [] |
1176 |
# convert params from [name, help] to (name, help)
|
1177 |
params = [tuple(v) for v in params] |
1178 |
all_os[name][node_name].append((path, status, diagnose, |
1179 |
variants, params, api_versions)) |
1180 |
return all_os
|
1181 |
|
1182 |
def _GetQueryData(self, lu): |
1183 |
"""Computes the list of nodes and their attributes.
|
1184 |
|
1185 |
"""
|
1186 |
# Locking is not used
|
1187 |
assert not (compat.any(lu.glm.is_owned(level) |
1188 |
for level in locking.LEVELS |
1189 |
if level != locking.LEVEL_CLUSTER) or |
1190 |
self.do_locking or self.use_locking) |
1191 |
|
1192 |
valid_nodes = [node.name |
1193 |
for node in lu.cfg.GetAllNodesInfo().values() |
1194 |
if not node.offline and node.vm_capable] |
1195 |
pol = self._DiagnoseByOS(lu.rpc.call_os_diagnose(valid_nodes))
|
1196 |
cluster = lu.cfg.GetClusterInfo() |
1197 |
|
1198 |
data = {} |
1199 |
|
1200 |
for (os_name, os_data) in pol.items(): |
1201 |
info = query.OsInfo(name=os_name, valid=True, node_status=os_data,
|
1202 |
hidden=(os_name in cluster.hidden_os),
|
1203 |
blacklisted=(os_name in cluster.blacklisted_os))
|
1204 |
|
1205 |
variants = set()
|
1206 |
parameters = set()
|
1207 |
api_versions = set()
|
1208 |
|
1209 |
for idx, osl in enumerate(os_data.values()): |
1210 |
info.valid = bool(info.valid and osl and osl[0][1]) |
1211 |
if not info.valid: |
1212 |
break
|
1213 |
|
1214 |
(node_variants, node_params, node_api) = osl[0][3:6] |
1215 |
if idx == 0: |
1216 |
# First entry
|
1217 |
variants.update(node_variants) |
1218 |
parameters.update(node_params) |
1219 |
api_versions.update(node_api) |
1220 |
else:
|
1221 |
# Filter out inconsistent values
|
1222 |
variants.intersection_update(node_variants) |
1223 |
parameters.intersection_update(node_params) |
1224 |
api_versions.intersection_update(node_api) |
1225 |
|
1226 |
info.variants = list(variants)
|
1227 |
info.parameters = list(parameters)
|
1228 |
info.api_versions = list(api_versions)
|
1229 |
|
1230 |
data[os_name] = info |
1231 |
|
1232 |
# Prepare data in requested order
|
1233 |
return [data[name] for name in self._GetNames(lu, pol.keys(), None) |
1234 |
if name in data] |
1235 |
|
1236 |
|
1237 |
class LUOsDiagnose(NoHooksLU): |
1238 |
"""Logical unit for OS diagnose/query.
|
1239 |
|
1240 |
"""
|
1241 |
REQ_BGL = False
|
1242 |
|
1243 |
@staticmethod
|
1244 |
def _BuildFilter(fields, names): |
1245 |
"""Builds a filter for querying OSes.
|
1246 |
|
1247 |
"""
|
1248 |
name_filter = qlang.MakeSimpleFilter("name", names)
|
1249 |
|
1250 |
# Legacy behaviour: Hide hidden, blacklisted or invalid OSes if the
|
1251 |
# respective field is not requested
|
1252 |
status_filter = [[qlang.OP_NOT, [qlang.OP_TRUE, fname]] |
1253 |
for fname in ["hidden", "blacklisted"] |
1254 |
if fname not in fields] |
1255 |
if "valid" not in fields: |
1256 |
status_filter.append([qlang.OP_TRUE, "valid"])
|
1257 |
|
1258 |
if status_filter:
|
1259 |
status_filter.insert(0, qlang.OP_AND)
|
1260 |
else:
|
1261 |
status_filter = None
|
1262 |
|
1263 |
if name_filter and status_filter: |
1264 |
return [qlang.OP_AND, name_filter, status_filter]
|
1265 |
elif name_filter:
|
1266 |
return name_filter
|
1267 |
else:
|
1268 |
return status_filter
|
1269 |
|
1270 |
def CheckArguments(self): |
1271 |
self.oq = _OsQuery(self._BuildFilter(self.op.output_fields, self.op.names), |
1272 |
self.op.output_fields, False) |
1273 |
|
1274 |
def ExpandNames(self): |
1275 |
self.oq.ExpandNames(self) |
1276 |
|
1277 |
def Exec(self, feedback_fn): |
1278 |
return self.oq.OldStyleQuery(self) |
1279 |
|
1280 |
|
1281 |
class _ExtStorageQuery(_QueryBase): |
1282 |
FIELDS = query.EXTSTORAGE_FIELDS |
1283 |
|
1284 |
def ExpandNames(self, lu): |
1285 |
# Lock all nodes in shared mode
|
1286 |
# Temporary removal of locks, should be reverted later
|
1287 |
# TODO: reintroduce locks when they are lighter-weight
|
1288 |
lu.needed_locks = {} |
1289 |
#self.share_locks[locking.LEVEL_NODE] = 1
|
1290 |
#self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
|
1291 |
|
1292 |
# The following variables interact with _QueryBase._GetNames
|
1293 |
if self.names: |
1294 |
self.wanted = self.names |
1295 |
else:
|
1296 |
self.wanted = locking.ALL_SET
|
1297 |
|
1298 |
self.do_locking = self.use_locking |
1299 |
|
1300 |
def DeclareLocks(self, lu, level): |
1301 |
pass
|
1302 |
|
1303 |
@staticmethod
|
1304 |
def _DiagnoseByProvider(rlist): |
1305 |
"""Remaps a per-node return list into an a per-provider per-node dictionary
|
1306 |
|
1307 |
@param rlist: a map with node names as keys and ExtStorage objects as values
|
1308 |
|
1309 |
@rtype: dict
|
1310 |
@return: a dictionary with extstorage providers as keys and as
|
1311 |
value another map, with nodes as keys and tuples of
|
1312 |
(path, status, diagnose, parameters) as values, eg::
|
1313 |
|
1314 |
{"provider1": {"node1": [(/usr/lib/..., True, "", [])]
|
1315 |
"node2": [(/srv/..., False, "missing file")]
|
1316 |
"node3": [(/srv/..., True, "", [])]
|
1317 |
}
|
1318 |
|
1319 |
"""
|
1320 |
all_es = {} |
1321 |
# we build here the list of nodes that didn't fail the RPC (at RPC
|
1322 |
# level), so that nodes with a non-responding node daemon don't
|
1323 |
# make all OSes invalid
|
1324 |
good_nodes = [node_name for node_name in rlist |
1325 |
if not rlist[node_name].fail_msg] |
1326 |
for node_name, nr in rlist.items(): |
1327 |
if nr.fail_msg or not nr.payload: |
1328 |
continue
|
1329 |
for (name, path, status, diagnose, params) in nr.payload: |
1330 |
if name not in all_es: |
1331 |
# build a list of nodes for this os containing empty lists
|
1332 |
# for each node in node_list
|
1333 |
all_es[name] = {} |
1334 |
for nname in good_nodes: |
1335 |
all_es[name][nname] = [] |
1336 |
# convert params from [name, help] to (name, help)
|
1337 |
params = [tuple(v) for v in params] |
1338 |
all_es[name][node_name].append((path, status, diagnose, params)) |
1339 |
return all_es
|
1340 |
|
1341 |
def _GetQueryData(self, lu): |
1342 |
"""Computes the list of nodes and their attributes.
|
1343 |
|
1344 |
"""
|
1345 |
# Locking is not used
|
1346 |
assert not (compat.any(lu.glm.is_owned(level) |
1347 |
for level in locking.LEVELS |
1348 |
if level != locking.LEVEL_CLUSTER) or |
1349 |
self.do_locking or self.use_locking) |
1350 |
|
1351 |
valid_nodes = [node.name |
1352 |
for node in lu.cfg.GetAllNodesInfo().values() |
1353 |
if not node.offline and node.vm_capable] |
1354 |
pol = self._DiagnoseByProvider(lu.rpc.call_extstorage_diagnose(valid_nodes))
|
1355 |
|
1356 |
data = {} |
1357 |
|
1358 |
nodegroup_list = lu.cfg.GetNodeGroupList() |
1359 |
|
1360 |
for (es_name, es_data) in pol.items(): |
1361 |
# For every provider compute the nodegroup validity.
|
1362 |
# To do this we need to check the validity of each node in es_data
|
1363 |
# and then construct the corresponding nodegroup dict:
|
1364 |
# { nodegroup1: status
|
1365 |
# nodegroup2: status
|
1366 |
# }
|
1367 |
ndgrp_data = {} |
1368 |
for nodegroup in nodegroup_list: |
1369 |
ndgrp = lu.cfg.GetNodeGroup(nodegroup) |
1370 |
|
1371 |
nodegroup_nodes = ndgrp.members |
1372 |
nodegroup_name = ndgrp.name |
1373 |
node_statuses = [] |
1374 |
|
1375 |
for node in nodegroup_nodes: |
1376 |
if node in valid_nodes: |
1377 |
if es_data[node] != []:
|
1378 |
node_status = es_data[node][0][1] |
1379 |
node_statuses.append(node_status) |
1380 |
else:
|
1381 |
node_statuses.append(False)
|
1382 |
|
1383 |
if False in node_statuses: |
1384 |
ndgrp_data[nodegroup_name] = False
|
1385 |
else:
|
1386 |
ndgrp_data[nodegroup_name] = True
|
1387 |
|
1388 |
# Compute the provider's parameters
|
1389 |
parameters = set()
|
1390 |
for idx, esl in enumerate(es_data.values()): |
1391 |
valid = bool(esl and esl[0][1]) |
1392 |
if not valid: |
1393 |
break
|
1394 |
|
1395 |
node_params = esl[0][3] |
1396 |
if idx == 0: |
1397 |
# First entry
|
1398 |
parameters.update(node_params) |
1399 |
else:
|
1400 |
# Filter out inconsistent values
|
1401 |
parameters.intersection_update(node_params) |
1402 |
|
1403 |
params = list(parameters)
|
1404 |
|
1405 |
# Now fill all the info for this provider
|
1406 |
info = query.ExtStorageInfo(name=es_name, node_status=es_data, |
1407 |
nodegroup_status=ndgrp_data, |
1408 |
parameters=params) |
1409 |
|
1410 |
data[es_name] = info |
1411 |
|
1412 |
# Prepare data in requested order
|
1413 |
return [data[name] for name in self._GetNames(lu, pol.keys(), None) |
1414 |
if name in data] |
1415 |
|
1416 |
|
1417 |
class LUExtStorageDiagnose(NoHooksLU): |
1418 |
"""Logical unit for ExtStorage diagnose/query.
|
1419 |
|
1420 |
"""
|
1421 |
REQ_BGL = False
|
1422 |
|
1423 |
def CheckArguments(self): |
1424 |
self.eq = _ExtStorageQuery(qlang.MakeSimpleFilter("name", self.op.names), |
1425 |
self.op.output_fields, False) |
1426 |
|
1427 |
def ExpandNames(self): |
1428 |
self.eq.ExpandNames(self) |
1429 |
|
1430 |
def Exec(self, feedback_fn): |
1431 |
return self.eq.OldStyleQuery(self) |
1432 |
|
1433 |
|
1434 |
class LUNodeRemove(LogicalUnit): |
1435 |
"""Logical unit for removing a node.
|
1436 |
|
1437 |
"""
|
1438 |
HPATH = "node-remove"
|
1439 |
HTYPE = constants.HTYPE_NODE |
1440 |
|
1441 |
def BuildHooksEnv(self): |
1442 |
"""Build hooks env.
|
1443 |
|
1444 |
"""
|
1445 |
return {
|
1446 |
"OP_TARGET": self.op.node_name, |
1447 |
"NODE_NAME": self.op.node_name, |
1448 |
} |
1449 |
|
1450 |
def BuildHooksNodes(self): |
1451 |
"""Build hooks nodes.
|
1452 |
|
1453 |
This doesn't run on the target node in the pre phase as a failed
|
1454 |
node would then be impossible to remove.
|
1455 |
|
1456 |
"""
|
1457 |
all_nodes = self.cfg.GetNodeList()
|
1458 |
try:
|
1459 |
all_nodes.remove(self.op.node_name)
|
1460 |
except ValueError: |
1461 |
pass
|
1462 |
return (all_nodes, all_nodes)
|
1463 |
|
1464 |
def CheckPrereq(self): |
1465 |
"""Check prerequisites.
|
1466 |
|
1467 |
This checks:
|
1468 |
- the node exists in the configuration
|
1469 |
- it does not have primary or secondary instances
|
1470 |
- it's not the master
|
1471 |
|
1472 |
Any errors are signaled by raising errors.OpPrereqError.
|
1473 |
|
1474 |
"""
|
1475 |
self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name) |
1476 |
node = self.cfg.GetNodeInfo(self.op.node_name) |
1477 |
assert node is not None |
1478 |
|
1479 |
masternode = self.cfg.GetMasterNode()
|
1480 |
if node.name == masternode:
|
1481 |
raise errors.OpPrereqError("Node is the master node, failover to another" |
1482 |
" node is required", errors.ECODE_INVAL)
|
1483 |
|
1484 |
for instance_name, instance in self.cfg.GetAllInstancesInfo().items(): |
1485 |
if node.name in instance.all_nodes: |
1486 |
raise errors.OpPrereqError("Instance %s is still running on the node," |
1487 |
" please remove first" % instance_name,
|
1488 |
errors.ECODE_INVAL) |
1489 |
self.op.node_name = node.name
|
1490 |
self.node = node
|
1491 |
|
1492 |
def Exec(self, feedback_fn): |
1493 |
"""Removes the node from the cluster.
|
1494 |
|
1495 |
"""
|
1496 |
node = self.node
|
1497 |
logging.info("Stopping the node daemon and removing configs from node %s",
|
1498 |
node.name) |
1499 |
|
1500 |
modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
|
1501 |
|
1502 |
assert locking.BGL in self.owned_locks(locking.LEVEL_CLUSTER), \ |
1503 |
"Not owning BGL"
|
1504 |
|
1505 |
# Promote nodes to master candidate as needed
|
1506 |
_AdjustCandidatePool(self, exceptions=[node.name])
|
1507 |
self.context.RemoveNode(node.name)
|
1508 |
|
1509 |
# Run post hooks on the node before it's removed
|
1510 |
_RunPostHook(self, node.name)
|
1511 |
|
1512 |
result = self.rpc.call_node_leave_cluster(node.name, modify_ssh_setup)
|
1513 |
msg = result.fail_msg |
1514 |
if msg:
|
1515 |
self.LogWarning("Errors encountered on the remote node while leaving" |
1516 |
" the cluster: %s", msg)
|
1517 |
|
1518 |
# Remove node from our /etc/hosts
|
1519 |
if self.cfg.GetClusterInfo().modify_etc_hosts: |
1520 |
master_node = self.cfg.GetMasterNode()
|
1521 |
result = self.rpc.call_etc_hosts_modify(master_node,
|
1522 |
constants.ETC_HOSTS_REMOVE, |
1523 |
node.name, None)
|
1524 |
result.Raise("Can't update hosts file with new host data")
|
1525 |
_RedistributeAncillaryFiles(self)
|
1526 |
|
1527 |
|
1528 |
class _NodeQuery(_QueryBase): |
1529 |
FIELDS = query.NODE_FIELDS |
1530 |
|
1531 |
def ExpandNames(self, lu): |
1532 |
lu.needed_locks = {} |
1533 |
lu.share_locks = _ShareAll() |
1534 |
|
1535 |
if self.names: |
1536 |
self.wanted = _GetWantedNodes(lu, self.names) |
1537 |
else:
|
1538 |
self.wanted = locking.ALL_SET
|
1539 |
|
1540 |
self.do_locking = (self.use_locking and |
1541 |
query.NQ_LIVE in self.requested_data) |
1542 |
|
1543 |
if self.do_locking: |
1544 |
# If any non-static field is requested we need to lock the nodes
|
1545 |
lu.needed_locks[locking.LEVEL_NODE] = self.wanted
|
1546 |
lu.needed_locks[locking.LEVEL_NODE_ALLOC] = locking.ALL_SET |
1547 |
|
1548 |
def DeclareLocks(self, lu, level): |
1549 |
pass
|
1550 |
|
1551 |
def _GetQueryData(self, lu): |
1552 |
"""Computes the list of nodes and their attributes.
|
1553 |
|
1554 |
"""
|
1555 |
all_info = lu.cfg.GetAllNodesInfo() |
1556 |
|
1557 |
nodenames = self._GetNames(lu, all_info.keys(), locking.LEVEL_NODE)
|
1558 |
|
1559 |
# Gather data as requested
|
1560 |
if query.NQ_LIVE in self.requested_data: |
1561 |
# filter out non-vm_capable nodes
|
1562 |
toquery_nodes = [name for name in nodenames if all_info[name].vm_capable] |
1563 |
|
1564 |
es_flags = rpc.GetExclusiveStorageForNodeNames(lu.cfg, toquery_nodes) |
1565 |
node_data = lu.rpc.call_node_info(toquery_nodes, [lu.cfg.GetVGName()], |
1566 |
[lu.cfg.GetHypervisorType()], es_flags) |
1567 |
live_data = dict((name, rpc.MakeLegacyNodeInfo(nresult.payload))
|
1568 |
for (name, nresult) in node_data.items() |
1569 |
if not nresult.fail_msg and nresult.payload) |
1570 |
else:
|
1571 |
live_data = None
|
1572 |
|
1573 |
if query.NQ_INST in self.requested_data: |
1574 |
node_to_primary = dict([(name, set()) for name in nodenames]) |
1575 |
node_to_secondary = dict([(name, set()) for name in nodenames]) |
1576 |
|
1577 |
inst_data = lu.cfg.GetAllInstancesInfo() |
1578 |
|
1579 |
for inst in inst_data.values(): |
1580 |
if inst.primary_node in node_to_primary: |
1581 |
node_to_primary[inst.primary_node].add(inst.name) |
1582 |
for secnode in inst.secondary_nodes: |
1583 |
if secnode in node_to_secondary: |
1584 |
node_to_secondary[secnode].add(inst.name) |
1585 |
else:
|
1586 |
node_to_primary = None
|
1587 |
node_to_secondary = None
|
1588 |
|
1589 |
if query.NQ_OOB in self.requested_data: |
1590 |
oob_support = dict((name, bool(_SupportsOob(lu.cfg, node))) |
1591 |
for name, node in all_info.iteritems()) |
1592 |
else:
|
1593 |
oob_support = None
|
1594 |
|
1595 |
if query.NQ_GROUP in self.requested_data: |
1596 |
groups = lu.cfg.GetAllNodeGroupsInfo() |
1597 |
else:
|
1598 |
groups = {} |
1599 |
|
1600 |
return query.NodeQueryData([all_info[name] for name in nodenames], |
1601 |
live_data, lu.cfg.GetMasterNode(), |
1602 |
node_to_primary, node_to_secondary, groups, |
1603 |
oob_support, lu.cfg.GetClusterInfo()) |
1604 |
|
1605 |
|
1606 |
class LUNodeQuery(NoHooksLU): |
1607 |
"""Logical unit for querying nodes.
|
1608 |
|
1609 |
"""
|
1610 |
# pylint: disable=W0142
|
1611 |
REQ_BGL = False
|
1612 |
|
1613 |
def CheckArguments(self): |
1614 |
self.nq = _NodeQuery(qlang.MakeSimpleFilter("name", self.op.names), |
1615 |
self.op.output_fields, self.op.use_locking) |
1616 |
|
1617 |
def ExpandNames(self): |
1618 |
self.nq.ExpandNames(self) |
1619 |
|
1620 |
def DeclareLocks(self, level): |
1621 |
self.nq.DeclareLocks(self, level) |
1622 |
|
1623 |
def Exec(self, feedback_fn): |
1624 |
return self.nq.OldStyleQuery(self) |
1625 |
|
1626 |
|
1627 |
class LUNodeQueryvols(NoHooksLU): |
1628 |
"""Logical unit for getting volumes on node(s).
|
1629 |
|
1630 |
"""
|
1631 |
REQ_BGL = False
|
1632 |
_FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance") |
1633 |
_FIELDS_STATIC = utils.FieldSet("node")
|
1634 |
|
1635 |
def CheckArguments(self): |
1636 |
_CheckOutputFields(static=self._FIELDS_STATIC,
|
1637 |
dynamic=self._FIELDS_DYNAMIC,
|
1638 |
selected=self.op.output_fields)
|
1639 |
|
1640 |
def ExpandNames(self): |
1641 |
self.share_locks = _ShareAll()
|
1642 |
|
1643 |
if self.op.nodes: |
1644 |
self.needed_locks = {
|
1645 |
locking.LEVEL_NODE: _GetWantedNodes(self, self.op.nodes), |
1646 |
} |
1647 |
else:
|
1648 |
self.needed_locks = {
|
1649 |
locking.LEVEL_NODE: locking.ALL_SET, |
1650 |
locking.LEVEL_NODE_ALLOC: locking.ALL_SET, |
1651 |
} |
1652 |
|
1653 |
def Exec(self, feedback_fn): |
1654 |
"""Computes the list of nodes and their attributes.
|
1655 |
|
1656 |
"""
|
1657 |
nodenames = self.owned_locks(locking.LEVEL_NODE)
|
1658 |
volumes = self.rpc.call_node_volumes(nodenames)
|
1659 |
|
1660 |
ilist = self.cfg.GetAllInstancesInfo()
|
1661 |
vol2inst = _MapInstanceDisksToNodes(ilist.values()) |
1662 |
|
1663 |
output = [] |
1664 |
for node in nodenames: |
1665 |
nresult = volumes[node] |
1666 |
if nresult.offline:
|
1667 |
continue
|
1668 |
msg = nresult.fail_msg |
1669 |
if msg:
|
1670 |
self.LogWarning("Can't compute volume data on node %s: %s", node, msg) |
1671 |
continue
|
1672 |
|
1673 |
node_vols = sorted(nresult.payload,
|
1674 |
key=operator.itemgetter("dev"))
|
1675 |
|
1676 |
for vol in node_vols: |
1677 |
node_output = [] |
1678 |
for field in self.op.output_fields: |
1679 |
if field == "node": |
1680 |
val = node |
1681 |
elif field == "phys": |
1682 |
val = vol["dev"]
|
1683 |
elif field == "vg": |
1684 |
val = vol["vg"]
|
1685 |
elif field == "name": |
1686 |
val = vol["name"]
|
1687 |
elif field == "size": |
1688 |
val = int(float(vol["size"])) |
1689 |
elif field == "instance": |
1690 |
val = vol2inst.get((node, vol["vg"] + "/" + vol["name"]), "-") |
1691 |
else:
|
1692 |
raise errors.ParameterError(field)
|
1693 |
node_output.append(str(val))
|
1694 |
|
1695 |
output.append(node_output) |
1696 |
|
1697 |
return output
|
1698 |
|
1699 |
|
1700 |
class LUNodeQueryStorage(NoHooksLU): |
1701 |
"""Logical unit for getting information on storage units on node(s).
|
1702 |
|
1703 |
"""
|
1704 |
_FIELDS_STATIC = utils.FieldSet(constants.SF_NODE) |
1705 |
REQ_BGL = False
|
1706 |
|
1707 |
def CheckArguments(self): |
1708 |
_CheckOutputFields(static=self._FIELDS_STATIC,
|
1709 |
dynamic=utils.FieldSet(*constants.VALID_STORAGE_FIELDS), |
1710 |
selected=self.op.output_fields)
|
1711 |
|
1712 |
def ExpandNames(self): |
1713 |
self.share_locks = _ShareAll()
|
1714 |
|
1715 |
if self.op.nodes: |
1716 |
self.needed_locks = {
|
1717 |
locking.LEVEL_NODE: _GetWantedNodes(self, self.op.nodes), |
1718 |
} |
1719 |
else:
|
1720 |
self.needed_locks = {
|
1721 |
locking.LEVEL_NODE: locking.ALL_SET, |
1722 |
locking.LEVEL_NODE_ALLOC: locking.ALL_SET, |
1723 |
} |
1724 |
|
1725 |
def Exec(self, feedback_fn): |
1726 |
"""Computes the list of nodes and their attributes.
|
1727 |
|
1728 |
"""
|
1729 |
self.nodes = self.owned_locks(locking.LEVEL_NODE) |
1730 |
|
1731 |
# Always get name to sort by
|
1732 |
if constants.SF_NAME in self.op.output_fields: |
1733 |
fields = self.op.output_fields[:]
|
1734 |
else:
|
1735 |
fields = [constants.SF_NAME] + self.op.output_fields
|
1736 |
|
1737 |
# Never ask for node or type as it's only known to the LU
|
1738 |
for extra in [constants.SF_NODE, constants.SF_TYPE]: |
1739 |
while extra in fields: |
1740 |
fields.remove(extra) |
1741 |
|
1742 |
field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)]) |
1743 |
name_idx = field_idx[constants.SF_NAME] |
1744 |
|
1745 |
st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type) |
1746 |
data = self.rpc.call_storage_list(self.nodes, |
1747 |
self.op.storage_type, st_args,
|
1748 |
self.op.name, fields)
|
1749 |
|
1750 |
result = [] |
1751 |
|
1752 |
for node in utils.NiceSort(self.nodes): |
1753 |
nresult = data[node] |
1754 |
if nresult.offline:
|
1755 |
continue
|
1756 |
|
1757 |
msg = nresult.fail_msg |
1758 |
if msg:
|
1759 |
self.LogWarning("Can't get storage data from node %s: %s", node, msg) |
1760 |
continue
|
1761 |
|
1762 |
rows = dict([(row[name_idx], row) for row in nresult.payload]) |
1763 |
|
1764 |
for name in utils.NiceSort(rows.keys()): |
1765 |
row = rows[name] |
1766 |
|
1767 |
out = [] |
1768 |
|
1769 |
for field in self.op.output_fields: |
1770 |
if field == constants.SF_NODE:
|
1771 |
val = node |
1772 |
elif field == constants.SF_TYPE:
|
1773 |
val = self.op.storage_type
|
1774 |
elif field in field_idx: |
1775 |
val = row[field_idx[field]] |
1776 |
else:
|
1777 |
raise errors.ParameterError(field)
|
1778 |
|
1779 |
out.append(val) |
1780 |
|
1781 |
result.append(out) |
1782 |
|
1783 |
return result
|
1784 |
|
1785 |
|
1786 |
class _InstanceQuery(_QueryBase): |
1787 |
FIELDS = query.INSTANCE_FIELDS |
1788 |
|
1789 |
def ExpandNames(self, lu): |
1790 |
lu.needed_locks = {} |
1791 |
lu.share_locks = _ShareAll() |
1792 |
|
1793 |
if self.names: |
1794 |
self.wanted = _GetWantedInstances(lu, self.names) |
1795 |
else:
|
1796 |
self.wanted = locking.ALL_SET
|
1797 |
|
1798 |
self.do_locking = (self.use_locking and |
1799 |
query.IQ_LIVE in self.requested_data) |
1800 |
if self.do_locking: |
1801 |
lu.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
|
1802 |
lu.needed_locks[locking.LEVEL_NODEGROUP] = [] |
1803 |
lu.needed_locks[locking.LEVEL_NODE] = [] |
1804 |
lu.needed_locks[locking.LEVEL_NETWORK] = [] |
1805 |
lu.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE |
1806 |
|
1807 |
self.do_grouplocks = (self.do_locking and |
1808 |
query.IQ_NODES in self.requested_data) |
1809 |
|
1810 |
def DeclareLocks(self, lu, level): |
1811 |
if self.do_locking: |
1812 |
if level == locking.LEVEL_NODEGROUP and self.do_grouplocks: |
1813 |
assert not lu.needed_locks[locking.LEVEL_NODEGROUP] |
1814 |
|
1815 |
# Lock all groups used by instances optimistically; this requires going
|
1816 |
# via the node before it's locked, requiring verification later on
|
1817 |
lu.needed_locks[locking.LEVEL_NODEGROUP] = \ |
1818 |
set(group_uuid
|
1819 |
for instance_name in lu.owned_locks(locking.LEVEL_INSTANCE) |
1820 |
for group_uuid in lu.cfg.GetInstanceNodeGroups(instance_name)) |
1821 |
elif level == locking.LEVEL_NODE:
|
1822 |
lu._LockInstancesNodes() # pylint: disable=W0212
|
1823 |
|
1824 |
elif level == locking.LEVEL_NETWORK:
|
1825 |
lu.needed_locks[locking.LEVEL_NETWORK] = \ |
1826 |
frozenset(net_uuid
|
1827 |
for instance_name in lu.owned_locks(locking.LEVEL_INSTANCE) |
1828 |
for net_uuid in lu.cfg.GetInstanceNetworks(instance_name)) |
1829 |
|
1830 |
@staticmethod
|
1831 |
def _CheckGroupLocks(lu): |
1832 |
owned_instances = frozenset(lu.owned_locks(locking.LEVEL_INSTANCE))
|
1833 |
owned_groups = frozenset(lu.owned_locks(locking.LEVEL_NODEGROUP))
|
1834 |
|
1835 |
# Check if node groups for locked instances are still correct
|
1836 |
for instance_name in owned_instances: |
1837 |
_CheckInstanceNodeGroups(lu.cfg, instance_name, owned_groups) |
1838 |
|
1839 |
def _GetQueryData(self, lu): |
1840 |
"""Computes the list of instances and their attributes.
|
1841 |
|
1842 |
"""
|
1843 |
if self.do_grouplocks: |
1844 |
self._CheckGroupLocks(lu)
|
1845 |
|
1846 |
cluster = lu.cfg.GetClusterInfo() |
1847 |
all_info = lu.cfg.GetAllInstancesInfo() |
1848 |
|
1849 |
instance_names = self._GetNames(lu, all_info.keys(), locking.LEVEL_INSTANCE)
|
1850 |
|
1851 |
instance_list = [all_info[name] for name in instance_names] |
1852 |
nodes = frozenset(itertools.chain(*(inst.all_nodes
|
1853 |
for inst in instance_list))) |
1854 |
hv_list = list(set([inst.hypervisor for inst in instance_list])) |
1855 |
bad_nodes = [] |
1856 |
offline_nodes = [] |
1857 |
wrongnode_inst = set()
|
1858 |
|
1859 |
# Gather data as requested
|
1860 |
if self.requested_data & set([query.IQ_LIVE, query.IQ_CONSOLE]): |
1861 |
live_data = {} |
1862 |
node_data = lu.rpc.call_all_instances_info(nodes, hv_list) |
1863 |
for name in nodes: |
1864 |
result = node_data[name] |
1865 |
if result.offline:
|
1866 |
# offline nodes will be in both lists
|
1867 |
assert result.fail_msg
|
1868 |
offline_nodes.append(name) |
1869 |
if result.fail_msg:
|
1870 |
bad_nodes.append(name) |
1871 |
elif result.payload:
|
1872 |
for inst in result.payload: |
1873 |
if inst in all_info: |
1874 |
if all_info[inst].primary_node == name:
|
1875 |
live_data.update(result.payload) |
1876 |
else:
|
1877 |
wrongnode_inst.add(inst) |
1878 |
else:
|
1879 |
# orphan instance; we don't list it here as we don't
|
1880 |
# handle this case yet in the output of instance listing
|
1881 |
logging.warning("Orphan instance '%s' found on node %s",
|
1882 |
inst, name) |
1883 |
# else no instance is alive
|
1884 |
else:
|
1885 |
live_data = {} |
1886 |
|
1887 |
if query.IQ_DISKUSAGE in self.requested_data: |
1888 |
gmi = ganeti.masterd.instance |
1889 |
disk_usage = dict((inst.name,
|
1890 |
gmi.ComputeDiskSize(inst.disk_template, |
1891 |
[{constants.IDISK_SIZE: disk.size} |
1892 |
for disk in inst.disks])) |
1893 |
for inst in instance_list) |
1894 |
else:
|
1895 |
disk_usage = None
|
1896 |
|
1897 |
if query.IQ_CONSOLE in self.requested_data: |
1898 |
consinfo = {} |
1899 |
for inst in instance_list: |
1900 |
if inst.name in live_data: |
1901 |
# Instance is running
|
1902 |
consinfo[inst.name] = _GetInstanceConsole(cluster, inst) |
1903 |
else:
|
1904 |
consinfo[inst.name] = None
|
1905 |
assert set(consinfo.keys()) == set(instance_names) |
1906 |
else:
|
1907 |
consinfo = None
|
1908 |
|
1909 |
if query.IQ_NODES in self.requested_data: |
1910 |
node_names = set(itertools.chain(*map(operator.attrgetter("all_nodes"), |
1911 |
instance_list))) |
1912 |
nodes = dict(lu.cfg.GetMultiNodeInfo(node_names))
|
1913 |
groups = dict((uuid, lu.cfg.GetNodeGroup(uuid))
|
1914 |
for uuid in set(map(operator.attrgetter("group"), |
1915 |
nodes.values()))) |
1916 |
else:
|
1917 |
nodes = None
|
1918 |
groups = None
|
1919 |
|
1920 |
if query.IQ_NETWORKS in self.requested_data: |
1921 |
net_uuids = itertools.chain(*(lu.cfg.GetInstanceNetworks(i.name) |
1922 |
for i in instance_list)) |
1923 |
networks = dict((uuid, lu.cfg.GetNetwork(uuid)) for uuid in net_uuids) |
1924 |
else:
|
1925 |
networks = None
|
1926 |
|
1927 |
return query.InstanceQueryData(instance_list, lu.cfg.GetClusterInfo(),
|
1928 |
disk_usage, offline_nodes, bad_nodes, |
1929 |
live_data, wrongnode_inst, consinfo, |
1930 |
nodes, groups, networks) |
1931 |
|
1932 |
|
1933 |
class LUQuery(NoHooksLU): |
1934 |
"""Query for resources/items of a certain kind.
|
1935 |
|
1936 |
"""
|
1937 |
# pylint: disable=W0142
|
1938 |
REQ_BGL = False
|
1939 |
|
1940 |
def CheckArguments(self): |
1941 |
qcls = _GetQueryImplementation(self.op.what)
|
1942 |
|
1943 |
self.impl = qcls(self.op.qfilter, self.op.fields, self.op.use_locking) |
1944 |
|
1945 |
def ExpandNames(self): |
1946 |
self.impl.ExpandNames(self) |
1947 |
|
1948 |
def DeclareLocks(self, level): |
1949 |
self.impl.DeclareLocks(self, level) |
1950 |
|
1951 |
def Exec(self, feedback_fn): |
1952 |
return self.impl.NewStyleQuery(self) |
1953 |
|
1954 |
|
1955 |
class LUQueryFields(NoHooksLU): |
1956 |
"""Query for resources/items of a certain kind.
|
1957 |
|
1958 |
"""
|
1959 |
# pylint: disable=W0142
|
1960 |
REQ_BGL = False
|
1961 |
|
1962 |
def CheckArguments(self): |
1963 |
self.qcls = _GetQueryImplementation(self.op.what) |
1964 |
|
1965 |
def ExpandNames(self): |
1966 |
self.needed_locks = {}
|
1967 |
|
1968 |
def Exec(self, feedback_fn): |
1969 |
return query.QueryFields(self.qcls.FIELDS, self.op.fields) |
1970 |
|
1971 |
|
1972 |
class LUNodeModifyStorage(NoHooksLU): |
1973 |
"""Logical unit for modifying a storage volume on a node.
|
1974 |
|
1975 |
"""
|
1976 |
REQ_BGL = False
|
1977 |
|
1978 |
def CheckArguments(self): |
1979 |
self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name) |
1980 |
|
1981 |
storage_type = self.op.storage_type
|
1982 |
|
1983 |
try:
|
1984 |
modifiable = constants.MODIFIABLE_STORAGE_FIELDS[storage_type] |
1985 |
except KeyError: |
1986 |
raise errors.OpPrereqError("Storage units of type '%s' can not be" |
1987 |
" modified" % storage_type,
|
1988 |
errors.ECODE_INVAL) |
1989 |
|
1990 |
diff = set(self.op.changes.keys()) - modifiable |
1991 |
if diff:
|
1992 |
raise errors.OpPrereqError("The following fields can not be modified for" |
1993 |
" storage units of type '%s': %r" %
|
1994 |
(storage_type, list(diff)),
|
1995 |
errors.ECODE_INVAL) |
1996 |
|
1997 |
def ExpandNames(self): |
1998 |
self.needed_locks = {
|
1999 |
locking.LEVEL_NODE: self.op.node_name,
|
2000 |
} |
2001 |
|
2002 |
def Exec(self, feedback_fn): |
2003 |
"""Computes the list of nodes and their attributes.
|
2004 |
|
2005 |
"""
|
2006 |
st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type) |
2007 |
result = self.rpc.call_storage_modify(self.op.node_name, |
2008 |
self.op.storage_type, st_args,
|
2009 |
self.op.name, self.op.changes) |
2010 |
result.Raise("Failed to modify storage unit '%s' on %s" %
|
2011 |
(self.op.name, self.op.node_name)) |
2012 |
|
2013 |
|
2014 |
class LUNodeAdd(LogicalUnit): |
2015 |
"""Logical unit for adding node to the cluster.
|
2016 |
|
2017 |
"""
|
2018 |
HPATH = "node-add"
|
2019 |
HTYPE = constants.HTYPE_NODE |
2020 |
_NFLAGS = ["master_capable", "vm_capable"] |
2021 |
|
2022 |
def CheckArguments(self): |
2023 |
self.primary_ip_family = self.cfg.GetPrimaryIPFamily() |
2024 |
# validate/normalize the node name
|
2025 |
self.hostname = netutils.GetHostname(name=self.op.node_name, |
2026 |
family=self.primary_ip_family)
|
2027 |
self.op.node_name = self.hostname.name |
2028 |
|
2029 |
if self.op.readd and self.op.node_name == self.cfg.GetMasterNode(): |
2030 |
raise errors.OpPrereqError("Cannot readd the master node", |
2031 |
errors.ECODE_STATE) |
2032 |
|
2033 |
if self.op.readd and self.op.group: |
2034 |
raise errors.OpPrereqError("Cannot pass a node group when a node is" |
2035 |
" being readded", errors.ECODE_INVAL)
|
2036 |
|
2037 |
def BuildHooksEnv(self): |
2038 |
"""Build hooks env.
|
2039 |
|
2040 |
This will run on all nodes before, and on all nodes + the new node after.
|
2041 |
|
2042 |
"""
|
2043 |
return {
|
2044 |
"OP_TARGET": self.op.node_name, |
2045 |
"NODE_NAME": self.op.node_name, |
2046 |
"NODE_PIP": self.op.primary_ip, |
2047 |
"NODE_SIP": self.op.secondary_ip, |
2048 |
"MASTER_CAPABLE": str(self.op.master_capable), |
2049 |
"VM_CAPABLE": str(self.op.vm_capable), |
2050 |
} |
2051 |
|
2052 |
def BuildHooksNodes(self): |
2053 |
"""Build hooks nodes.
|
2054 |
|
2055 |
"""
|
2056 |
# Exclude added node
|
2057 |
pre_nodes = list(set(self.cfg.GetNodeList()) - set([self.op.node_name])) |
2058 |
post_nodes = pre_nodes + [self.op.node_name, ]
|
2059 |
|
2060 |
return (pre_nodes, post_nodes)
|
2061 |
|
2062 |
def CheckPrereq(self): |
2063 |
"""Check prerequisites.
|
2064 |
|
2065 |
This checks:
|
2066 |
- the new node is not already in the config
|
2067 |
- it is resolvable
|
2068 |
- its parameters (single/dual homed) matches the cluster
|
2069 |
|
2070 |
Any errors are signaled by raising errors.OpPrereqError.
|
2071 |
|
2072 |
"""
|
2073 |
cfg = self.cfg
|
2074 |
hostname = self.hostname
|
2075 |
node = hostname.name |
2076 |
primary_ip = self.op.primary_ip = hostname.ip
|
2077 |
if self.op.secondary_ip is None: |
2078 |
if self.primary_ip_family == netutils.IP6Address.family: |
2079 |
raise errors.OpPrereqError("When using a IPv6 primary address, a valid" |
2080 |
" IPv4 address must be given as secondary",
|
2081 |
errors.ECODE_INVAL) |
2082 |
self.op.secondary_ip = primary_ip
|
2083 |
|
2084 |
secondary_ip = self.op.secondary_ip
|
2085 |
if not netutils.IP4Address.IsValid(secondary_ip): |
2086 |
raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4" |
2087 |
" address" % secondary_ip, errors.ECODE_INVAL)
|
2088 |
|
2089 |
node_list = cfg.GetNodeList() |
2090 |
if not self.op.readd and node in node_list: |
2091 |
raise errors.OpPrereqError("Node %s is already in the configuration" % |
2092 |
node, errors.ECODE_EXISTS) |
2093 |
elif self.op.readd and node not in node_list: |
2094 |
raise errors.OpPrereqError("Node %s is not in the configuration" % node, |
2095 |
errors.ECODE_NOENT) |
2096 |
|
2097 |
self.changed_primary_ip = False |
2098 |
|
2099 |
for existing_node_name, existing_node in cfg.GetMultiNodeInfo(node_list): |
2100 |
if self.op.readd and node == existing_node_name: |
2101 |
if existing_node.secondary_ip != secondary_ip:
|
2102 |
raise errors.OpPrereqError("Readded node doesn't have the same IP" |
2103 |
" address configuration as before",
|
2104 |
errors.ECODE_INVAL) |
2105 |
if existing_node.primary_ip != primary_ip:
|
2106 |
self.changed_primary_ip = True |
2107 |
|
2108 |
continue
|
2109 |
|
2110 |
if (existing_node.primary_ip == primary_ip or |
2111 |
existing_node.secondary_ip == primary_ip or
|
2112 |
existing_node.primary_ip == secondary_ip or
|
2113 |
existing_node.secondary_ip == secondary_ip): |
2114 |
raise errors.OpPrereqError("New node ip address(es) conflict with" |
2115 |
" existing node %s" % existing_node.name,
|
2116 |
errors.ECODE_NOTUNIQUE) |
2117 |
|
2118 |
# After this 'if' block, None is no longer a valid value for the
|
2119 |
# _capable op attributes
|
2120 |
if self.op.readd: |
2121 |
old_node = self.cfg.GetNodeInfo(node)
|
2122 |
assert old_node is not None, "Can't retrieve locked node %s" % node |
2123 |
for attr in self._NFLAGS: |
2124 |
if getattr(self.op, attr) is None: |
2125 |
setattr(self.op, attr, getattr(old_node, attr)) |
2126 |
else:
|
2127 |
for attr in self._NFLAGS: |
2128 |
if getattr(self.op, attr) is None: |
2129 |
setattr(self.op, attr, True) |
2130 |
|
2131 |
if self.op.readd and not self.op.vm_capable: |
2132 |
pri, sec = cfg.GetNodeInstances(node) |
2133 |
if pri or sec: |
2134 |
raise errors.OpPrereqError("Node %s being re-added with vm_capable" |
2135 |
" flag set to false, but it already holds"
|
2136 |
" instances" % node,
|
2137 |
errors.ECODE_STATE) |
2138 |
|
2139 |
# check that the type of the node (single versus dual homed) is the
|
2140 |
# same as for the master
|
2141 |
myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
|
2142 |
master_singlehomed = myself.secondary_ip == myself.primary_ip |
2143 |
newbie_singlehomed = secondary_ip == primary_ip |
2144 |
if master_singlehomed != newbie_singlehomed:
|
2145 |
if master_singlehomed:
|
2146 |
raise errors.OpPrereqError("The master has no secondary ip but the" |
2147 |
" new node has one",
|
2148 |
errors.ECODE_INVAL) |
2149 |
else:
|
2150 |
raise errors.OpPrereqError("The master has a secondary ip but the" |
2151 |
" new node doesn't have one",
|
2152 |
errors.ECODE_INVAL) |
2153 |
|
2154 |
# checks reachability
|
2155 |
if not netutils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT): |
2156 |
raise errors.OpPrereqError("Node not reachable by ping", |
2157 |
errors.ECODE_ENVIRON) |
2158 |
|
2159 |
if not newbie_singlehomed: |
2160 |
# check reachability from my secondary ip to newbie's secondary ip
|
2161 |
if not netutils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT, |
2162 |
source=myself.secondary_ip): |
2163 |
raise errors.OpPrereqError("Node secondary ip not reachable by TCP" |
2164 |
" based ping to node daemon port",
|
2165 |
errors.ECODE_ENVIRON) |
2166 |
|
2167 |
if self.op.readd: |
2168 |
exceptions = [node] |
2169 |
else:
|
2170 |
exceptions = [] |
2171 |
|
2172 |
if self.op.master_capable: |
2173 |
self.master_candidate = _DecideSelfPromotion(self, exceptions=exceptions) |
2174 |
else:
|
2175 |
self.master_candidate = False |
2176 |
|
2177 |
if self.op.readd: |
2178 |
self.new_node = old_node
|
2179 |
else:
|
2180 |
node_group = cfg.LookupNodeGroup(self.op.group)
|
2181 |
self.new_node = objects.Node(name=node,
|
2182 |
primary_ip=primary_ip, |
2183 |
secondary_ip=secondary_ip, |
2184 |
master_candidate=self.master_candidate,
|
2185 |
offline=False, drained=False, |
2186 |
group=node_group, ndparams={}) |
2187 |
|
2188 |
if self.op.ndparams: |
2189 |
utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
|
2190 |
_CheckParamsNotGlobal(self.op.ndparams, constants.NDC_GLOBALS, "node", |
2191 |
"node", "cluster or group") |
2192 |
|
2193 |
if self.op.hv_state: |
2194 |
self.new_hv_state = _MergeAndVerifyHvState(self.op.hv_state, None) |
2195 |
|
2196 |
if self.op.disk_state: |
2197 |
self.new_disk_state = _MergeAndVerifyDiskState(self.op.disk_state, None) |
2198 |
|
2199 |
# TODO: If we need to have multiple DnsOnlyRunner we probably should make
|
2200 |
# it a property on the base class.
|
2201 |
rpcrunner = rpc.DnsOnlyRunner() |
2202 |
result = rpcrunner.call_version([node])[node] |
2203 |
result.Raise("Can't get version information from node %s" % node)
|
2204 |
if constants.PROTOCOL_VERSION == result.payload:
|
2205 |
logging.info("Communication to node %s fine, sw version %s match",
|
2206 |
node, result.payload) |
2207 |
else:
|
2208 |
raise errors.OpPrereqError("Version mismatch master version %s," |
2209 |
" node version %s" %
|
2210 |
(constants.PROTOCOL_VERSION, result.payload), |
2211 |
errors.ECODE_ENVIRON) |
2212 |
|
2213 |
vg_name = cfg.GetVGName() |
2214 |
if vg_name is not None: |
2215 |
vparams = {constants.NV_PVLIST: [vg_name]} |
2216 |
excl_stor = _IsExclusiveStorageEnabledNode(cfg, self.new_node)
|
2217 |
cname = self.cfg.GetClusterName()
|
2218 |
result = rpcrunner.call_node_verify_light([node], vparams, cname)[node] |
2219 |
(errmsgs, _) = _CheckNodePVs(result.payload, excl_stor) |
2220 |
if errmsgs:
|
2221 |
raise errors.OpPrereqError("Checks on node PVs failed: %s" % |
2222 |
"; ".join(errmsgs), errors.ECODE_ENVIRON)
|
2223 |
|
2224 |
def Exec(self, feedback_fn): |
2225 |
"""Adds the new node to the cluster.
|
2226 |
|
2227 |
"""
|
2228 |
new_node = self.new_node
|
2229 |
node = new_node.name |
2230 |
|
2231 |
assert locking.BGL in self.owned_locks(locking.LEVEL_CLUSTER), \ |
2232 |
"Not owning BGL"
|
2233 |
|
2234 |
# We adding a new node so we assume it's powered
|
2235 |
new_node.powered = True
|
2236 |
|
2237 |
# for re-adds, reset the offline/drained/master-candidate flags;
|
2238 |
# we need to reset here, otherwise offline would prevent RPC calls
|
2239 |
# later in the procedure; this also means that if the re-add
|
2240 |
# fails, we are left with a non-offlined, broken node
|
2241 |
if self.op.readd: |
2242 |
new_node.drained = new_node.offline = False # pylint: disable=W0201 |
2243 |
self.LogInfo("Readding a node, the offline/drained flags were reset") |
2244 |
# if we demote the node, we do cleanup later in the procedure
|
2245 |
new_node.master_candidate = self.master_candidate
|
2246 |
if self.changed_primary_ip: |
2247 |
new_node.primary_ip = self.op.primary_ip
|
2248 |
|
2249 |
# copy the master/vm_capable flags
|
2250 |
for attr in self._NFLAGS: |
2251 |
setattr(new_node, attr, getattr(self.op, attr)) |
2252 |
|
2253 |
# notify the user about any possible mc promotion
|
2254 |
if new_node.master_candidate:
|
2255 |
self.LogInfo("Node will be a master candidate") |
2256 |
|
2257 |
if self.op.ndparams: |
2258 |
new_node.ndparams = self.op.ndparams
|
2259 |
else:
|
2260 |
new_node.ndparams = {} |
2261 |
|
2262 |
if self.op.hv_state: |
2263 |
new_node.hv_state_static = self.new_hv_state
|
2264 |
|
2265 |
if self.op.disk_state: |
2266 |
new_node.disk_state_static = self.new_disk_state
|
2267 |
|
2268 |
# Add node to our /etc/hosts, and add key to known_hosts
|
2269 |
if self.cfg.GetClusterInfo().modify_etc_hosts: |
2270 |
master_node = self.cfg.GetMasterNode()
|
2271 |
result = self.rpc.call_etc_hosts_modify(master_node,
|
2272 |
constants.ETC_HOSTS_ADD, |
2273 |
self.hostname.name,
|
2274 |
self.hostname.ip)
|
2275 |
result.Raise("Can't update hosts file with new host data")
|
2276 |
|
2277 |
if new_node.secondary_ip != new_node.primary_ip:
|
2278 |
_CheckNodeHasSecondaryIP(self, new_node.name, new_node.secondary_ip,
|
2279 |
False)
|
2280 |
|
2281 |
node_verify_list = [self.cfg.GetMasterNode()]
|
2282 |
node_verify_param = { |
2283 |
constants.NV_NODELIST: ([node], {}), |
2284 |
# TODO: do a node-net-test as well?
|
2285 |
} |
2286 |
|
2287 |
result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
|
2288 |
self.cfg.GetClusterName())
|
2289 |
for verifier in node_verify_list: |
2290 |
result[verifier].Raise("Cannot communicate with node %s" % verifier)
|
2291 |
nl_payload = result[verifier].payload[constants.NV_NODELIST] |
2292 |
if nl_payload:
|
2293 |
for failed in nl_payload: |
2294 |
feedback_fn("ssh/hostname verification failed"
|
2295 |
" (checking from %s): %s" %
|
2296 |
(verifier, nl_payload[failed])) |
2297 |
raise errors.OpExecError("ssh/hostname verification failed") |
2298 |
|
2299 |
if self.op.readd: |
2300 |
_RedistributeAncillaryFiles(self)
|
2301 |
self.context.ReaddNode(new_node)
|
2302 |
# make sure we redistribute the config
|
2303 |
self.cfg.Update(new_node, feedback_fn)
|
2304 |
# and make sure the new node will not have old files around
|
2305 |
if not new_node.master_candidate: |
2306 |
result = self.rpc.call_node_demote_from_mc(new_node.name)
|
2307 |
msg = result.fail_msg |
2308 |
if msg:
|
2309 |
self.LogWarning("Node failed to demote itself from master" |
2310 |
" candidate status: %s" % msg)
|
2311 |
else:
|
2312 |
_RedistributeAncillaryFiles(self, additional_nodes=[node],
|
2313 |
additional_vm=self.op.vm_capable)
|
2314 |
self.context.AddNode(new_node, self.proc.GetECId()) |
2315 |
|
2316 |
|
2317 |
class LUNodeSetParams(LogicalUnit): |
2318 |
"""Modifies the parameters of a node.
|
2319 |
|
2320 |
@cvar _F2R: a dictionary from tuples of flags (mc, drained, offline)
|
2321 |
to the node role (as _ROLE_*)
|
2322 |
@cvar _R2F: a dictionary from node role to tuples of flags
|
2323 |
@cvar _FLAGS: a list of attribute names corresponding to the flags
|
2324 |
|
2325 |
"""
|
2326 |
HPATH = "node-modify"
|
2327 |
HTYPE = constants.HTYPE_NODE |
2328 |
REQ_BGL = False
|
2329 |
(_ROLE_CANDIDATE, _ROLE_DRAINED, _ROLE_OFFLINE, _ROLE_REGULAR) = range(4) |
2330 |
_F2R = { |
2331 |
(True, False, False): _ROLE_CANDIDATE, |
2332 |
(False, True, False): _ROLE_DRAINED, |
2333 |
(False, False, True): _ROLE_OFFLINE, |
2334 |
(False, False, False): _ROLE_REGULAR, |
2335 |
} |
2336 |
_R2F = dict((v, k) for k, v in _F2R.items()) |
2337 |
_FLAGS = ["master_candidate", "drained", "offline"] |
2338 |
|
2339 |
def CheckArguments(self): |
2340 |
self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name) |
2341 |
all_mods = [self.op.offline, self.op.master_candidate, self.op.drained, |
2342 |
self.op.master_capable, self.op.vm_capable, |
2343 |
self.op.secondary_ip, self.op.ndparams, self.op.hv_state, |
2344 |
self.op.disk_state]
|
2345 |
if all_mods.count(None) == len(all_mods): |
2346 |
raise errors.OpPrereqError("Please pass at least one modification", |
2347 |
errors.ECODE_INVAL) |
2348 |
if all_mods.count(True) > 1: |
2349 |
raise errors.OpPrereqError("Can't set the node into more than one" |
2350 |
" state at the same time",
|
2351 |
errors.ECODE_INVAL) |
2352 |
|
2353 |
# Boolean value that tells us whether we might be demoting from MC
|
2354 |
self.might_demote = (self.op.master_candidate is False or |
2355 |
self.op.offline is True or |
2356 |
self.op.drained is True or |
2357 |
self.op.master_capable is False) |
2358 |
|
2359 |
if self.op.secondary_ip: |
2360 |
if not netutils.IP4Address.IsValid(self.op.secondary_ip): |
2361 |
raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4" |
2362 |
" address" % self.op.secondary_ip, |
2363 |
errors.ECODE_INVAL) |
2364 |
|
2365 |
self.lock_all = self.op.auto_promote and self.might_demote |
2366 |
self.lock_instances = self.op.secondary_ip is not None |
2367 |
|
2368 |
def _InstanceFilter(self, instance): |
2369 |
"""Filter for getting affected instances.
|
2370 |
|
2371 |
"""
|
2372 |
return (instance.disk_template in constants.DTS_INT_MIRROR and |
2373 |
self.op.node_name in instance.all_nodes) |
2374 |
|
2375 |
def ExpandNames(self): |
2376 |
if self.lock_all: |
2377 |
self.needed_locks = {
|
2378 |
locking.LEVEL_NODE: locking.ALL_SET, |
2379 |
|
2380 |
# Block allocations when all nodes are locked
|
2381 |
locking.LEVEL_NODE_ALLOC: locking.ALL_SET, |
2382 |
} |
2383 |
else:
|
2384 |
self.needed_locks = {
|
2385 |
locking.LEVEL_NODE: self.op.node_name,
|
2386 |
} |
2387 |
|
2388 |
# Since modifying a node can have severe effects on currently running
|
2389 |
# operations the resource lock is at least acquired in shared mode
|
2390 |
self.needed_locks[locking.LEVEL_NODE_RES] = \
|
2391 |
self.needed_locks[locking.LEVEL_NODE]
|
2392 |
|
2393 |
# Get all locks except nodes in shared mode; they are not used for anything
|
2394 |
# but read-only access
|
2395 |
self.share_locks = _ShareAll()
|
2396 |
self.share_locks[locking.LEVEL_NODE] = 0 |
2397 |
self.share_locks[locking.LEVEL_NODE_RES] = 0 |
2398 |
self.share_locks[locking.LEVEL_NODE_ALLOC] = 0 |
2399 |
|
2400 |
if self.lock_instances: |
2401 |
self.needed_locks[locking.LEVEL_INSTANCE] = \
|
2402 |
frozenset(self.cfg.GetInstancesInfoByFilter(self._InstanceFilter)) |
2403 |
|
2404 |
def BuildHooksEnv(self): |
2405 |
"""Build hooks env.
|
2406 |
|
2407 |
This runs on the master node.
|
2408 |
|
2409 |
"""
|
2410 |
return {
|
2411 |
"OP_TARGET": self.op.node_name, |
2412 |
"MASTER_CANDIDATE": str(self.op.master_candidate), |
2413 |
"OFFLINE": str(self.op.offline), |
2414 |
"DRAINED": str(self.op.drained), |
2415 |
"MASTER_CAPABLE": str(self.op.master_capable), |
2416 |
"VM_CAPABLE": str(self.op.vm_capable), |
2417 |
} |
2418 |
|
2419 |
def BuildHooksNodes(self): |
2420 |
"""Build hooks nodes.
|
2421 |
|
2422 |
"""
|
2423 |
nl = [self.cfg.GetMasterNode(), self.op.node_name] |
2424 |
return (nl, nl)
|
2425 |
|
2426 |
def CheckPrereq(self): |
2427 |
"""Check prerequisites.
|
2428 |
|
2429 |
This only checks the instance list against the existing names.
|
2430 |
|
2431 |
"""
|
2432 |
node = self.node = self.cfg.GetNodeInfo(self.op.node_name) |
2433 |
|
2434 |
if self.lock_instances: |
2435 |
affected_instances = \ |
2436 |
self.cfg.GetInstancesInfoByFilter(self._InstanceFilter) |
2437 |
|
2438 |
# Verify instance locks
|
2439 |
owned_instances = self.owned_locks(locking.LEVEL_INSTANCE)
|
2440 |
wanted_instances = frozenset(affected_instances.keys())
|
2441 |
if wanted_instances - owned_instances:
|
2442 |
raise errors.OpPrereqError("Instances affected by changing node %s's" |
2443 |
" secondary IP address have changed since"
|
2444 |
" locks were acquired, wanted '%s', have"
|
2445 |
" '%s'; retry the operation" %
|
2446 |
(self.op.node_name,
|
2447 |
utils.CommaJoin(wanted_instances), |
2448 |
utils.CommaJoin(owned_instances)), |
2449 |
errors.ECODE_STATE) |
2450 |
else:
|
2451 |
affected_instances = None
|
2452 |
|
2453 |
if (self.op.master_candidate is not None or |
2454 |
self.op.drained is not None or |
2455 |
self.op.offline is not None): |
2456 |
# we can't change the master's node flags
|
2457 |
if self.op.node_name == self.cfg.GetMasterNode(): |
2458 |
raise errors.OpPrereqError("The master role can be changed" |
2459 |
" only via master-failover",
|
2460 |
errors.ECODE_INVAL) |
2461 |
|
2462 |
if self.op.master_candidate and not node.master_capable: |
2463 |
raise errors.OpPrereqError("Node %s is not master capable, cannot make" |
2464 |
" it a master candidate" % node.name,
|
2465 |
errors.ECODE_STATE) |
2466 |
|
2467 |
if self.op.vm_capable is False: |
2468 |
(ipri, isec) = self.cfg.GetNodeInstances(self.op.node_name) |
2469 |
if ipri or isec: |
2470 |
raise errors.OpPrereqError("Node %s hosts instances, cannot unset" |
2471 |
" the vm_capable flag" % node.name,
|
2472 |
errors.ECODE_STATE) |
2473 |
|
2474 |
if node.master_candidate and self.might_demote and not self.lock_all: |
2475 |
assert not self.op.auto_promote, "auto_promote set but lock_all not" |
2476 |
# check if after removing the current node, we're missing master
|
2477 |
# candidates
|
2478 |
(mc_remaining, mc_should, _) = \ |
2479 |
self.cfg.GetMasterCandidateStats(exceptions=[node.name])
|
2480 |
if mc_remaining < mc_should:
|
2481 |
raise errors.OpPrereqError("Not enough master candidates, please" |
2482 |
" pass auto promote option to allow"
|
2483 |
" promotion (--auto-promote or RAPI"
|
2484 |
" auto_promote=True)", errors.ECODE_STATE)
|
2485 |
|
2486 |
self.old_flags = old_flags = (node.master_candidate,
|
2487 |
node.drained, node.offline) |
2488 |
assert old_flags in self._F2R, "Un-handled old flags %s" % str(old_flags) |
2489 |
self.old_role = old_role = self._F2R[old_flags] |
2490 |
|
2491 |
# Check for ineffective changes
|
2492 |
for attr in self._FLAGS: |
2493 |
if (getattr(self.op, attr) is False and getattr(node, attr) is False): |
2494 |
self.LogInfo("Ignoring request to unset flag %s, already unset", attr) |
2495 |
setattr(self.op, attr, None) |
2496 |
|
2497 |
# Past this point, any flag change to False means a transition
|
2498 |
# away from the respective state, as only real changes are kept
|
2499 |
|
2500 |
# TODO: We might query the real power state if it supports OOB
|
2501 |
if _SupportsOob(self.cfg, node): |
2502 |
if self.op.offline is False and not (node.powered or |
2503 |
self.op.powered is True): |
2504 |
raise errors.OpPrereqError(("Node %s needs to be turned on before its" |
2505 |
" offline status can be reset") %
|
2506 |
self.op.node_name, errors.ECODE_STATE)
|
2507 |
elif self.op.powered is not None: |
2508 |
raise errors.OpPrereqError(("Unable to change powered state for node %s" |
2509 |
" as it does not support out-of-band"
|
2510 |
" handling") % self.op.node_name, |
2511 |
errors.ECODE_STATE) |
2512 |
|
2513 |
# If we're being deofflined/drained, we'll MC ourself if needed
|
2514 |
if (self.op.drained is False or self.op.offline is False or |
2515 |
(self.op.master_capable and not node.master_capable)): |
2516 |
if _DecideSelfPromotion(self): |
2517 |
self.op.master_candidate = True |
2518 |
self.LogInfo("Auto-promoting node to master candidate") |
2519 |
|
2520 |
# If we're no longer master capable, we'll demote ourselves from MC
|
2521 |
if self.op.master_capable is False and node.master_candidate: |
2522 |
self.LogInfo("Demoting from master candidate") |
2523 |
self.op.master_candidate = False |
2524 |
|
2525 |
# Compute new role
|
2526 |
assert [getattr(self.op, attr) for attr in self._FLAGS].count(True) <= 1 |
2527 |
if self.op.master_candidate: |
2528 |
new_role = self._ROLE_CANDIDATE
|
2529 |
elif self.op.drained: |
2530 |
new_role = self._ROLE_DRAINED
|
2531 |
elif self.op.offline: |
2532 |
new_role = self._ROLE_OFFLINE
|
2533 |
elif False in [self.op.master_candidate, self.op.drained, self.op.offline]: |
2534 |
# False is still in new flags, which means we're un-setting (the
|
2535 |
# only) True flag
|
2536 |
new_role = self._ROLE_REGULAR
|
2537 |
else: # no new flags, nothing, keep old role |
2538 |
new_role = old_role |
2539 |
|
2540 |
self.new_role = new_role
|
2541 |
|
2542 |
if old_role == self._ROLE_OFFLINE and new_role != old_role: |
2543 |
# Trying to transition out of offline status
|
2544 |
result = self.rpc.call_version([node.name])[node.name]
|
2545 |
if result.fail_msg:
|
2546 |
raise errors.OpPrereqError("Node %s is being de-offlined but fails" |
2547 |
" to report its version: %s" %
|
2548 |
(node.name, result.fail_msg), |
2549 |
errors.ECODE_STATE) |
2550 |
else:
|
2551 |
self.LogWarning("Transitioning node from offline to online state" |
2552 |
" without using re-add. Please make sure the node"
|
2553 |
" is healthy!")
|
2554 |
|
2555 |
# When changing the secondary ip, verify if this is a single-homed to
|
2556 |
# multi-homed transition or vice versa, and apply the relevant
|
2557 |
# restrictions.
|
2558 |
if self.op.secondary_ip: |
2559 |
# Ok even without locking, because this can't be changed by any LU
|
2560 |
master = self.cfg.GetNodeInfo(self.cfg.GetMasterNode()) |
2561 |
master_singlehomed = master.secondary_ip == master.primary_ip |
2562 |
if master_singlehomed and self.op.secondary_ip != node.primary_ip: |
2563 |
if self.op.force and node.name == master.name: |
2564 |
self.LogWarning("Transitioning from single-homed to multi-homed" |
2565 |
" cluster; all nodes will require a secondary IP"
|
2566 |
" address")
|
2567 |
else:
|
2568 |
raise errors.OpPrereqError("Changing the secondary ip on a" |
2569 |
" single-homed cluster requires the"
|
2570 |
" --force option to be passed, and the"
|
2571 |
" target node to be the master",
|
2572 |
errors.ECODE_INVAL) |
2573 |
elif not master_singlehomed and self.op.secondary_ip == node.primary_ip: |
2574 |
if self.op.force and node.name == master.name: |
2575 |
self.LogWarning("Transitioning from multi-homed to single-homed" |
2576 |
" cluster; secondary IP addresses will have to be"
|
2577 |
" removed")
|
2578 |
else:
|
2579 |
raise errors.OpPrereqError("Cannot set the secondary IP to be the" |
2580 |
" same as the primary IP on a multi-homed"
|
2581 |
" cluster, unless the --force option is"
|
2582 |
" passed, and the target node is the"
|
2583 |
" master", errors.ECODE_INVAL)
|
2584 |
|
2585 |
assert not (frozenset(affected_instances) - |
2586 |
self.owned_locks(locking.LEVEL_INSTANCE))
|
2587 |
|
2588 |
if node.offline:
|
2589 |
if affected_instances:
|
2590 |
msg = ("Cannot change secondary IP address: offline node has"
|
2591 |
" instances (%s) configured to use it" %
|
2592 |
utils.CommaJoin(affected_instances.keys())) |
2593 |
raise errors.OpPrereqError(msg, errors.ECODE_STATE)
|
2594 |
else:
|
2595 |
# On online nodes, check that no instances are running, and that
|
2596 |
# the node has the new ip and we can reach it.
|
2597 |
for instance in affected_instances.values(): |
2598 |
_CheckInstanceState(self, instance, INSTANCE_DOWN,
|
2599 |
msg="cannot change secondary ip")
|
2600 |
|
2601 |
_CheckNodeHasSecondaryIP(self, node.name, self.op.secondary_ip, True) |
2602 |
if master.name != node.name:
|
2603 |
# check reachability from master secondary ip to new secondary ip
|
2604 |
if not netutils.TcpPing(self.op.secondary_ip, |
2605 |
constants.DEFAULT_NODED_PORT, |
2606 |
source=master.secondary_ip): |
2607 |
raise errors.OpPrereqError("Node secondary ip not reachable by TCP" |
2608 |
" based ping to node daemon port",
|
2609 |
errors.ECODE_ENVIRON) |
2610 |
|
2611 |
if self.op.ndparams: |
2612 |
new_ndparams = _GetUpdatedParams(self.node.ndparams, self.op.ndparams) |
2613 |
utils.ForceDictType(new_ndparams, constants.NDS_PARAMETER_TYPES) |
2614 |
_CheckParamsNotGlobal(self.op.ndparams, constants.NDC_GLOBALS, "node", |
2615 |
"node", "cluster or group") |
2616 |
self.new_ndparams = new_ndparams
|
2617 |
|
2618 |
if self.op.hv_state: |
2619 |
self.new_hv_state = _MergeAndVerifyHvState(self.op.hv_state, |
2620 |
self.node.hv_state_static)
|
2621 |
|
2622 |
if self.op.disk_state: |
2623 |
self.new_disk_state = \
|
2624 |
_MergeAndVerifyDiskState(self.op.disk_state,
|
2625 |
self.node.disk_state_static)
|
2626 |
|
2627 |
def Exec(self, feedback_fn): |
2628 |
"""Modifies a node.
|
2629 |
|
2630 |
"""
|
2631 |
node = self.node
|
2632 |
old_role = self.old_role
|
2633 |
new_role = self.new_role
|
2634 |
|
2635 |
result = [] |
2636 |
|
2637 |
if self.op.ndparams: |
2638 |
node.ndparams = self.new_ndparams
|
2639 |
|
2640 |
if self.op.powered is not None: |
2641 |
node.powered = self.op.powered
|
2642 |
|
2643 |
if self.op.hv_state: |
2644 |
node.hv_state_static = self.new_hv_state
|
2645 |
|
2646 |
if self.op.disk_state: |
2647 |
node.disk_state_static = self.new_disk_state
|
2648 |
|
2649 |
for attr in ["master_capable", "vm_capable"]: |
2650 |
val = getattr(self.op, attr) |
2651 |
if val is not None: |
2652 |
setattr(node, attr, val)
|
2653 |
result.append((attr, str(val)))
|
2654 |
|
2655 |
if new_role != old_role:
|
2656 |
# Tell the node to demote itself, if no longer MC and not offline
|
2657 |
if old_role == self._ROLE_CANDIDATE and new_role != self._ROLE_OFFLINE: |
2658 |
msg = self.rpc.call_node_demote_from_mc(node.name).fail_msg
|
2659 |
if msg:
|
2660 |
self.LogWarning("Node failed to demote itself: %s", msg) |
2661 |
|
2662 |
new_flags = self._R2F[new_role]
|
2663 |
for of, nf, desc in zip(self.old_flags, new_flags, self._FLAGS): |
2664 |
if of != nf:
|
2665 |
result.append((desc, str(nf)))
|
2666 |
(node.master_candidate, node.drained, node.offline) = new_flags |
2667 |
|
2668 |
# we locked all nodes, we adjust the CP before updating this node
|
2669 |
if self.lock_all: |
2670 |
_AdjustCandidatePool(self, [node.name])
|
2671 |
|
2672 |
if self.op.secondary_ip: |
2673 |
node.secondary_ip = self.op.secondary_ip
|
2674 |
result.append(("secondary_ip", self.op.secondary_ip)) |
2675 |
|
2676 |
# this will trigger configuration file update, if needed
|
2677 |
self.cfg.Update(node, feedback_fn)
|
2678 |
|
2679 |
# this will trigger job queue propagation or cleanup if the mc
|
2680 |
# flag changed
|
2681 |
if [old_role, new_role].count(self._ROLE_CANDIDATE) == 1: |
2682 |
self.context.ReaddNode(node)
|
2683 |
|
2684 |
return result
|
2685 |
|
2686 |
|
2687 |
class LUNodePowercycle(NoHooksLU): |
2688 |
"""Powercycles a node.
|
2689 |
|
2690 |
"""
|
2691 |
REQ_BGL = False
|
2692 |
|
2693 |
def CheckArguments(self): |
2694 |
self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name) |
2695 |
if self.op.node_name == self.cfg.GetMasterNode() and not self.op.force: |
2696 |
raise errors.OpPrereqError("The node is the master and the force" |
2697 |
" parameter was not set",
|
2698 |
errors.ECODE_INVAL) |
2699 |
|
2700 |
def ExpandNames(self): |
2701 |
"""Locking for PowercycleNode.
|
2702 |
|
2703 |
This is a last-resort option and shouldn't block on other
|
2704 |
jobs. Therefore, we grab no locks.
|
2705 |
|
2706 |
"""
|
2707 |
self.needed_locks = {}
|
2708 |
|
2709 |
def Exec(self, feedback_fn): |
2710 |
"""Reboots a node.
|
2711 |
|
2712 |
"""
|
2713 |
result = self.rpc.call_node_powercycle(self.op.node_name, |
2714 |
self.cfg.GetHypervisorType())
|
2715 |
result.Raise("Failed to schedule the reboot")
|
2716 |
return result.payload
|
2717 |
|
2718 |
|
2719 |
class LUInstanceActivateDisks(NoHooksLU): |
2720 |
"""Bring up an instance's disks.
|
2721 |
|
2722 |
"""
|
2723 |
REQ_BGL = False
|
2724 |
|
2725 |
def ExpandNames(self): |
2726 |
self._ExpandAndLockInstance()
|
2727 |
self.needed_locks[locking.LEVEL_NODE] = []
|
2728 |
self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
|
2729 |
|
2730 |
def DeclareLocks(self, level): |
2731 |
if level == locking.LEVEL_NODE:
|
2732 |
self._LockInstancesNodes()
|
2733 |
|
2734 |
def CheckPrereq(self): |
2735 |
"""Check prerequisites.
|
2736 |
|
2737 |
This checks that the instance is in the cluster.
|
2738 |
|
2739 |
"""
|
2740 |
self.instance = self.cfg.GetInstanceInfo(self.op.instance_name) |
2741 |
assert self.instance is not None, \ |
2742 |
"Cannot retrieve locked instance %s" % self.op.instance_name |
2743 |
_CheckNodeOnline(self, self.instance.primary_node) |
2744 |
|
2745 |
def Exec(self, feedback_fn): |
2746 |
"""Activate the disks.
|
2747 |
|
2748 |
"""
|
2749 |
disks_ok, disks_info = \ |
2750 |
_AssembleInstanceDisks(self, self.instance, |
2751 |
ignore_size=self.op.ignore_size)
|
2752 |
if not disks_ok: |
2753 |
raise errors.OpExecError("Cannot activate block devices") |
2754 |
|
2755 |
if self.op.wait_for_sync: |
2756 |
if not _WaitForSync(self, self.instance): |
2757 |
raise errors.OpExecError("Some disks of the instance are degraded!") |
2758 |
|
2759 |
return disks_info
|
2760 |
|
2761 |
|
2762 |
def _AssembleInstanceDisks(lu, instance, disks=None, ignore_secondaries=False, |
2763 |
ignore_size=False):
|
2764 |
"""Prepare the block devices for an instance.
|
2765 |
|
2766 |
This sets up the block devices on all nodes.
|
2767 |
|
2768 |
@type lu: L{LogicalUnit}
|
2769 |
@param lu: the logical unit on whose behalf we execute
|
2770 |
@type instance: L{objects.Instance}
|
2771 |
@param instance: the instance for whose disks we assemble
|
2772 |
@type disks: list of L{objects.Disk} or None
|
2773 |
@param disks: which disks to assemble (or all, if None)
|
2774 |
@type ignore_secondaries: boolean
|
2775 |
@param ignore_secondaries: if true, errors on secondary nodes
|
2776 |
won't result in an error return from the function
|
2777 |
@type ignore_size: boolean
|
2778 |
@param ignore_size: if true, the current known size of the disk
|
2779 |
will not be used during the disk activation, useful for cases
|
2780 |
when the size is wrong
|
2781 |
@return: False if the operation failed, otherwise a list of
|
2782 |
(host, instance_visible_name, node_visible_name)
|
2783 |
with the mapping from node devices to instance devices
|
2784 |
|
2785 |
"""
|
2786 |
device_info = [] |
2787 |
disks_ok = True
|
2788 |
iname = instance.name |
2789 |
disks = _ExpandCheckDisks(instance, disks) |
2790 |
|
2791 |
# With the two passes mechanism we try to reduce the window of
|
2792 |
# opportunity for the race condition of switching DRBD to primary
|
2793 |
# before handshaking occured, but we do not eliminate it
|
2794 |
|
2795 |
# The proper fix would be to wait (with some limits) until the
|
2796 |
# connection has been made and drbd transitions from WFConnection
|
2797 |
# into any other network-connected state (Connected, SyncTarget,
|
2798 |
# SyncSource, etc.)
|
2799 |
|
2800 |
# 1st pass, assemble on all nodes in secondary mode
|
2801 |
for idx, inst_disk in enumerate(disks): |
2802 |
for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node): |
2803 |
if ignore_size:
|
2804 |
node_disk = node_disk.Copy() |
2805 |
node_disk.UnsetSize() |
2806 |
lu.cfg.SetDiskID(node_disk, node) |
2807 |
result = lu.rpc.call_blockdev_assemble(node, (node_disk, instance), iname, |
2808 |
False, idx)
|
2809 |
msg = result.fail_msg |
2810 |
if msg:
|
2811 |
is_offline_secondary = (node in instance.secondary_nodes and |
2812 |
result.offline) |
2813 |
lu.LogWarning("Could not prepare block device %s on node %s"
|
2814 |
" (is_primary=False, pass=1): %s",
|
2815 |
inst_disk.iv_name, node, msg) |
2816 |
if not (ignore_secondaries or is_offline_secondary): |
2817 |
disks_ok = False
|
2818 |
|
2819 |
# FIXME: race condition on drbd migration to primary
|
2820 |
|
2821 |
# 2nd pass, do only the primary node
|
2822 |
for idx, inst_disk in enumerate(disks): |
2823 |
dev_path = None
|
2824 |
|
2825 |
for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node): |
2826 |
if node != instance.primary_node:
|
2827 |
continue
|
2828 |
if ignore_size:
|
2829 |
node_disk = node_disk.Copy() |
2830 |
node_disk.UnsetSize() |
2831 |
lu.cfg.SetDiskID(node_disk, node) |
2832 |
result = lu.rpc.call_blockdev_assemble(node, (node_disk, instance), iname, |
2833 |
True, idx)
|
2834 |
msg = result.fail_msg |
2835 |
if msg:
|
2836 |
lu.LogWarning("Could not prepare block device %s on node %s"
|
2837 |
" (is_primary=True, pass=2): %s",
|
2838 |
inst_disk.iv_name, node, msg) |
2839 |
disks_ok = False
|
2840 |
else:
|
2841 |
dev_path = result.payload |
2842 |
|
2843 |
device_info.append((instance.primary_node, inst_disk.iv_name, dev_path)) |
2844 |
|
2845 |
# leave the disks configured for the primary node
|
2846 |
# this is a workaround that would be fixed better by
|
2847 |
# improving the logical/physical id handling
|
2848 |
for disk in disks: |
2849 |
lu.cfg.SetDiskID(disk, instance.primary_node) |
2850 |
|
2851 |
return disks_ok, device_info
|
2852 |
|
2853 |
|
2854 |
def _StartInstanceDisks(lu, instance, force): |
2855 |
"""Start the disks of an instance.
|
2856 |
|
2857 |
"""
|
2858 |
disks_ok, _ = _AssembleInstanceDisks(lu, instance, |
2859 |
ignore_secondaries=force) |
2860 |
if not disks_ok: |
2861 |
_ShutdownInstanceDisks(lu, instance) |
2862 |
if force is not None and not force: |
2863 |
lu.LogWarning("",
|
2864 |
hint=("If the message above refers to a secondary node,"
|
2865 |
" you can retry the operation using '--force'"))
|
2866 |
raise errors.OpExecError("Disk consistency error") |
2867 |
|
2868 |
|
2869 |
class LUInstanceDeactivateDisks(NoHooksLU): |
2870 |
"""Shutdown an instance's disks.
|
2871 |
|
2872 |
"""
|
2873 |
REQ_BGL = False
|
2874 |
|
2875 |
def ExpandNames(self): |
2876 |
self._ExpandAndLockInstance()
|
2877 |
self.needed_locks[locking.LEVEL_NODE] = []
|
2878 |
self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
|
2879 |
|
2880 |
def DeclareLocks(self, level): |
2881 |
if level == locking.LEVEL_NODE:
|
2882 |
self._LockInstancesNodes()
|
2883 |
|
2884 |
def CheckPrereq(self): |
2885 |
"""Check prerequisites.
|
2886 |
|
2887 |
This checks that the instance is in the cluster.
|
2888 |
|
2889 |
"""
|
2890 |
self.instance = self.cfg.GetInstanceInfo(self.op.instance_name) |
2891 |
assert self.instance is not None, \ |
2892 |
"Cannot retrieve locked instance %s" % self.op.instance_name |
2893 |
|
2894 |
def Exec(self, feedback_fn): |
2895 |
"""Deactivate the disks
|
2896 |
|
2897 |
"""
|
2898 |
instance = self.instance
|
2899 |
if self.op.force: |
2900 |
_ShutdownInstanceDisks(self, instance)
|
2901 |
else:
|
2902 |
_SafeShutdownInstanceDisks(self, instance)
|
2903 |
|
2904 |
|
2905 |
def _SafeShutdownInstanceDisks(lu, instance, disks=None): |
2906 |
"""Shutdown block devices of an instance.
|
2907 |
|
2908 |
This function checks if an instance is running, before calling
|
2909 |
_ShutdownInstanceDisks.
|
2910 |
|
2911 |
"""
|
2912 |
_CheckInstanceState(lu, instance, INSTANCE_DOWN, msg="cannot shutdown disks")
|
2913 |
_ShutdownInstanceDisks(lu, instance, disks=disks) |
2914 |
|
2915 |
|
2916 |
def _ExpandCheckDisks(instance, disks): |
2917 |
"""Return the instance disks selected by the disks list
|
2918 |
|
2919 |
@type disks: list of L{objects.Disk} or None
|
2920 |
@param disks: selected disks
|
2921 |
@rtype: list of L{objects.Disk}
|
2922 |
@return: selected instance disks to act on
|
2923 |
|
2924 |
"""
|
2925 |
if disks is None: |
2926 |
return instance.disks
|
2927 |
else:
|
2928 |
if not set(disks).issubset(instance.disks): |
2929 |
raise errors.ProgrammerError("Can only act on disks belonging to the" |
2930 |
" target instance")
|
2931 |
return disks
|
2932 |
|
2933 |
|
2934 |
def _ShutdownInstanceDisks(lu, instance, disks=None, ignore_primary=False): |
2935 |
"""Shutdown block devices of an instance.
|
2936 |
|
2937 |
This does the shutdown on all nodes of the instance.
|
2938 |
|
2939 |
If the ignore_primary is false, errors on the primary node are
|
2940 |
ignored.
|
2941 |
|
2942 |
"""
|
2943 |
all_result = True
|
2944 |
disks = _ExpandCheckDisks(instance, disks) |
2945 |
|
2946 |
for disk in disks: |
2947 |
for node, top_disk in disk.ComputeNodeTree(instance.primary_node): |
2948 |
lu.cfg.SetDiskID(top_disk, node) |
2949 |
result = lu.rpc.call_blockdev_shutdown(node, (top_disk, instance)) |
2950 |
msg = result.fail_msg |
2951 |
if msg:
|
2952 |
lu.LogWarning("Could not shutdown block device %s on node %s: %s",
|
2953 |
disk.iv_name, node, msg) |
2954 |
if ((node == instance.primary_node and not ignore_primary) or |
2955 |
(node != instance.primary_node and not result.offline)): |
2956 |
all_result = False
|
2957 |
return all_result
|
2958 |
|
2959 |
|
2960 |
def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name): |
2961 |
"""Checks if a node has enough free memory.
|
2962 |
|
2963 |
This function checks if a given node has the needed amount of free
|
2964 |
memory. In case the node has less memory or we cannot get the
|
2965 |
information from the node, this function raises an OpPrereqError
|
2966 |
exception.
|
2967 |
|
2968 |
@type lu: C{LogicalUnit}
|
2969 |
@param lu: a logical unit from which we get configuration data
|
2970 |
@type node: C{str}
|
2971 |
@param node: the node to check
|
2972 |
@type reason: C{str}
|
2973 |
@param reason: string to use in the error message
|
2974 |
@type requested: C{int}
|
2975 |
@param requested: the amount of memory in MiB to check for
|
2976 |
@type hypervisor_name: C{str}
|
2977 |
@param hypervisor_name: the hypervisor to ask for memory stats
|
2978 |
@rtype: integer
|
2979 |
@return: node current free memory
|
2980 |
@raise errors.OpPrereqError: if the node doesn't have enough memory, or
|
2981 |
we cannot check the node
|
2982 |
|
2983 |
"""
|
2984 |
nodeinfo = lu.rpc.call_node_info([node], None, [hypervisor_name], False) |
2985 |
nodeinfo[node].Raise("Can't get data from node %s" % node,
|
2986 |
prereq=True, ecode=errors.ECODE_ENVIRON)
|
2987 |
(_, _, (hv_info, )) = nodeinfo[node].payload |
2988 |
|
2989 |
free_mem = hv_info.get("memory_free", None) |
2990 |
if not isinstance(free_mem, int): |
2991 |
raise errors.OpPrereqError("Can't compute free memory on node %s, result" |
2992 |
" was '%s'" % (node, free_mem),
|
2993 |
errors.ECODE_ENVIRON) |
2994 |
if requested > free_mem:
|
2995 |
raise errors.OpPrereqError("Not enough memory on node %s for %s:" |
2996 |
" needed %s MiB, available %s MiB" %
|
2997 |
(node, reason, requested, free_mem), |
2998 |
errors.ECODE_NORES) |
2999 |
return free_mem
|
3000 |
|
3001 |
|
3002 |
def _CheckNodesFreeDiskPerVG(lu, nodenames, req_sizes): |
3003 |
"""Checks if nodes have enough free disk space in all the VGs.
|
3004 |
|
3005 |
This function checks if all given nodes have the needed amount of
|
3006 |
free disk. In case any node has less disk or we cannot get the
|
3007 |
information from the node, this function raises an OpPrereqError
|
3008 |
exception.
|
3009 |
|
3010 |
@type lu: C{LogicalUnit}
|
3011 |
@param lu: a logical unit from which we get configuration data
|
3012 |
@type nodenames: C{list}
|
3013 |
@param nodenames: the list of node names to check
|
3014 |
@type req_sizes: C{dict}
|
3015 |
@param req_sizes: the hash of vg and corresponding amount of disk in
|
3016 |
MiB to check for
|
3017 |
@raise errors.OpPrereqError: if the node doesn't have enough disk,
|
3018 |
or we cannot check the node
|
3019 |
|
3020 |
"""
|
3021 |
for vg, req_size in req_sizes.items(): |
3022 |
_CheckNodesFreeDiskOnVG(lu, nodenames, vg, req_size) |
3023 |
|
3024 |
|
3025 |
def _CheckNodesFreeDiskOnVG(lu, nodenames, vg, requested): |
3026 |
"""Checks if nodes have enough free disk space in the specified VG.
|
3027 |
|
3028 |
This function checks if all given nodes have the needed amount of
|
3029 |
free disk. In case any node has less disk or we cannot get the
|
3030 |
information from the node, this function raises an OpPrereqError
|
3031 |
exception.
|
3032 |
|
3033 |
@type lu: C{LogicalUnit}
|
3034 |
@param lu: a logical unit from which we get configuration data
|
3035 |
@type nodenames: C{list}
|
3036 |
@param nodenames: the list of node names to check
|
3037 |
@type vg: C{str}
|
3038 |
@param vg: the volume group to check
|
3039 |
@type requested: C{int}
|
3040 |
@param requested: the amount of disk in MiB to check for
|
3041 |
@raise errors.OpPrereqError: if the node doesn't have enough disk,
|
3042 |
or we cannot check the node
|
3043 |
|
3044 |
"""
|
3045 |
es_flags = rpc.GetExclusiveStorageForNodeNames(lu.cfg, nodenames) |
3046 |
nodeinfo = lu.rpc.call_node_info(nodenames, [vg], None, es_flags)
|
3047 |
for node in nodenames: |
3048 |
info = nodeinfo[node] |
3049 |
info.Raise("Cannot get current information from node %s" % node,
|
3050 |
prereq=True, ecode=errors.ECODE_ENVIRON)
|
3051 |
(_, (vg_info, ), _) = info.payload |
3052 |
vg_free = vg_info.get("vg_free", None) |
3053 |
if not isinstance(vg_free, int): |
3054 |
raise errors.OpPrereqError("Can't compute free disk space on node" |
3055 |
" %s for vg %s, result was '%s'" %
|
3056 |
(node, vg, vg_free), errors.ECODE_ENVIRON) |
3057 |
if requested > vg_free:
|
3058 |
raise errors.OpPrereqError("Not enough disk space on target node %s" |
3059 |
" vg %s: required %d MiB, available %d MiB" %
|
3060 |
(node, vg, requested, vg_free), |
3061 |
errors.ECODE_NORES) |
3062 |
|
3063 |
|
3064 |
def _CheckNodesPhysicalCPUs(lu, nodenames, requested, hypervisor_name): |
3065 |
"""Checks if nodes have enough physical CPUs
|
3066 |
|
3067 |
This function checks if all given nodes have the needed number of
|
3068 |
physical CPUs. In case any node has less CPUs or we cannot get the
|
3069 |
information from the node, this function raises an OpPrereqError
|
3070 |
exception.
|
3071 |
|
3072 |
@type lu: C{LogicalUnit}
|
3073 |
@param lu: a logical unit from which we get configuration data
|
3074 |
@type nodenames: C{list}
|
3075 |
@param nodenames: the list of node names to check
|
3076 |
@type requested: C{int}
|
3077 |
@param requested: the minimum acceptable number of physical CPUs
|
3078 |
@raise errors.OpPrereqError: if the node doesn't have enough CPUs,
|
3079 |
or we cannot check the node
|
3080 |
|
3081 |
"""
|
3082 |
nodeinfo = lu.rpc.call_node_info(nodenames, None, [hypervisor_name], None) |
3083 |
for node in nodenames: |
3084 |
info = nodeinfo[node] |
3085 |
info.Raise("Cannot get current information from node %s" % node,
|
3086 |
prereq=True, ecode=errors.ECODE_ENVIRON)
|
3087 |
(_, _, (hv_info, )) = info.payload |
3088 |
num_cpus = hv_info.get("cpu_total", None) |
3089 |
if not isinstance(num_cpus, int): |
3090 |
raise errors.OpPrereqError("Can't compute the number of physical CPUs" |
3091 |
" on node %s, result was '%s'" %
|
3092 |
(node, num_cpus), errors.ECODE_ENVIRON) |
3093 |
if requested > num_cpus:
|
3094 |
raise errors.OpPrereqError("Node %s has %s physical CPUs, but %s are " |
3095 |
"required" % (node, num_cpus, requested),
|
3096 |
errors.ECODE_NORES) |
3097 |
|
3098 |
|
3099 |
class LUInstanceStartup(LogicalUnit): |
3100 |
"""Starts an instance.
|
3101 |
|
3102 |
"""
|
3103 |
HPATH = "instance-start"
|
3104 |
HTYPE = constants.HTYPE_INSTANCE |
3105 |
REQ_BGL = False
|
3106 |
|
3107 |
def CheckArguments(self): |
3108 |
# extra beparams
|
3109 |
if self.op.beparams: |
3110 |
# fill the beparams dict
|
3111 |
objects.UpgradeBeParams(self.op.beparams)
|
3112 |
utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
|
3113 |
|
3114 |
def ExpandNames(self): |
3115 |
self._ExpandAndLockInstance()
|
3116 |
self.recalculate_locks[locking.LEVEL_NODE_RES] = constants.LOCKS_REPLACE
|
3117 |
|
3118 |
def DeclareLocks(self, level): |
3119 |
if level == locking.LEVEL_NODE_RES:
|
3120 |
self._LockInstancesNodes(primary_only=True, level=locking.LEVEL_NODE_RES) |
3121 |
|
3122 |
def BuildHooksEnv(self): |
3123 |
"""Build hooks env.
|
3124 |
|
3125 |
This runs on master, primary and secondary nodes of the instance.
|
3126 |
|
3127 |
"""
|
3128 |
env = { |
3129 |
"FORCE": self.op.force, |
3130 |
} |
3131 |
|
3132 |
env.update(_BuildInstanceHookEnvByObject(self, self.instance)) |
3133 |
|
3134 |
return env
|
3135 |
|
3136 |
def BuildHooksNodes(self): |
3137 |
"""Build hooks nodes.
|
3138 |
|
3139 |
"""
|
3140 |
nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes) |
3141 |
return (nl, nl)
|
3142 |
|
3143 |
def CheckPrereq(self): |
3144 |
"""Check prerequisites.
|
3145 |
|
3146 |
This checks that the instance is in the cluster.
|
3147 |
|
3148 |
"""
|
3149 |
self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name) |
3150 |
assert self.instance is not None, \ |
3151 |
"Cannot retrieve locked instance %s" % self.op.instance_name |
3152 |
|
3153 |
# extra hvparams
|
3154 |
if self.op.hvparams: |
3155 |
# check hypervisor parameter syntax (locally)
|
3156 |
cluster = self.cfg.GetClusterInfo()
|
3157 |
utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
|
3158 |
filled_hvp = cluster.FillHV(instance) |
3159 |
filled_hvp.update(self.op.hvparams)
|
3160 |
hv_type = hypervisor.GetHypervisorClass(instance.hypervisor) |
3161 |
hv_type.CheckParameterSyntax(filled_hvp) |
3162 |
_CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp)
|
3163 |
|
3164 |
_CheckInstanceState(self, instance, INSTANCE_ONLINE)
|
3165 |
|
3166 |
self.primary_offline = self.cfg.GetNodeInfo(instance.primary_node).offline |
3167 |
|
3168 |
if self.primary_offline and self.op.ignore_offline_nodes: |
3169 |
self.LogWarning("Ignoring offline primary node") |
3170 |
|
3171 |
if self.op.hvparams or self.op.beparams: |
3172 |
self.LogWarning("Overridden parameters are ignored") |
3173 |
else:
|
3174 |
_CheckNodeOnline(self, instance.primary_node)
|
3175 |
|
3176 |
bep = self.cfg.GetClusterInfo().FillBE(instance)
|
3177 |
bep.update(self.op.beparams)
|
3178 |
|
3179 |
# check bridges existence
|
3180 |
_CheckInstanceBridgesExist(self, instance)
|
3181 |
|
3182 |
remote_info = self.rpc.call_instance_info(instance.primary_node,
|
3183 |
instance.name, |
3184 |
instance.hypervisor) |
3185 |
remote_info.Raise("Error checking node %s" % instance.primary_node,
|
3186 |
prereq=True, ecode=errors.ECODE_ENVIRON)
|
3187 |
if not remote_info.payload: # not running already |
3188 |
_CheckNodeFreeMemory(self, instance.primary_node,
|
3189 |
"starting instance %s" % instance.name,
|
3190 |
bep[constants.BE_MINMEM], instance.hypervisor) |
3191 |
|
3192 |
def Exec(self, feedback_fn): |
3193 |
"""Start the instance.
|
3194 |
|
3195 |
"""
|
3196 |
instance = self.instance
|
3197 |
force = self.op.force
|
3198 |
reason = self.op.reason
|
3199 |
|
3200 |
if not self.op.no_remember: |
3201 |
self.cfg.MarkInstanceUp(instance.name)
|
3202 |
|
3203 |
if self.primary_offline: |
3204 |
assert self.op.ignore_offline_nodes |
3205 |
self.LogInfo("Primary node offline, marked instance as started") |
3206 |
else:
|
3207 |
node_current = instance.primary_node |
3208 |
|
3209 |
_StartInstanceDisks(self, instance, force)
|
3210 |
|
3211 |
result = \ |
3212 |
self.rpc.call_instance_start(node_current,
|
3213 |
(instance, self.op.hvparams,
|
3214 |
self.op.beparams),
|
3215 |
self.op.startup_paused, reason)
|
3216 |
msg = result.fail_msg |
3217 |
if msg:
|
3218 |
_ShutdownInstanceDisks(self, instance)
|
3219 |
raise errors.OpExecError("Could not start instance: %s" % msg) |
3220 |
|
3221 |
|
3222 |
class LUInstanceReboot(LogicalUnit): |
3223 |
"""Reboot an instance.
|
3224 |
|
3225 |
"""
|
3226 |
HPATH = "instance-reboot"
|
3227 |
HTYPE = constants.HTYPE_INSTANCE |
3228 |
REQ_BGL = False
|
3229 |
|
3230 |
def ExpandNames(self): |
3231 |
self._ExpandAndLockInstance()
|
3232 |
|
3233 |
def BuildHooksEnv(self): |
3234 |
"""Build hooks env.
|
3235 |
|
3236 |
This runs on master, primary and secondary nodes of the instance.
|
3237 |
|
3238 |
"""
|
3239 |
env = { |
3240 |
"IGNORE_SECONDARIES": self.op.ignore_secondaries, |
3241 |
"REBOOT_TYPE": self.op.reboot_type, |
3242 |
"SHUTDOWN_TIMEOUT": self.op.shutdown_timeout, |
3243 |
} |
3244 |
|
3245 |
env.update(_BuildInstanceHookEnvByObject(self, self.instance)) |
3246 |
|
3247 |
return env
|
3248 |
|
3249 |
def BuildHooksNodes(self): |
3250 |
"""Build hooks nodes.
|
3251 |
|
3252 |
"""
|
3253 |
nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes) |
3254 |
return (nl, nl)
|
3255 |
|
3256 |
def CheckPrereq(self): |
3257 |
"""Check prerequisites.
|
3258 |
|
3259 |
This checks that the instance is in the cluster.
|
3260 |
|
3261 |
"""
|
3262 |
self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name) |
3263 |
assert self.instance is not None, \ |
3264 |
"Cannot retrieve locked instance %s" % self.op.instance_name |
3265 |
_CheckInstanceState(self, instance, INSTANCE_ONLINE)
|
3266 |
_CheckNodeOnline(self, instance.primary_node)
|
3267 |
|
3268 |
# check bridges existence
|
3269 |
_CheckInstanceBridgesExist(self, instance)
|
3270 |
|
3271 |
def Exec(self, feedback_fn): |
3272 |
"""Reboot the instance.
|
3273 |
|
3274 |
"""
|
3275 |
instance = self.instance
|
3276 |
ignore_secondaries = self.op.ignore_secondaries
|
3277 |
reboot_type = self.op.reboot_type
|
3278 |
reason = self.op.reason
|
3279 |
|
3280 |
remote_info = self.rpc.call_instance_info(instance.primary_node,
|
3281 |
instance.name, |
3282 |
instance.hypervisor) |
3283 |
remote_info.Raise("Error checking node %s" % instance.primary_node)
|
3284 |
instance_running = bool(remote_info.payload)
|
3285 |
|
3286 |
node_current = instance.primary_node |
3287 |
|
3288 |
if instance_running and reboot_type in [constants.INSTANCE_REBOOT_SOFT, |
3289 |
constants.INSTANCE_REBOOT_HARD]: |
3290 |
for disk in instance.disks: |
3291 |
self.cfg.SetDiskID(disk, node_current)
|
3292 |
result = self.rpc.call_instance_reboot(node_current, instance,
|
3293 |
reboot_type, |
3294 |
self.op.shutdown_timeout, reason)
|
3295 |
result.Raise("Could not reboot instance")
|
3296 |
else:
|
3297 |
if instance_running:
|
3298 |
result = self.rpc.call_instance_shutdown(node_current, instance,
|
3299 |
self.op.shutdown_timeout,
|
3300 |
reason) |
3301 |
result.Raise("Could not shutdown instance for full reboot")
|
3302 |
_ShutdownInstanceDisks(self, instance)
|
3303 |
else:
|
3304 |
self.LogInfo("Instance %s was already stopped, starting now", |
3305 |
instance.name) |
3306 |
_StartInstanceDisks(self, instance, ignore_secondaries)
|
3307 |
result = self.rpc.call_instance_start(node_current,
|
3308 |
(instance, None, None), False, |
3309 |
reason) |
3310 |
msg = result.fail_msg |
3311 |
if msg:
|
3312 |
_ShutdownInstanceDisks(self, instance)
|
3313 |
raise errors.OpExecError("Could not start instance for" |
3314 |
" full reboot: %s" % msg)
|
3315 |
|
3316 |
self.cfg.MarkInstanceUp(instance.name)
|
3317 |
|
3318 |
|
3319 |
class LUInstanceShutdown(LogicalUnit): |
3320 |
"""Shutdown an instance.
|
3321 |
|
3322 |
"""
|
3323 |
HPATH = "instance-stop"
|
3324 |
HTYPE = constants.HTYPE_INSTANCE |
3325 |
REQ_BGL = False
|
3326 |
|
3327 |
def ExpandNames(self): |
3328 |
self._ExpandAndLockInstance()
|
3329 |
|
3330 |
def BuildHooksEnv(self): |
3331 |
"""Build hooks env.
|
3332 |
|
3333 |
This runs on master, primary and secondary nodes of the instance.
|
3334 |
|
3335 |
"""
|
3336 |
env = _BuildInstanceHookEnvByObject(self, self.instance) |
3337 |
env["TIMEOUT"] = self.op.timeout |
3338 |
return env
|
3339 |
|
3340 |
def BuildHooksNodes(self): |
3341 |
"""Build hooks nodes.
|
3342 |
|
3343 |
"""
|
3344 |
nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes) |
3345 |
return (nl, nl)
|
3346 |
|
3347 |
def CheckPrereq(self): |
3348 |
"""Check prerequisites.
|
3349 |
|
3350 |
This checks that the instance is in the cluster.
|
3351 |
|
3352 |
"""
|
3353 |
self.instance = self.cfg.GetInstanceInfo(self.op.instance_name) |
3354 |
assert self.instance is not None, \ |
3355 |
"Cannot retrieve locked instance %s" % self.op.instance_name |
3356 |
|
3357 |
if not self.op.force: |
3358 |
_CheckInstanceState(self, self.instance, INSTANCE_ONLINE) |
3359 |
else:
|
3360 |
self.LogWarning("Ignoring offline instance check") |
3361 |
|
3362 |
self.primary_offline = \
|
3363 |
self.cfg.GetNodeInfo(self.instance.primary_node).offline |
3364 |
|
3365 |
if self.primary_offline and self.op.ignore_offline_nodes: |
3366 |
self.LogWarning("Ignoring offline primary node") |
3367 |
else:
|
3368 |
_CheckNodeOnline(self, self.instance.primary_node) |
3369 |
|
3370 |
def Exec(self, feedback_fn): |
3371 |
"""Shutdown the instance.
|
3372 |
|
3373 |
"""
|
3374 |
instance = self.instance
|
3375 |
node_current = instance.primary_node |
3376 |
timeout = self.op.timeout
|
3377 |
reason = self.op.reason
|
3378 |
|
3379 |
# If the instance is offline we shouldn't mark it as down, as that
|
3380 |
# resets the offline flag.
|
3381 |
if not self.op.no_remember and instance.admin_state in INSTANCE_ONLINE: |
3382 |
self.cfg.MarkInstanceDown(instance.name)
|
3383 |
|
3384 |
if self.primary_offline: |
3385 |
assert self.op.ignore_offline_nodes |
3386 |
self.LogInfo("Primary node offline, marked instance as stopped") |
3387 |
else:
|
3388 |
result = self.rpc.call_instance_shutdown(node_current, instance, timeout,
|
3389 |
reason) |
3390 |
msg = result.fail_msg |
3391 |
if msg:
|
3392 |
self.LogWarning("Could not shutdown instance: %s", msg) |
3393 |
|
3394 |
_ShutdownInstanceDisks(self, instance)
|
3395 |
|
3396 |
|
3397 |
class LUInstanceReinstall(LogicalUnit): |
3398 |
"""Reinstall an instance.
|
3399 |
|
3400 |
"""
|
3401 |
HPATH = "instance-reinstall"
|
3402 |
HTYPE = constants.HTYPE_INSTANCE |
3403 |
REQ_BGL = False
|
3404 |
|
3405 |
def ExpandNames(self): |
3406 |
self._ExpandAndLockInstance()
|
3407 |
|
3408 |
def BuildHooksEnv(self): |
3409 |
"""Build hooks env.
|
3410 |
|
3411 |
This runs on master, primary and secondary nodes of the instance.
|
3412 |
|
3413 |
"""
|
3414 |
return _BuildInstanceHookEnvByObject(self, self.instance) |
3415 |
|
3416 |
def BuildHooksNodes(self): |
3417 |
"""Build hooks nodes.
|
3418 |
|
3419 |
"""
|
3420 |
nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes) |
3421 |
return (nl, nl)
|
3422 |
|
3423 |
def CheckPrereq(self): |
3424 |
"""Check prerequisites.
|
3425 |
|
3426 |
This checks that the instance is in the cluster and is not running.
|
3427 |
|
3428 |
"""
|
3429 |
instance = self.cfg.GetInstanceInfo(self.op.instance_name) |
3430 |
assert instance is not None, \ |
3431 |
"Cannot retrieve locked instance %s" % self.op.instance_name |
3432 |
_CheckNodeOnline(self, instance.primary_node, "Instance primary node" |
3433 |
" offline, cannot reinstall")
|
3434 |
|
3435 |
if instance.disk_template == constants.DT_DISKLESS:
|
3436 |
raise errors.OpPrereqError("Instance '%s' has no disks" % |
3437 |
self.op.instance_name,
|
3438 |
errors.ECODE_INVAL) |
3439 |
_CheckInstanceState(self, instance, INSTANCE_DOWN, msg="cannot reinstall") |
3440 |
|
3441 |
if self.op.os_type is not None: |
3442 |
# OS verification
|
3443 |
pnode = _ExpandNodeName(self.cfg, instance.primary_node)
|
3444 |
_CheckNodeHasOS(self, pnode, self.op.os_type, self.op.force_variant) |
3445 |
instance_os = self.op.os_type
|
3446 |
else:
|
3447 |
instance_os = instance.os |
3448 |
|
3449 |
nodelist = list(instance.all_nodes)
|
3450 |
|
3451 |
if self.op.osparams: |
3452 |
i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
|
3453 |
_CheckOSParams(self, True, nodelist, instance_os, i_osdict) |
3454 |
self.os_inst = i_osdict # the new dict (without defaults) |
3455 |
else:
|
3456 |
self.os_inst = None |
3457 |
|
3458 |
self.instance = instance
|
3459 |
|
3460 |
def Exec(self, feedback_fn): |
3461 |
"""Reinstall the instance.
|
3462 |
|
3463 |
"""
|
3464 |
inst = self.instance
|
3465 |
|
3466 |
if self.op.os_type is not None: |
3467 |
feedback_fn("Changing OS to '%s'..." % self.op.os_type) |
3468 |
inst.os = self.op.os_type
|
3469 |
# Write to configuration
|
3470 |
self.cfg.Update(inst, feedback_fn)
|
3471 |
|
3472 |
_StartInstanceDisks(self, inst, None) |
3473 |
try:
|
3474 |
feedback_fn("Running the instance OS create scripts...")
|
3475 |
# FIXME: pass debug option from opcode to backend
|
3476 |
result = self.rpc.call_instance_os_add(inst.primary_node,
|
3477 |
(inst, self.os_inst), True, |
3478 |
self.op.debug_level)
|
3479 |
result.Raise("Could not install OS for instance %s on node %s" %
|
3480 |
(inst.name, inst.primary_node)) |
3481 |
finally:
|
3482 |
_ShutdownInstanceDisks(self, inst)
|
3483 |
|
3484 |
|
3485 |
class LUInstanceRecreateDisks(LogicalUnit): |
3486 |
"""Recreate an instance's missing disks.
|
3487 |
|
3488 |
"""
|
3489 |
HPATH = "instance-recreate-disks"
|
3490 |
HTYPE = constants.HTYPE_INSTANCE |
3491 |
REQ_BGL = False
|
3492 |
|
3493 |
_MODIFYABLE = compat.UniqueFrozenset([ |
3494 |
constants.IDISK_SIZE, |
3495 |
constants.IDISK_MODE, |
3496 |
]) |
3497 |
|
3498 |
# New or changed disk parameters may have different semantics
|
3499 |
assert constants.IDISK_PARAMS == (_MODIFYABLE | frozenset([ |
3500 |
constants.IDISK_ADOPT, |
3501 |
|
3502 |
# TODO: Implement support changing VG while recreating
|
3503 |
constants.IDISK_VG, |
3504 |
constants.IDISK_METAVG, |
3505 |
constants.IDISK_PROVIDER, |
3506 |
constants.IDISK_NAME, |
3507 |
])) |
3508 |
|
3509 |
def _RunAllocator(self): |
3510 |
"""Run the allocator based on input opcode.
|
3511 |
|
3512 |
"""
|
3513 |
be_full = self.cfg.GetClusterInfo().FillBE(self.instance) |
3514 |
|
3515 |
# FIXME
|
3516 |
# The allocator should actually run in "relocate" mode, but current
|
3517 |
# allocators don't support relocating all the nodes of an instance at
|
3518 |
# the same time. As a workaround we use "allocate" mode, but this is
|
3519 |
# suboptimal for two reasons:
|
3520 |
# - The instance name passed to the allocator is present in the list of
|
3521 |
# existing instances, so there could be a conflict within the
|
3522 |
# internal structures of the allocator. This doesn't happen with the
|
3523 |
# current allocators, but it's a liability.
|
3524 |
# - The allocator counts the resources used by the instance twice: once
|
3525 |
# because the instance exists already, and once because it tries to
|
3526 |
# allocate a new instance.
|
3527 |
# The allocator could choose some of the nodes on which the instance is
|
3528 |
# running, but that's not a problem. If the instance nodes are broken,
|
3529 |
# they should be already be marked as drained or offline, and hence
|
3530 |
# skipped by the allocator. If instance disks have been lost for other
|
3531 |
# reasons, then recreating the disks on the same nodes should be fine.
|
3532 |
disk_template = self.instance.disk_template
|
3533 |
spindle_use = be_full[constants.BE_SPINDLE_USE] |
3534 |
req = iallocator.IAReqInstanceAlloc(name=self.op.instance_name,
|
3535 |
disk_template=disk_template, |
3536 |
tags=list(self.instance.GetTags()), |
3537 |
os=self.instance.os,
|
3538 |
nics=[{}], |
3539 |
vcpus=be_full[constants.BE_VCPUS], |
3540 |
memory=be_full[constants.BE_MAXMEM], |
3541 |
spindle_use=spindle_use, |
3542 |
disks=[{constants.IDISK_SIZE: d.size, |
3543 |
constants.IDISK_MODE: d.mode} |
3544 |
for d in self.instance.disks], |
3545 |
hypervisor=self.instance.hypervisor,
|
3546 |
node_whitelist=None)
|
3547 |
ial = iallocator.IAllocator(self.cfg, self.rpc, req) |
3548 |
|
3549 |
ial.Run(self.op.iallocator)
|
3550 |
|
3551 |
assert req.RequiredNodes() == len(self.instance.all_nodes) |
3552 |
|
3553 |
if not ial.success: |
3554 |
raise errors.OpPrereqError("Can't compute nodes using iallocator '%s':" |
3555 |
" %s" % (self.op.iallocator, ial.info), |
3556 |
errors.ECODE_NORES) |
3557 |
|
3558 |
self.op.nodes = ial.result
|
3559 |
self.LogInfo("Selected nodes for instance %s via iallocator %s: %s", |
3560 |
self.op.instance_name, self.op.iallocator, |
3561 |
utils.CommaJoin(ial.result)) |
3562 |
|
3563 |
def CheckArguments(self): |
3564 |
if self.op.disks and ht.TNonNegativeInt(self.op.disks[0]): |
3565 |
# Normalize and convert deprecated list of disk indices
|
3566 |
self.op.disks = [(idx, {}) for idx in sorted(frozenset(self.op.disks))] |
3567 |
|
3568 |
duplicates = utils.FindDuplicates(map(compat.fst, self.op.disks)) |
3569 |
if duplicates:
|
3570 |
raise errors.OpPrereqError("Some disks have been specified more than" |
3571 |
" once: %s" % utils.CommaJoin(duplicates),
|
3572 |
errors.ECODE_INVAL) |
3573 |
|
3574 |
# We don't want _CheckIAllocatorOrNode selecting the default iallocator
|
3575 |
# when neither iallocator nor nodes are specified
|
3576 |
if self.op.iallocator or self.op.nodes: |
3577 |
_CheckIAllocatorOrNode(self, "iallocator", "nodes") |
3578 |
|
3579 |
for (idx, params) in self.op.disks: |
3580 |
utils.ForceDictType(params, constants.IDISK_PARAMS_TYPES) |
3581 |
unsupported = frozenset(params.keys()) - self._MODIFYABLE |
3582 |
if unsupported:
|
3583 |
raise errors.OpPrereqError("Parameters for disk %s try to change" |
3584 |
" unmodifyable parameter(s): %s" %
|
3585 |
(idx, utils.CommaJoin(unsupported)), |
3586 |
errors.ECODE_INVAL) |
3587 |
|
3588 |
def ExpandNames(self): |
3589 |
self._ExpandAndLockInstance()
|
3590 |
self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
|
3591 |
|
3592 |
if self.op.nodes: |
3593 |
self.op.nodes = [_ExpandNodeName(self.cfg, n) for n in self.op.nodes] |
3594 |
self.needed_locks[locking.LEVEL_NODE] = list(self.op.nodes) |
3595 |
else:
|
3596 |
self.needed_locks[locking.LEVEL_NODE] = []
|
3597 |
if self.op.iallocator: |
3598 |
# iallocator will select a new node in the same group
|
3599 |
self.needed_locks[locking.LEVEL_NODEGROUP] = []
|
3600 |
self.needed_locks[locking.LEVEL_NODE_ALLOC] = locking.ALL_SET
|
3601 |
|
3602 |
self.needed_locks[locking.LEVEL_NODE_RES] = []
|
3603 |
|
3604 |
def DeclareLocks(self, level): |
3605 |
if level == locking.LEVEL_NODEGROUP:
|
3606 |
assert self.op.iallocator is not None |
3607 |
assert not self.op.nodes |
3608 |
assert not self.needed_locks[locking.LEVEL_NODEGROUP] |
3609 |
self.share_locks[locking.LEVEL_NODEGROUP] = 1 |
3610 |
# Lock the primary group used by the instance optimistically; this
|
3611 |
# requires going via the node before it's locked, requiring
|
3612 |
# verification later on
|
3613 |
self.needed_locks[locking.LEVEL_NODEGROUP] = \
|
3614 |
self.cfg.GetInstanceNodeGroups(self.op.instance_name, primary_only=True) |
3615 |
|
3616 |
elif level == locking.LEVEL_NODE:
|
3617 |
# If an allocator is used, then we lock all the nodes in the current
|
3618 |
# instance group, as we don't know yet which ones will be selected;
|
3619 |
# if we replace the nodes without using an allocator, locks are
|
3620 |
# already declared in ExpandNames; otherwise, we need to lock all the
|
3621 |
# instance nodes for disk re-creation
|
3622 |
if self.op.iallocator: |
3623 |
assert not self.op.nodes |
3624 |
assert not self.needed_locks[locking.LEVEL_NODE] |
3625 |
assert len(self.owned_locks(locking.LEVEL_NODEGROUP)) == 1 |
3626 |
|
3627 |
# Lock member nodes of the group of the primary node
|
3628 |
for group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP): |
3629 |
self.needed_locks[locking.LEVEL_NODE].extend(
|
3630 |
self.cfg.GetNodeGroup(group_uuid).members)
|
3631 |
|
3632 |
assert locking.NAL in self.owned_locks(locking.LEVEL_NODE_ALLOC) |
3633 |
elif not self.op.nodes: |
3634 |
self._LockInstancesNodes(primary_only=False) |
3635 |
elif level == locking.LEVEL_NODE_RES:
|
3636 |
# Copy node locks
|
3637 |
self.needed_locks[locking.LEVEL_NODE_RES] = \
|
3638 |
_CopyLockList(self.needed_locks[locking.LEVEL_NODE])
|
3639 |
|
3640 |
def BuildHooksEnv(self): |
3641 |
"""Build hooks env.
|
3642 |
|
3643 |
This runs on master, primary and secondary nodes of the instance.
|
3644 |
|
3645 |
"""
|
3646 |
return _BuildInstanceHookEnvByObject(self, self.instance) |
3647 |
|
3648 |
def BuildHooksNodes(self): |
3649 |
"""Build hooks nodes.
|
3650 |
|
3651 |
"""
|
3652 |
nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes) |
3653 |
return (nl, nl)
|
3654 |
|
3655 |
def CheckPrereq(self): |
3656 |
"""Check prerequisites.
|
3657 |
|
3658 |
This checks that the instance is in the cluster and is not running.
|
3659 |
|
3660 |
"""
|
3661 |
instance = self.cfg.GetInstanceInfo(self.op.instance_name) |
3662 |
assert instance is not None, \ |
3663 |
"Cannot retrieve locked instance %s" % self.op.instance_name |
3664 |
if self.op.nodes: |
3665 |
if len(self.op.nodes) != len(instance.all_nodes): |
3666 |
raise errors.OpPrereqError("Instance %s currently has %d nodes, but" |
3667 |
" %d replacement nodes were specified" %
|
3668 |
(instance.name, len(instance.all_nodes),
|
3669 |
len(self.op.nodes)), |
3670 |
errors.ECODE_INVAL) |
3671 |
assert instance.disk_template != constants.DT_DRBD8 or \ |
3672 |
len(self.op.nodes) == 2 |
3673 |
assert instance.disk_template != constants.DT_PLAIN or \ |
3674 |
len(self.op.nodes) == 1 |
3675 |
primary_node = self.op.nodes[0] |
3676 |
else:
|
3677 |
primary_node = instance.primary_node |
3678 |
if not self.op.iallocator: |
3679 |
_CheckNodeOnline(self, primary_node)
|
3680 |
|
3681 |
if instance.disk_template == constants.DT_DISKLESS:
|
3682 |
raise errors.OpPrereqError("Instance '%s' has no disks" % |
3683 |
self.op.instance_name, errors.ECODE_INVAL)
|
3684 |
|
3685 |
# Verify if node group locks are still correct
|
3686 |
owned_groups = self.owned_locks(locking.LEVEL_NODEGROUP)
|
3687 |
if owned_groups:
|
3688 |
# Node group locks are acquired only for the primary node (and only
|
3689 |
# when the allocator is used)
|
3690 |
_CheckInstanceNodeGroups(self.cfg, self.op.instance_name, owned_groups, |
3691 |
primary_only=True)
|
3692 |
|
3693 |
# if we replace nodes *and* the old primary is offline, we don't
|
3694 |
# check the instance state
|
3695 |
old_pnode = self.cfg.GetNodeInfo(instance.primary_node)
|
3696 |
if not ((self.op.iallocator or self.op.nodes) and old_pnode.offline): |
3697 |
_CheckInstanceState(self, instance, INSTANCE_NOT_RUNNING,
|
3698 |
msg="cannot recreate disks")
|
3699 |
|
3700 |
if self.op.disks: |
3701 |
self.disks = dict(self.op.disks) |
3702 |
else:
|
3703 |
self.disks = dict((idx, {}) for idx in range(len(instance.disks))) |
3704 |
|
3705 |
maxidx = max(self.disks.keys()) |
3706 |
if maxidx >= len(instance.disks): |
3707 |
raise errors.OpPrereqError("Invalid disk index '%s'" % maxidx, |
3708 |
errors.ECODE_INVAL) |
3709 |
|
3710 |
if ((self.op.nodes or self.op.iallocator) and |
3711 |
sorted(self.disks.keys()) != range(len(instance.disks))): |
3712 |
raise errors.OpPrereqError("Can't recreate disks partially and" |
3713 |
" change the nodes at the same time",
|
3714 |
errors.ECODE_INVAL) |
3715 |
|
3716 |
self.instance = instance
|
3717 |
|
3718 |
if self.op.iallocator: |
3719 |
self._RunAllocator()
|
3720 |
# Release unneeded node and node resource locks
|
3721 |
_ReleaseLocks(self, locking.LEVEL_NODE, keep=self.op.nodes) |
3722 |
_ReleaseLocks(self, locking.LEVEL_NODE_RES, keep=self.op.nodes) |
3723 |
_ReleaseLocks(self, locking.LEVEL_NODE_ALLOC)
|
3724 |
|
3725 |
assert not self.glm.is_owned(locking.LEVEL_NODE_ALLOC) |
3726 |
|
3727 |
def Exec(self, feedback_fn): |
3728 |
"""Recreate the disks.
|
3729 |
|
3730 |
"""
|
3731 |
instance = self.instance
|
3732 |
|
3733 |
assert (self.owned_locks(locking.LEVEL_NODE) == |
3734 |
self.owned_locks(locking.LEVEL_NODE_RES))
|
3735 |
|
3736 |
to_skip = [] |
3737 |
mods = [] # keeps track of needed changes
|
3738 |
|
3739 |
for idx, disk in enumerate(instance.disks): |
3740 |
try:
|
3741 |
changes = self.disks[idx]
|
3742 |
except KeyError: |
3743 |
# Disk should not be recreated
|
3744 |
to_skip.append(idx) |
3745 |
continue
|
3746 |
|
3747 |
# update secondaries for disks, if needed
|
3748 |
if self.op.nodes and disk.dev_type == constants.LD_DRBD8: |
3749 |
# need to update the nodes and minors
|
3750 |
assert len(self.op.nodes) == 2 |
3751 |
assert len(disk.logical_id) == 6 # otherwise disk internals |
3752 |
# have changed
|
3753 |
(_, _, old_port, _, _, old_secret) = disk.logical_id |
3754 |
new_minors = self.cfg.AllocateDRBDMinor(self.op.nodes, instance.name) |
3755 |
new_id = (self.op.nodes[0], self.op.nodes[1], old_port, |
3756 |
new_minors[0], new_minors[1], old_secret) |
3757 |
assert len(disk.logical_id) == len(new_id) |
3758 |
else:
|
3759 |
new_id = None
|
3760 |
|
3761 |
mods.append((idx, new_id, changes)) |
3762 |
|
3763 |
# now that we have passed all asserts above, we can apply the mods
|
3764 |
# in a single run (to avoid partial changes)
|
3765 |
for idx, new_id, changes in mods: |
3766 |
disk = instance.disks[idx] |
3767 |
if new_id is not None: |
3768 |
assert disk.dev_type == constants.LD_DRBD8
|
3769 |
disk.logical_id = new_id |
3770 |
if changes:
|
3771 |
disk.Update(size=changes.get(constants.IDISK_SIZE, None),
|
3772 |
mode=changes.get(constants.IDISK_MODE, None))
|
3773 |
|
3774 |
# change primary node, if needed
|
3775 |
if self.op.nodes: |
3776 |
instance.primary_node = self.op.nodes[0] |
3777 |
self.LogWarning("Changing the instance's nodes, you will have to" |
3778 |
" remove any disks left on the older nodes manually")
|
3779 |
|
3780 |
if self.op.nodes: |
3781 |
self.cfg.Update(instance, feedback_fn)
|
3782 |
|
3783 |
# All touched nodes must be locked
|
3784 |
mylocks = self.owned_locks(locking.LEVEL_NODE)
|
3785 |
assert mylocks.issuperset(frozenset(instance.all_nodes)) |
3786 |
_CreateDisks(self, instance, to_skip=to_skip)
|
3787 |
|
3788 |
|
3789 |
class LUInstanceRename(LogicalUnit): |
3790 |
"""Rename an instance.
|
3791 |
|
3792 |
"""
|
3793 |
HPATH = "instance-rename"
|
3794 |
HTYPE = constants.HTYPE_INSTANCE |
3795 |
|
3796 |
def CheckArguments(self): |
3797 |
"""Check arguments.
|
3798 |
|
3799 |
"""
|
3800 |
if self.op.ip_check and not self.op.name_check: |
3801 |
# TODO: make the ip check more flexible and not depend on the name check
|
3802 |
raise errors.OpPrereqError("IP address check requires a name check", |
3803 |
errors.ECODE_INVAL) |
3804 |
|
3805 |
def BuildHooksEnv(self): |
3806 |
"""Build hooks env.
|
3807 |
|
3808 |
This runs on master, primary and secondary nodes of the instance.
|
3809 |
|
3810 |
"""
|
3811 |
env = _BuildInstanceHookEnvByObject(self, self.instance) |
3812 |
env["INSTANCE_NEW_NAME"] = self.op.new_name |
3813 |
return env
|
3814 |
|
3815 |
def BuildHooksNodes(self): |
3816 |
"""Build hooks nodes.
|
3817 |
|
3818 |
"""
|
3819 |
nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes) |
3820 |
return (nl, nl)
|
3821 |
|
3822 |
def CheckPrereq(self): |
3823 |
"""Check prerequisites.
|
3824 |
|
3825 |
This checks that the instance is in the cluster and is not running.
|
3826 |
|
3827 |
"""
|
3828 |
self.op.instance_name = _ExpandInstanceName(self.cfg, |
3829 |
self.op.instance_name)
|
3830 |
instance = self.cfg.GetInstanceInfo(self.op.instance_name) |
3831 |
assert instance is not None |
3832 |
_CheckNodeOnline(self, instance.primary_node)
|
3833 |
_CheckInstanceState(self, instance, INSTANCE_NOT_RUNNING,
|
3834 |
msg="cannot rename")
|
3835 |
self.instance = instance
|
3836 |
|
3837 |
new_name = self.op.new_name
|
3838 |
if self.op.name_check: |
3839 |
hostname = _CheckHostnameSane(self, new_name)
|
3840 |
new_name = self.op.new_name = hostname.name
|
3841 |
if (self.op.ip_check and |
3842 |
netutils.TcpPing(hostname.ip, constants.DEFAULT_NODED_PORT)): |
3843 |
raise errors.OpPrereqError("IP %s of instance %s already in use" % |
3844 |
(hostname.ip, new_name), |
3845 |
errors.ECODE_NOTUNIQUE) |
3846 |
|
3847 |
instance_list = self.cfg.GetInstanceList()
|
3848 |
if new_name in instance_list and new_name != instance.name: |
3849 |
raise errors.OpPrereqError("Instance '%s' is already in the cluster" % |
3850 |
new_name, errors.ECODE_EXISTS) |
3851 |
|
3852 |
def Exec(self, feedback_fn): |
3853 |
"""Rename the instance.
|
3854 |
|
3855 |
"""
|
3856 |
inst = self.instance
|
3857 |
old_name = inst.name |
3858 |
|
3859 |
rename_file_storage = False
|
3860 |
if (inst.disk_template in constants.DTS_FILEBASED and |
3861 |
self.op.new_name != inst.name):
|
3862 |
old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1]) |
3863 |
rename_file_storage = True
|
3864 |
|
3865 |
self.cfg.RenameInstance(inst.name, self.op.new_name) |
3866 |
# Change the instance lock. This is definitely safe while we hold the BGL.
|
3867 |
# Otherwise the new lock would have to be added in acquired mode.
|
3868 |
assert self.REQ_BGL |
3869 |
assert locking.BGL in self.owned_locks(locking.LEVEL_CLUSTER) |
3870 |
self.glm.remove(locking.LEVEL_INSTANCE, old_name)
|
3871 |
self.glm.add(locking.LEVEL_INSTANCE, self.op.new_name) |
3872 |
|
3873 |
# re-read the instance from the configuration after rename
|
3874 |
inst = self.cfg.GetInstanceInfo(self.op.new_name) |
3875 |
|
3876 |
if rename_file_storage:
|
3877 |
new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1]) |
3878 |
result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
|
3879 |
old_file_storage_dir, |
3880 |
new_file_storage_dir) |
3881 |
result.Raise("Could not rename on node %s directory '%s' to '%s'"
|
3882 |
" (but the instance has been renamed in Ganeti)" %
|
3883 |
(inst.primary_node, old_file_storage_dir, |
3884 |
new_file_storage_dir)) |
3885 |
|
3886 |
_StartInstanceDisks(self, inst, None) |
3887 |
# update info on disks
|
3888 |
info = _GetInstanceInfoText(inst) |
3889 |
for (idx, disk) in enumerate(inst.disks): |
3890 |
for node in inst.all_nodes: |
3891 |
self.cfg.SetDiskID(disk, node)
|
3892 |
result = self.rpc.call_blockdev_setinfo(node, disk, info)
|
3893 |
if result.fail_msg:
|
3894 |
self.LogWarning("Error setting info on node %s for disk %s: %s", |
3895 |
node, idx, result.fail_msg) |
3896 |
try:
|
3897 |
result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
|
3898 |
old_name, self.op.debug_level)
|
3899 |
msg = result.fail_msg |
3900 |
if msg:
|
3901 |
msg = ("Could not run OS rename script for instance %s on node %s"
|
3902 |
" (but the instance has been renamed in Ganeti): %s" %
|
3903 |
(inst.name, inst.primary_node, msg)) |
3904 |
self.LogWarning(msg)
|
3905 |
finally:
|
3906 |
_ShutdownInstanceDisks(self, inst)
|
3907 |
|
3908 |
return inst.name
|
3909 |
|
3910 |
|
3911 |
class LUInstanceRemove(LogicalUnit): |
3912 |
"""Remove an instance.
|
3913 |
|
3914 |
"""
|
3915 |
HPATH = "instance-remove"
|
3916 |
HTYPE = constants.HTYPE_INSTANCE |
3917 |
REQ_BGL = False
|
3918 |
|
3919 |
def ExpandNames(self): |
3920 |
self._ExpandAndLockInstance()
|
3921 |
self.needed_locks[locking.LEVEL_NODE] = []
|
3922 |
self.needed_locks[locking.LEVEL_NODE_RES] = []
|
3923 |
self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
|
3924 |
|
3925 |
def DeclareLocks(self, level): |
3926 |
if level == locking.LEVEL_NODE:
|
3927 |
self._LockInstancesNodes()
|
3928 |
elif level == locking.LEVEL_NODE_RES:
|
3929 |
# Copy node locks
|
3930 |
self.needed_locks[locking.LEVEL_NODE_RES] = \
|
3931 |
_CopyLockList(self.needed_locks[locking.LEVEL_NODE])
|
3932 |
|
3933 |
def BuildHooksEnv(self): |
3934 |
"""Build hooks env.
|
3935 |
|
3936 |
This runs on master, primary and secondary nodes of the instance.
|
3937 |
|
3938 |
"""
|
3939 |
env = _BuildInstanceHookEnvByObject(self, self.instance) |
3940 |
env["SHUTDOWN_TIMEOUT"] = self.op.shutdown_timeout |
3941 |
return env
|
3942 |
|
3943 |
def BuildHooksNodes(self): |
3944 |
"""Build hooks nodes.
|
3945 |
|
3946 |
"""
|
3947 |
nl = [self.cfg.GetMasterNode()]
|
3948 |
nl_post = list(self.instance.all_nodes) + nl |
3949 |
return (nl, nl_post)
|
3950 |
|
3951 |
def CheckPrereq(self): |
3952 |
"""Check prerequisites.
|
3953 |
|
3954 |
This checks that the instance is in the cluster.
|
3955 |
|
3956 |
"""
|
3957 |
self.instance = self.cfg.GetInstanceInfo(self.op.instance_name) |
3958 |
assert self.instance is not None, \ |
3959 |
"Cannot retrieve locked instance %s" % self.op.instance_name |
3960 |
|
3961 |
def Exec(self, feedback_fn): |
3962 |
"""Remove the instance.
|
3963 |
|
3964 |
"""
|
3965 |
instance = self.instance
|
3966 |
logging.info("Shutting down instance %s on node %s",
|
3967 |
instance.name, instance.primary_node) |
3968 |
|
3969 |
result = self.rpc.call_instance_shutdown(instance.primary_node, instance,
|
3970 |
self.op.shutdown_timeout,
|
3971 |
self.op.reason)
|
3972 |
msg = result.fail_msg |
3973 |
if msg:
|
3974 |
if self.op.ignore_failures: |
3975 |
feedback_fn("Warning: can't shutdown instance: %s" % msg)
|
3976 |
else:
|
3977 |
raise errors.OpExecError("Could not shutdown instance %s on" |
3978 |
" node %s: %s" %
|
3979 |
(instance.name, instance.primary_node, msg)) |
3980 |
|
3981 |
assert (self.owned_locks(locking.LEVEL_NODE) == |
3982 |
self.owned_locks(locking.LEVEL_NODE_RES))
|
3983 |
assert not (set(instance.all_nodes) - |
3984 |
self.owned_locks(locking.LEVEL_NODE)), \
|
3985 |
"Not owning correct locks"
|
3986 |
|
3987 |
_RemoveInstance(self, feedback_fn, instance, self.op.ignore_failures) |
3988 |
|
3989 |
|
3990 |
def _RemoveInstance(lu, feedback_fn, instance, ignore_failures): |
3991 |
"""Utility function to remove an instance.
|
3992 |
|
3993 |
"""
|
3994 |
logging.info("Removing block devices for instance %s", instance.name)
|
3995 |
|
3996 |
if not _RemoveDisks(lu, instance, ignore_failures=ignore_failures): |
3997 |
if not ignore_failures: |
3998 |
raise errors.OpExecError("Can't remove instance's disks") |
3999 |
feedback_fn("Warning: can't remove instance's disks")
|
4000 |
|
4001 |
logging.info("Removing instance %s out of cluster config", instance.name)
|
4002 |
|
4003 |
lu.cfg.RemoveInstance(instance.name) |
4004 |
|
4005 |
assert not lu.remove_locks.get(locking.LEVEL_INSTANCE), \ |
4006 |
"Instance lock removal conflict"
|
4007 |
|
4008 |
# Remove lock for the instance
|
4009 |
lu.remove_locks[locking.LEVEL_INSTANCE] = instance.name |
4010 |
|
4011 |
|
4012 |
class LUInstanceQuery(NoHooksLU): |
4013 |
"""Logical unit for querying instances.
|
4014 |
|
4015 |
"""
|
4016 |
# pylint: disable=W0142
|
4017 |
REQ_BGL = False
|
4018 |
|
4019 |
def CheckArguments(self): |
4020 |
self.iq = _InstanceQuery(qlang.MakeSimpleFilter("name", self.op.names), |
4021 |
self.op.output_fields, self.op.use_locking) |
4022 |
|
4023 |
def ExpandNames(self): |
4024 |
self.iq.ExpandNames(self) |
4025 |
|
4026 |
def DeclareLocks(self, level): |
4027 |
self.iq.DeclareLocks(self, level) |
4028 |
|
4029 |
def Exec(self, feedback_fn): |
4030 |
return self.iq.OldStyleQuery(self) |
4031 |
|
4032 |
|
4033 |
def _ExpandNamesForMigration(lu): |
4034 |
"""Expands names for use with L{TLMigrateInstance}.
|
4035 |
|
4036 |
@type lu: L{LogicalUnit}
|
4037 |
|
4038 |
"""
|
4039 |
if lu.op.target_node is not None: |
4040 |
lu.op.target_node = _ExpandNodeName(lu.cfg, lu.op.target_node) |
4041 |
|
4042 |
lu.needed_locks[locking.LEVEL_NODE] = [] |
4043 |
lu.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE |
4044 |
|
4045 |
lu.needed_locks[locking.LEVEL_NODE_RES] = [] |
4046 |
lu.recalculate_locks[locking.LEVEL_NODE_RES] = constants.LOCKS_REPLACE |
4047 |
|
4048 |
# The node allocation lock is actually only needed for externally replicated
|
4049 |
# instances (e.g. sharedfile or RBD) and if an iallocator is used.
|
4050 |
lu.needed_locks[locking.LEVEL_NODE_ALLOC] = [] |
4051 |
|
4052 |
|
4053 |
def _DeclareLocksForMigration(lu, level): |
4054 |
"""Declares locks for L{TLMigrateInstance}.
|
4055 |
|
4056 |
@type lu: L{LogicalUnit}
|
4057 |
@param level: Lock level
|
4058 |
|
4059 |
"""
|
4060 |
if level == locking.LEVEL_NODE_ALLOC:
|
4061 |
assert lu.op.instance_name in lu.owned_locks(locking.LEVEL_INSTANCE) |
4062 |
|
4063 |
instance = lu.cfg.GetInstanceInfo(lu.op.instance_name) |
4064 |
|
4065 |
# Node locks are already declared here rather than at LEVEL_NODE as we need
|
4066 |
# the instance object anyway to declare the node allocation lock.
|
4067 |
if instance.disk_template in constants.DTS_EXT_MIRROR: |
4068 |
if lu.op.target_node is None: |
4069 |
lu.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET |
4070 |
lu.needed_locks[locking.LEVEL_NODE_ALLOC] = locking.ALL_SET |
4071 |
else:
|
4072 |
lu.needed_locks[locking.LEVEL_NODE] = [instance.primary_node, |
4073 |
lu.op.target_node] |
4074 |
del lu.recalculate_locks[locking.LEVEL_NODE]
|
4075 |
else:
|
4076 |
lu._LockInstancesNodes() # pylint: disable=W0212
|
4077 |
|
4078 |
elif level == locking.LEVEL_NODE:
|
4079 |
# Node locks are declared together with the node allocation lock
|
4080 |
assert (lu.needed_locks[locking.LEVEL_NODE] or |
4081 |
lu.needed_locks[locking.LEVEL_NODE] is locking.ALL_SET)
|
4082 |
|
4083 |
elif level == locking.LEVEL_NODE_RES:
|
4084 |
# Copy node locks
|
4085 |
lu.needed_locks[locking.LEVEL_NODE_RES] = \ |
4086 |
_CopyLockList(lu.needed_locks[locking.LEVEL_NODE]) |
4087 |
|
4088 |
|
4089 |
class LUInstanceFailover(LogicalUnit): |
4090 |
"""Failover an instance.
|
4091 |
|
4092 |
"""
|
4093 |
HPATH = "instance-failover"
|
4094 |
HTYPE = constants.HTYPE_INSTANCE |
4095 |
REQ_BGL = False
|
4096 |
|
4097 |
def CheckArguments(self): |
4098 |
"""Check the arguments.
|
4099 |
|
4100 |
"""
|
4101 |
self.iallocator = getattr(self.op, "iallocator", None) |
4102 |
self.target_node = getattr(self.op, "target_node", None) |
4103 |
|
4104 |
def ExpandNames(self): |
4105 |
self._ExpandAndLockInstance()
|
4106 |
_ExpandNamesForMigration(self)
|
4107 |
|
4108 |
self._migrater = \
|
4109 |
TLMigrateInstance(self, self.op.instance_name, False, True, False, |
4110 |
self.op.ignore_consistency, True, |
4111 |
self.op.shutdown_timeout, self.op.ignore_ipolicy) |
4112 |
|
4113 |
self.tasklets = [self._migrater] |
4114 |
|
4115 |
def DeclareLocks(self, level): |
4116 |
_DeclareLocksForMigration(self, level)
|
4117 |
|
4118 |
def BuildHooksEnv(self): |
4119 |
"""Build hooks env.
|
4120 |
|
4121 |
This runs on master, primary and secondary nodes of the instance.
|
4122 |
|
4123 |
"""
|
4124 |
instance = self._migrater.instance
|
4125 |
source_node = instance.primary_node |
4126 |
target_node = self.op.target_node
|
4127 |
env = { |
4128 |
"IGNORE_CONSISTENCY": self.op.ignore_consistency, |
4129 |
"SHUTDOWN_TIMEOUT": self.op.shutdown_timeout, |
4130 |
"OLD_PRIMARY": source_node,
|
4131 |
"NEW_PRIMARY": target_node,
|
4132 |
} |
4133 |
|
4134 |
if instance.disk_template in constants.DTS_INT_MIRROR: |
4135 |
env["OLD_SECONDARY"] = instance.secondary_nodes[0] |
4136 |
env["NEW_SECONDARY"] = source_node
|
4137 |
else:
|
4138 |
env["OLD_SECONDARY"] = env["NEW_SECONDARY"] = "" |
4139 |
|
4140 |
env.update(_BuildInstanceHookEnvByObject(self, instance))
|
4141 |
|
4142 |
return env
|
4143 |
|
4144 |
def BuildHooksNodes(self): |
4145 |
"""Build hooks nodes.
|
4146 |
|
4147 |
"""
|
4148 |
instance = self._migrater.instance
|
4149 |
nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes) |
4150 |
return (nl, nl + [instance.primary_node])
|
4151 |
|
4152 |
|
4153 |
class LUInstanceMigrate(LogicalUnit): |
4154 |
"""Migrate an instance.
|
4155 |
|
4156 |
This is migration without shutting down, compared to the failover,
|
4157 |
which is done with shutdown.
|
4158 |
|
4159 |
"""
|
4160 |
HPATH = "instance-migrate"
|
4161 |
HTYPE = constants.HTYPE_INSTANCE |
4162 |
REQ_BGL = False
|
4163 |
|
4164 |
def ExpandNames(self): |
4165 |
self._ExpandAndLockInstance()
|
4166 |
_ExpandNamesForMigration(self)
|
4167 |
|
4168 |
self._migrater = \
|
4169 |
TLMigrateInstance(self, self.op.instance_name, self.op.cleanup, |
4170 |
False, self.op.allow_failover, False, |
4171 |
self.op.allow_runtime_changes,
|
4172 |
constants.DEFAULT_SHUTDOWN_TIMEOUT, |
4173 |
self.op.ignore_ipolicy)
|
4174 |
|
4175 |
self.tasklets = [self._migrater] |
4176 |
|
4177 |
def DeclareLocks(self, level): |
4178 |
_DeclareLocksForMigration(self, level)
|
4179 |
|
4180 |
def BuildHooksEnv(self): |
4181 |
"""Build hooks env.
|
4182 |
|
4183 |
This runs on master, primary and secondary nodes of the instance.
|
4184 |
|
4185 |
"""
|
4186 |
instance = self._migrater.instance
|
4187 |
source_node = instance.primary_node |
4188 |
target_node = self.op.target_node
|
4189 |
env = _BuildInstanceHookEnvByObject(self, instance)
|
4190 |
env.update({ |
4191 |
"MIGRATE_LIVE": self._migrater.live, |
4192 |
"MIGRATE_CLEANUP": self.op.cleanup, |
4193 |
"OLD_PRIMARY": source_node,
|
4194 |
"NEW_PRIMARY": target_node,
|
4195 |
"ALLOW_RUNTIME_CHANGES": self.op.allow_runtime_changes, |
4196 |
}) |
4197 |
|
4198 |
if instance.disk_template in constants.DTS_INT_MIRROR: |
4199 |
env["OLD_SECONDARY"] = target_node
|
4200 |
env["NEW_SECONDARY"] = source_node
|
4201 |
else:
|
4202 |
env["OLD_SECONDARY"] = env["NEW_SECONDARY"] = None |
4203 |
|
4204 |
return env
|
4205 |
|
4206 |
def BuildHooksNodes(self): |
4207 |
"""Build hooks nodes.
|
4208 |
|
4209 |
"""
|
4210 |
instance = self._migrater.instance
|
4211 |
snodes = list(instance.secondary_nodes)
|
4212 |
nl = [self.cfg.GetMasterNode(), instance.primary_node] + snodes
|
4213 |
return (nl, nl)
|
4214 |
|
4215 |
|
4216 |
class LUInstanceMove(LogicalUnit): |
4217 |
"""Move an instance by data-copying.
|
4218 |
|
4219 |
"""
|
4220 |
HPATH = "instance-move"
|
4221 |
HTYPE = constants.HTYPE_INSTANCE |
4222 |
REQ_BGL = False
|
4223 |
|
4224 |
def ExpandNames(self): |
4225 |
self._ExpandAndLockInstance()
|
4226 |
target_node = _ExpandNodeName(self.cfg, self.op.target_node) |
4227 |
self.op.target_node = target_node
|
4228 |
self.needed_locks[locking.LEVEL_NODE] = [target_node]
|
4229 |
self.needed_locks[locking.LEVEL_NODE_RES] = []
|
4230 |
self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
|
4231 |
|
4232 |
def DeclareLocks(self, level): |
4233 |
if level == locking.LEVEL_NODE:
|
4234 |
self._LockInstancesNodes(primary_only=True) |
4235 |
elif level == locking.LEVEL_NODE_RES:
|
4236 |
# Copy node locks
|
4237 |
self.needed_locks[locking.LEVEL_NODE_RES] = \
|
4238 |
_CopyLockList(self.needed_locks[locking.LEVEL_NODE])
|
4239 |
|
4240 |
def BuildHooksEnv(self): |
4241 |
"""Build hooks env.
|
4242 |
|
4243 |
This runs on master, primary and secondary nodes of the instance.
|
4244 |
|
4245 |
"""
|
4246 |
env = { |
4247 |
"TARGET_NODE": self.op.target_node, |
4248 |
"SHUTDOWN_TIMEOUT": self.op.shutdown_timeout, |
4249 |
} |
4250 |
env.update(_BuildInstanceHookEnvByObject(self, self.instance)) |
4251 |
return env
|
4252 |
|
4253 |
def BuildHooksNodes(self): |
4254 |
"""Build hooks nodes.
|
4255 |
|
4256 |
"""
|
4257 |
nl = [ |
4258 |
self.cfg.GetMasterNode(),
|
4259 |
self.instance.primary_node,
|
4260 |
self.op.target_node,
|
4261 |
] |
4262 |
return (nl, nl)
|
4263 |
|
4264 |
def CheckPrereq(self): |
4265 |
"""Check prerequisites.
|
4266 |
|
4267 |
This checks that the instance is in the cluster.
|
4268 |
|
4269 |
"""
|
4270 |
self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name) |
4271 |
assert self.instance is not None, \ |
4272 |
"Cannot retrieve locked instance %s" % self.op.instance_name |
4273 |
|
4274 |
if instance.disk_template not in constants.DTS_COPYABLE: |
4275 |
raise errors.OpPrereqError("Disk template %s not suitable for copying" % |
4276 |
instance.disk_template, errors.ECODE_STATE) |
4277 |
|
4278 |
node = self.cfg.GetNodeInfo(self.op.target_node) |
4279 |
assert node is not None, \ |
4280 |
"Cannot retrieve locked node %s" % self.op.target_node |
4281 |
|
4282 |
self.target_node = target_node = node.name
|
4283 |
|
4284 |
if target_node == instance.primary_node:
|
4285 |
raise errors.OpPrereqError("Instance %s is already on the node %s" % |
4286 |
(instance.name, target_node), |
4287 |
errors.ECODE_STATE) |
4288 |
|
4289 |
bep = self.cfg.GetClusterInfo().FillBE(instance)
|
4290 |
|
4291 |
for idx, dsk in enumerate(instance.disks): |
4292 |
if dsk.dev_type not in (constants.LD_LV, constants.LD_FILE): |
4293 |
raise errors.OpPrereqError("Instance disk %d has a complex layout," |
4294 |
" cannot copy" % idx, errors.ECODE_STATE)
|
4295 |
|
4296 |
_CheckNodeOnline(self, target_node)
|
4297 |
_CheckNodeNotDrained(self, target_node)
|
4298 |
_CheckNodeVmCapable(self, target_node)
|
4299 |
cluster = self.cfg.GetClusterInfo()
|
4300 |
group_info = self.cfg.GetNodeGroup(node.group)
|
4301 |
ipolicy = ganeti.masterd.instance.CalculateGroupIPolicy(cluster, group_info) |
4302 |
_CheckTargetNodeIPolicy(self, ipolicy, instance, node, self.cfg, |
4303 |
ignore=self.op.ignore_ipolicy)
|
4304 |
|
4305 |
if instance.admin_state == constants.ADMINST_UP:
|
4306 |
# check memory requirements on the secondary node
|
4307 |
_CheckNodeFreeMemory(self, target_node, "failing over instance %s" % |
4308 |
instance.name, bep[constants.BE_MAXMEM], |
4309 |
instance.hypervisor) |
4310 |
else:
|
4311 |
self.LogInfo("Not checking memory on the secondary node as" |
4312 |
" instance will not be started")
|
4313 |
|
4314 |
# check bridge existance
|
4315 |
_CheckInstanceBridgesExist(self, instance, node=target_node)
|
4316 |
|
4317 |
def Exec(self, feedback_fn): |
4318 |
"""Move an instance.
|
4319 |
|
4320 |
The move is done by shutting it down on its present node, copying
|
4321 |
the data over (slow) and starting it on the new node.
|
4322 |
|
4323 |
"""
|
4324 |
instance = self.instance
|
4325 |
|
4326 |
source_node = instance.primary_node |
4327 |
target_node = self.target_node
|
4328 |
|
4329 |
self.LogInfo("Shutting down instance %s on source node %s", |
4330 |
instance.name, source_node) |
4331 |
|
4332 |
assert (self.owned_locks(locking.LEVEL_NODE) == |
4333 |
self.owned_locks(locking.LEVEL_NODE_RES))
|
4334 |
|
4335 |
result = self.rpc.call_instance_shutdown(source_node, instance,
|
4336 |
self.op.shutdown_timeout,
|
4337 |
self.op.reason)
|
4338 |
msg = result.fail_msg |
4339 |
if msg:
|
4340 |
if self.op.ignore_consistency: |
4341 |
self.LogWarning("Could not shutdown instance %s on node %s." |
4342 |
" Proceeding anyway. Please make sure node"
|
4343 |
" %s is down. Error details: %s",
|
4344 |
instance.name, source_node, source_node, msg) |
4345 |
else:
|
4346 |
raise errors.OpExecError("Could not shutdown instance %s on" |
4347 |
" node %s: %s" %
|
4348 |
(instance.name, source_node, msg)) |
4349 |
|
4350 |
# create the target disks
|
4351 |
try:
|
4352 |
_CreateDisks(self, instance, target_node=target_node)
|
4353 |
except errors.OpExecError:
|
4354 |
self.LogWarning("Device creation failed") |
4355 |
self.cfg.ReleaseDRBDMinors(instance.name)
|
4356 |
raise
|
4357 |
|
4358 |
cluster_name = self.cfg.GetClusterInfo().cluster_name
|
4359 |
|
4360 |
errs = [] |
4361 |
# activate, get path, copy the data over
|
4362 |
for idx, disk in enumerate(instance.disks): |
4363 |
self.LogInfo("Copying data for disk %d", idx) |
4364 |
result = self.rpc.call_blockdev_assemble(target_node, (disk, instance),
|
4365 |
instance.name, True, idx)
|
4366 |
if result.fail_msg:
|
4367 |
self.LogWarning("Can't assemble newly created disk %d: %s", |
4368 |
idx, result.fail_msg) |
4369 |
errs.append(result.fail_msg) |
4370 |
break
|
4371 |
dev_path = result.payload |
4372 |
result = self.rpc.call_blockdev_export(source_node, (disk, instance),
|
4373 |
target_node, dev_path, |
4374 |
cluster_name) |
4375 |
if result.fail_msg:
|
4376 |
self.LogWarning("Can't copy data over for disk %d: %s", |
4377 |
idx, result.fail_msg) |
4378 |
errs.append(result.fail_msg) |
4379 |
break
|
4380 |
|
4381 |
if errs:
|
4382 |
self.LogWarning("Some disks failed to copy, aborting") |
4383 |
try:
|
4384 |
_RemoveDisks(self, instance, target_node=target_node)
|
4385 |
finally:
|
4386 |
self.cfg.ReleaseDRBDMinors(instance.name)
|
4387 |
raise errors.OpExecError("Errors during disk copy: %s" % |
4388 |
(",".join(errs),))
|
4389 |
|
4390 |
instance.primary_node = target_node |
4391 |
self.cfg.Update(instance, feedback_fn)
|
4392 |
|
4393 |
self.LogInfo("Removing the disks on the original node") |
4394 |
_RemoveDisks(self, instance, target_node=source_node)
|
4395 |
|
4396 |
# Only start the instance if it's marked as up
|
4397 |
if instance.admin_state == constants.ADMINST_UP:
|
4398 |
self.LogInfo("Starting instance %s on node %s", |
4399 |
instance.name, target_node) |
4400 |
|
4401 |
disks_ok, _ = _AssembleInstanceDisks(self, instance,
|
4402 |
ignore_secondaries=True)
|
4403 |
if not disks_ok: |
4404 |
_ShutdownInstanceDisks(self, instance)
|
4405 |
raise errors.OpExecError("Can't activate the instance's disks") |
4406 |
|
4407 |
result = self.rpc.call_instance_start(target_node,
|
4408 |
(instance, None, None), False, |
4409 |
self.op.reason)
|
4410 |
msg = result.fail_msg |
4411 |
if msg:
|
4412 |
_ShutdownInstanceDisks(self, instance)
|
4413 |
raise errors.OpExecError("Could not start instance %s on node %s: %s" % |
4414 |
(instance.name, target_node, msg)) |
4415 |
|
4416 |
|
4417 |
class LUNodeMigrate(LogicalUnit): |
4418 |
"""Migrate all instances from a node.
|
4419 |
|
4420 |
"""
|
4421 |
HPATH = "node-migrate"
|
4422 |
HTYPE = constants.HTYPE_NODE |
4423 |
REQ_BGL = False
|
4424 |
|
4425 |
def CheckArguments(self): |
4426 |
pass
|
4427 |
|
4428 |
def ExpandNames(self): |
4429 |
self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name) |
4430 |
|
4431 |
self.share_locks = _ShareAll()
|
4432 |
self.needed_locks = {
|
4433 |
locking.LEVEL_NODE: [self.op.node_name],
|
4434 |
} |
4435 |
|
4436 |
def BuildHooksEnv(self): |
4437 |
"""Build hooks env.
|
4438 |
|
4439 |
This runs on the master, the primary and all the secondaries.
|
4440 |
|
4441 |
"""
|
4442 |
return {
|
4443 |
"NODE_NAME": self.op.node_name, |
4444 |
"ALLOW_RUNTIME_CHANGES": self.op.allow_runtime_changes, |
4445 |
} |
4446 |
|
4447 |
def BuildHooksNodes(self): |
4448 |
"""Build hooks nodes.
|
4449 |
|
4450 |
"""
|
4451 |
nl = [self.cfg.GetMasterNode()]
|
4452 |
return (nl, nl)
|
4453 |
|
4454 |
def CheckPrereq(self): |
4455 |
pass
|
4456 |
|
4457 |
def Exec(self, feedback_fn): |
4458 |
# Prepare jobs for migration instances
|
4459 |
allow_runtime_changes = self.op.allow_runtime_changes
|
4460 |
jobs = [ |
4461 |
[opcodes.OpInstanceMigrate(instance_name=inst.name, |
4462 |
mode=self.op.mode,
|
4463 |
live=self.op.live,
|
4464 |
iallocator=self.op.iallocator,
|
4465 |
target_node=self.op.target_node,
|
4466 |
allow_runtime_changes=allow_runtime_changes, |
4467 |
ignore_ipolicy=self.op.ignore_ipolicy)]
|
4468 |
for inst in _GetNodePrimaryInstances(self.cfg, self.op.node_name)] |
4469 |
|
4470 |
# TODO: Run iallocator in this opcode and pass correct placement options to
|
4471 |
# OpInstanceMigrate. Since other jobs can modify the cluster between
|
4472 |
# running the iallocator and the actual migration, a good consistency model
|
4473 |
# will have to be found.
|
4474 |
|
4475 |
assert (frozenset(self.owned_locks(locking.LEVEL_NODE)) == |
4476 |
frozenset([self.op.node_name])) |
4477 |
|
4478 |
return ResultWithJobs(jobs)
|
4479 |
|
4480 |
|
4481 |
class TLMigrateInstance(Tasklet): |
4482 |
"""Tasklet class for instance migration.
|
4483 |
|
4484 |
@type live: boolean
|
4485 |
@ivar live: whether the migration will be done live or non-live;
|
4486 |
this variable is initalized only after CheckPrereq has run
|
4487 |
@type cleanup: boolean
|
4488 |
@ivar cleanup: Wheater we cleanup from a failed migration
|
4489 |
@type iallocator: string
|
4490 |
@ivar iallocator: The iallocator used to determine target_node
|
4491 |
@type target_node: string
|
4492 |
@ivar target_node: If given, the target_node to reallocate the instance to
|
4493 |
@type failover: boolean
|
4494 |
@ivar failover: Whether operation results in failover or migration
|
4495 |
@type fallback: boolean
|
4496 |
@ivar fallback: Whether fallback to failover is allowed if migration not
|
4497 |
possible
|
4498 |
@type ignore_consistency: boolean
|
4499 |
@ivar ignore_consistency: Wheter we should ignore consistency between source
|
4500 |
and target node
|
4501 |
@type shutdown_timeout: int
|
4502 |
@ivar shutdown_timeout: In case of failover timeout of the shutdown
|
4503 |
@type ignore_ipolicy: bool
|
4504 |
@ivar ignore_ipolicy: If true, we can ignore instance policy when migrating
|
4505 |
|
4506 |
"""
|
4507 |
|
4508 |
# Constants
|
4509 |
_MIGRATION_POLL_INTERVAL = 1 # seconds |
4510 |
_MIGRATION_FEEDBACK_INTERVAL = 10 # seconds |
4511 |
|
4512 |
def __init__(self, lu, instance_name, cleanup, failover, fallback, |
4513 |
ignore_consistency, allow_runtime_changes, shutdown_timeout, |
4514 |
ignore_ipolicy): |
4515 |
"""Initializes this class.
|
4516 |
|
4517 |
"""
|
4518 |
Tasklet.__init__(self, lu)
|
4519 |
|
4520 |
# Parameters
|
4521 |
self.instance_name = instance_name
|
4522 |
self.cleanup = cleanup
|
4523 |
self.live = False # will be overridden later |
4524 |
self.failover = failover
|
4525 |
self.fallback = fallback
|
4526 |
self.ignore_consistency = ignore_consistency
|
4527 |
self.shutdown_timeout = shutdown_timeout
|
4528 |
self.ignore_ipolicy = ignore_ipolicy
|
4529 |
self.allow_runtime_changes = allow_runtime_changes
|
4530 |
|
4531 |
def CheckPrereq(self): |
4532 |
"""Check prerequisites.
|
4533 |
|
4534 |
This checks that the instance is in the cluster.
|
4535 |
|
4536 |
"""
|
4537 |
instance_name = _ExpandInstanceName(self.lu.cfg, self.instance_name) |
4538 |
instance = self.cfg.GetInstanceInfo(instance_name)
|
4539 |
assert instance is not None |
4540 |
self.instance = instance
|
4541 |
cluster = self.cfg.GetClusterInfo()
|
4542 |
|
4543 |
if (not self.cleanup and |
4544 |
not instance.admin_state == constants.ADMINST_UP and |
4545 |
not self.failover and self.fallback): |
4546 |
self.lu.LogInfo("Instance is marked down or offline, fallback allowed," |
4547 |
" switching to failover")
|
4548 |
self.failover = True |
4549 |
|
4550 |
if instance.disk_template not in constants.DTS_MIRRORED: |
4551 |
if self.failover: |
4552 |
text = "failovers"
|
4553 |
else:
|
4554 |
text = "migrations"
|
4555 |
raise errors.OpPrereqError("Instance's disk layout '%s' does not allow" |
4556 |
" %s" % (instance.disk_template, text),
|
4557 |
errors.ECODE_STATE) |
4558 |
|
4559 |
if instance.disk_template in constants.DTS_EXT_MIRROR: |
4560 |
_CheckIAllocatorOrNode(self.lu, "iallocator", "target_node") |
4561 |
|
4562 |
if self.lu.op.iallocator: |
4563 |
assert locking.NAL in self.lu.owned_locks(locking.LEVEL_NODE_ALLOC) |
4564 |
self._RunAllocator()
|
4565 |
else:
|
4566 |
# We set set self.target_node as it is required by
|
4567 |
# BuildHooksEnv
|
4568 |
self.target_node = self.lu.op.target_node |
4569 |
|
4570 |
# Check that the target node is correct in terms of instance policy
|
4571 |
nodeinfo = self.cfg.GetNodeInfo(self.target_node) |
4572 |
group_info = self.cfg.GetNodeGroup(nodeinfo.group)
|
4573 |
ipolicy = ganeti.masterd.instance.CalculateGroupIPolicy(cluster, |
4574 |
group_info) |
4575 |
_CheckTargetNodeIPolicy(self.lu, ipolicy, instance, nodeinfo, self.cfg, |
4576 |
ignore=self.ignore_ipolicy)
|
4577 |
|
4578 |
# self.target_node is already populated, either directly or by the
|
4579 |
# iallocator run
|
4580 |
target_node = self.target_node
|
4581 |
if self.target_node == instance.primary_node: |
4582 |
raise errors.OpPrereqError("Cannot migrate instance %s" |
4583 |
" to its primary (%s)" %
|
4584 |
(instance.name, instance.primary_node), |
4585 |
errors.ECODE_STATE) |
4586 |
|
4587 |
if len(self.lu.tasklets) == 1: |
4588 |
# It is safe to release locks only when we're the only tasklet
|
4589 |
# in the LU
|
4590 |
_ReleaseLocks(self.lu, locking.LEVEL_NODE,
|
4591 |
keep=[instance.primary_node, self.target_node])
|
4592 |
_ReleaseLocks(self.lu, locking.LEVEL_NODE_ALLOC)
|
4593 |
|
4594 |
else:
|
4595 |
assert not self.lu.glm.is_owned(locking.LEVEL_NODE_ALLOC) |
4596 |
|
4597 |
secondary_nodes = instance.secondary_nodes |
4598 |
if not secondary_nodes: |
4599 |
raise errors.ConfigurationError("No secondary node but using" |
4600 |
" %s disk template" %
|
4601 |
instance.disk_template) |
4602 |
target_node = secondary_nodes[0]
|
4603 |
if self.lu.op.iallocator or (self.lu.op.target_node and |
4604 |
self.lu.op.target_node != target_node):
|
4605 |
if self.failover: |
4606 |
text = "failed over"
|
4607 |
else:
|
4608 |
text = "migrated"
|
4609 |
raise errors.OpPrereqError("Instances with disk template %s cannot" |
4610 |
" be %s to arbitrary nodes"
|
4611 |
" (neither an iallocator nor a target"
|
4612 |
" node can be passed)" %
|
4613 |
(instance.disk_template, text), |
4614 |
errors.ECODE_INVAL) |
4615 |
nodeinfo = self.cfg.GetNodeInfo(target_node)
|
4616 |
group_info = self.cfg.GetNodeGroup(nodeinfo.group)
|
4617 |
ipolicy = ganeti.masterd.instance.CalculateGroupIPolicy(cluster, |
4618 |
group_info) |
4619 |
_CheckTargetNodeIPolicy(self.lu, ipolicy, instance, nodeinfo, self.cfg, |
4620 |
ignore=self.ignore_ipolicy)
|
4621 |
|
4622 |
i_be = cluster.FillBE(instance) |
4623 |
|
4624 |
# check memory requirements on the secondary node
|
4625 |
if (not self.cleanup and |
4626 |
(not self.failover or instance.admin_state == constants.ADMINST_UP)): |
4627 |
self.tgt_free_mem = _CheckNodeFreeMemory(self.lu, target_node, |
4628 |
"migrating instance %s" %
|
4629 |
instance.name, |
4630 |
i_be[constants.BE_MINMEM], |
4631 |
instance.hypervisor) |
4632 |
else:
|
4633 |
self.lu.LogInfo("Not checking memory on the secondary node as" |
4634 |
" instance will not be started")
|
4635 |
|
4636 |
# check if failover must be forced instead of migration
|
4637 |
if (not self.cleanup and not self.failover and |
4638 |
i_be[constants.BE_ALWAYS_FAILOVER]): |
4639 |
self.lu.LogInfo("Instance configured to always failover; fallback" |
4640 |
" to failover")
|
4641 |
self.failover = True |
4642 |
|
4643 |
# check bridge existance
|
4644 |
_CheckInstanceBridgesExist(self.lu, instance, node=target_node)
|
4645 |
|
4646 |
if not self.cleanup: |
4647 |
_CheckNodeNotDrained(self.lu, target_node)
|
4648 |
if not self.failover: |
4649 |
result = self.rpc.call_instance_migratable(instance.primary_node,
|
4650 |
instance) |
4651 |
if result.fail_msg and self.fallback: |
4652 |
self.lu.LogInfo("Can't migrate, instance offline, fallback to" |
4653 |
" failover")
|
4654 |
self.failover = True |
4655 |
else:
|
4656 |
result.Raise("Can't migrate, please use failover",
|
4657 |
prereq=True, ecode=errors.ECODE_STATE)
|
4658 |
|
4659 |
assert not (self.failover and self.cleanup) |
4660 |
|
4661 |
if not self.failover: |
4662 |
if self.lu.op.live is not None and self.lu.op.mode is not None: |
4663 |
raise errors.OpPrereqError("Only one of the 'live' and 'mode'" |
4664 |
" parameters are accepted",
|
4665 |
errors.ECODE_INVAL) |
4666 |
if self.lu.op.live is not None: |
4667 |
if self.lu.op.live: |
4668 |
self.lu.op.mode = constants.HT_MIGRATION_LIVE
|
4669 |
else:
|
4670 |
self.lu.op.mode = constants.HT_MIGRATION_NONLIVE
|
4671 |
# reset the 'live' parameter to None so that repeated
|
4672 |
# invocations of CheckPrereq do not raise an exception
|
4673 |
self.lu.op.live = None |
4674 |
elif self.lu.op.mode is None: |
4675 |
# read the default value from the hypervisor
|
4676 |
i_hv = cluster.FillHV(self.instance, skip_globals=False) |
4677 |
self.lu.op.mode = i_hv[constants.HV_MIGRATION_MODE]
|
4678 |
|
4679 |
self.live = self.lu.op.mode == constants.HT_MIGRATION_LIVE |
4680 |
else:
|
4681 |
# Failover is never live
|
4682 |
self.live = False |
4683 |
|
4684 |
if not (self.failover or self.cleanup): |
4685 |
remote_info = self.rpc.call_instance_info(instance.primary_node,
|
4686 |
instance.name, |
4687 |
instance.hypervisor) |
4688 |
remote_info.Raise("Error checking instance on node %s" %
|
4689 |
instance.primary_node) |
4690 |
instance_running = bool(remote_info.payload)
|
4691 |
if instance_running:
|
4692 |
self.current_mem = int(remote_info.payload["memory"]) |
4693 |
|
4694 |
def _RunAllocator(self): |
4695 |
"""Run the allocator based on input opcode.
|
4696 |
|
4697 |
"""
|
4698 |
assert locking.NAL in self.lu.owned_locks(locking.LEVEL_NODE_ALLOC) |
4699 |
|
4700 |
# FIXME: add a self.ignore_ipolicy option
|
4701 |
req = iallocator.IAReqRelocate(name=self.instance_name,
|
4702 |
relocate_from=[self.instance.primary_node])
|
4703 |
ial = iallocator.IAllocator(self.cfg, self.rpc, req) |
4704 |
|
4705 |
ial.Run(self.lu.op.iallocator)
|
4706 |
|
4707 |
if not ial.success: |
4708 |
raise errors.OpPrereqError("Can't compute nodes using" |
4709 |
" iallocator '%s': %s" %
|
4710 |
(self.lu.op.iallocator, ial.info),
|
4711 |
errors.ECODE_NORES) |
4712 |
self.target_node = ial.result[0] |
4713 |
self.lu.LogInfo("Selected nodes for instance %s via iallocator %s: %s", |
4714 |
self.instance_name, self.lu.op.iallocator, |
4715 |
utils.CommaJoin(ial.result)) |
4716 |
|
4717 |
def _WaitUntilSync(self): |
4718 |
"""Poll with custom rpc for disk sync.
|
4719 |
|
4720 |
This uses our own step-based rpc call.
|
4721 |
|
4722 |
"""
|
4723 |
self.feedback_fn("* wait until resync is done") |
4724 |
all_done = False
|
4725 |
while not all_done: |
4726 |
all_done = True
|
4727 |
result = self.rpc.call_drbd_wait_sync(self.all_nodes, |
4728 |
self.nodes_ip,
|
4729 |
(self.instance.disks,
|
4730 |
self.instance))
|
4731 |
min_percent = 100
|
4732 |
for node, nres in result.items(): |
4733 |
nres.Raise("Cannot resync disks on node %s" % node)
|
4734 |
node_done, node_percent = nres.payload |
4735 |
all_done = all_done and node_done
|
4736 |
if node_percent is not None: |
4737 |
min_percent = min(min_percent, node_percent)
|
4738 |
if not all_done: |
4739 |
if min_percent < 100: |
4740 |
self.feedback_fn(" - progress: %.1f%%" % min_percent) |
4741 |
time.sleep(2)
|
4742 |
|
4743 |
def _EnsureSecondary(self, node): |
4744 |
"""Demote a node to secondary.
|
4745 |
|
4746 |
"""
|
4747 |
self.feedback_fn("* switching node %s to secondary mode" % node) |
4748 |
|
4749 |
for dev in self.instance.disks: |
4750 |
self.cfg.SetDiskID(dev, node)
|
4751 |
|
4752 |
result = self.rpc.call_blockdev_close(node, self.instance.name, |
4753 |
self.instance.disks)
|
4754 |
result.Raise("Cannot change disk to secondary on node %s" % node)
|
4755 |
|
4756 |
def _GoStandalone(self): |
4757 |
"""Disconnect from the network.
|
4758 |
|
4759 |
"""
|
4760 |
self.feedback_fn("* changing into standalone mode") |
4761 |
result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip, |
4762 |
self.instance.disks)
|
4763 |
for node, nres in result.items(): |
4764 |
nres.Raise("Cannot disconnect disks node %s" % node)
|
4765 |
|
4766 |
def _GoReconnect(self, multimaster): |
4767 |
"""Reconnect to the network.
|
4768 |
|
4769 |
"""
|
4770 |
if multimaster:
|
4771 |
msg = "dual-master"
|
4772 |
else:
|
4773 |
msg = "single-master"
|
4774 |
self.feedback_fn("* changing disks into %s mode" % msg) |
4775 |
result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip, |
4776 |
(self.instance.disks, self.instance), |
4777 |
self.instance.name, multimaster)
|
4778 |
for node, nres in result.items(): |
4779 |
nres.Raise("Cannot change disks config on node %s" % node)
|
4780 |
|
4781 |
def _ExecCleanup(self): |
4782 |
"""Try to cleanup after a failed migration.
|
4783 |
|
4784 |
The cleanup is done by:
|
4785 |
- check that the instance is running only on one node
|
4786 |
(and update the config if needed)
|
4787 |
- change disks on its secondary node to secondary
|
4788 |
- wait until disks are fully synchronized
|
4789 |
- disconnect from the network
|
4790 |
- change disks into single-master mode
|
4791 |
- wait again until disks are fully synchronized
|
4792 |
|
4793 |
"""
|
4794 |
instance = self.instance
|
4795 |
target_node = self.target_node
|
4796 |
source_node = self.source_node
|
4797 |
|
4798 |
# check running on only one node
|
4799 |
self.feedback_fn("* checking where the instance actually runs" |
4800 |
" (if this hangs, the hypervisor might be in"
|
4801 |
" a bad state)")
|
4802 |
ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor]) |
4803 |
for node, result in ins_l.items(): |
4804 |
result.Raise("Can't contact node %s" % node)
|
4805 |
|
4806 |
runningon_source = instance.name in ins_l[source_node].payload
|
4807 |
runningon_target = instance.name in ins_l[target_node].payload
|
4808 |
|
4809 |
if runningon_source and runningon_target: |
4810 |
raise errors.OpExecError("Instance seems to be running on two nodes," |
4811 |
" or the hypervisor is confused; you will have"
|
4812 |
" to ensure manually that it runs only on one"
|
4813 |
" and restart this operation")
|
4814 |
|
4815 |
if not (runningon_source or runningon_target): |
4816 |
raise errors.OpExecError("Instance does not seem to be running at all;" |
4817 |
" in this case it's safer to repair by"
|
4818 |
" running 'gnt-instance stop' to ensure disk"
|
4819 |
" shutdown, and then restarting it")
|
4820 |
|
4821 |
if runningon_target:
|
4822 |
# the migration has actually succeeded, we need to update the config
|
4823 |
self.feedback_fn("* instance running on secondary node (%s)," |
4824 |
" updating config" % target_node)
|
4825 |
instance.primary_node = target_node |
4826 |
self.cfg.Update(instance, self.feedback_fn) |
4827 |
demoted_node = source_node |
4828 |
else:
|
4829 |
self.feedback_fn("* instance confirmed to be running on its" |
4830 |
" primary node (%s)" % source_node)
|
4831 |
demoted_node = target_node |
4832 |
|
4833 |
if instance.disk_template in constants.DTS_INT_MIRROR: |
4834 |
self._EnsureSecondary(demoted_node)
|
4835 |
try:
|
4836 |
self._WaitUntilSync()
|
4837 |
except errors.OpExecError:
|
4838 |
# we ignore here errors, since if the device is standalone, it
|
4839 |
# won't be able to sync
|
4840 |
pass
|
4841 |
self._GoStandalone()
|
4842 |
self._GoReconnect(False) |
4843 |
self._WaitUntilSync()
|
4844 |
|
4845 |
self.feedback_fn("* done") |
4846 |
|
4847 |
def _RevertDiskStatus(self): |
4848 |
"""Try to revert the disk status after a failed migration.
|
4849 |
|
4850 |
"""
|
4851 |
target_node = self.target_node
|
4852 |
if self.instance.disk_template in constants.DTS_EXT_MIRROR: |
4853 |
return
|
4854 |
|
4855 |
try:
|
4856 |
self._EnsureSecondary(target_node)
|
4857 |
self._GoStandalone()
|
4858 |
self._GoReconnect(False) |
4859 |
self._WaitUntilSync()
|
4860 |
except errors.OpExecError, err:
|
4861 |
self.lu.LogWarning("Migration failed and I can't reconnect the drives," |
4862 |
" please try to recover the instance manually;"
|
4863 |
" error '%s'" % str(err)) |
4864 |
|
4865 |
def _AbortMigration(self): |
4866 |
"""Call the hypervisor code to abort a started migration.
|
4867 |
|
4868 |
"""
|
4869 |
instance = self.instance
|
4870 |
target_node = self.target_node
|
4871 |
source_node = self.source_node
|
4872 |
migration_info = self.migration_info
|
4873 |
|
4874 |
abort_result = self.rpc.call_instance_finalize_migration_dst(target_node,
|
4875 |
instance, |
4876 |
migration_info, |
4877 |
False)
|
4878 |
abort_msg = abort_result.fail_msg |
4879 |
if abort_msg:
|
4880 |
logging.error("Aborting migration failed on target node %s: %s",
|
4881 |
target_node, abort_msg) |
4882 |
# Don't raise an exception here, as we stil have to try to revert the
|
4883 |
# disk status, even if this step failed.
|
4884 |
|
4885 |
abort_result = self.rpc.call_instance_finalize_migration_src(
|
4886 |
source_node, instance, False, self.live) |
4887 |
abort_msg = abort_result.fail_msg |
4888 |
if abort_msg:
|
4889 |
logging.error("Aborting migration failed on source node %s: %s",
|
4890 |
source_node, abort_msg) |
4891 |
|
4892 |
def _ExecMigration(self): |
4893 |
"""Migrate an instance.
|
4894 |
|
4895 |
The migrate is done by:
|
4896 |
- change the disks into dual-master mode
|
4897 |
- wait until disks are fully synchronized again
|
4898 |
- migrate the instance
|
4899 |
- change disks on the new secondary node (the old primary) to secondary
|
4900 |
- wait until disks are fully synchronized
|
4901 |
- change disks into single-master mode
|
4902 |
|
4903 |
"""
|
4904 |
instance = self.instance
|
4905 |
target_node = self.target_node
|
4906 |
source_node = self.source_node
|
4907 |
|
4908 |
# Check for hypervisor version mismatch and warn the user.
|
4909 |
nodeinfo = self.rpc.call_node_info([source_node, target_node],
|
4910 |
None, [self.instance.hypervisor], False) |
4911 |
for ninfo in nodeinfo.values(): |
4912 |
ninfo.Raise("Unable to retrieve node information from node '%s'" %
|
4913 |
ninfo.node) |
4914 |
(_, _, (src_info, )) = nodeinfo[source_node].payload |
4915 |
(_, _, (dst_info, )) = nodeinfo[target_node].payload |
4916 |
|
4917 |
if ((constants.HV_NODEINFO_KEY_VERSION in src_info) and |
4918 |
(constants.HV_NODEINFO_KEY_VERSION in dst_info)):
|
4919 |
src_version = src_info[constants.HV_NODEINFO_KEY_VERSION] |
4920 |
dst_version = dst_info[constants.HV_NODEINFO_KEY_VERSION] |
4921 |
if src_version != dst_version:
|
4922 |
self.feedback_fn("* warning: hypervisor version mismatch between" |
4923 |
" source (%s) and target (%s) node" %
|
4924 |
(src_version, dst_version)) |
4925 |
|
4926 |
self.feedback_fn("* checking disk consistency between source and target") |
4927 |
for (idx, dev) in enumerate(instance.disks): |
4928 |
if not _CheckDiskConsistency(self.lu, instance, dev, target_node, False): |
4929 |
raise errors.OpExecError("Disk %s is degraded or not fully" |
4930 |
" synchronized on target node,"
|
4931 |
" aborting migration" % idx)
|
4932 |
|
4933 |
if self.current_mem > self.tgt_free_mem: |
4934 |
if not self.allow_runtime_changes: |
4935 |
raise errors.OpExecError("Memory ballooning not allowed and not enough" |
4936 |
" free memory to fit instance %s on target"
|
4937 |
" node %s (have %dMB, need %dMB)" %
|
4938 |
(instance.name, target_node, |
4939 |
self.tgt_free_mem, self.current_mem)) |
4940 |
self.feedback_fn("* setting instance memory to %s" % self.tgt_free_mem) |
4941 |
rpcres = self.rpc.call_instance_balloon_memory(instance.primary_node,
|
4942 |
instance, |
4943 |
self.tgt_free_mem)
|
4944 |
rpcres.Raise("Cannot modify instance runtime memory")
|
4945 |
|
4946 |
# First get the migration information from the remote node
|
4947 |
result = self.rpc.call_migration_info(source_node, instance)
|
4948 |
msg = result.fail_msg |
4949 |
if msg:
|
4950 |
log_err = ("Failed fetching source migration information from %s: %s" %
|
4951 |
(source_node, msg)) |
4952 |
logging.error(log_err) |
4953 |
raise errors.OpExecError(log_err)
|
4954 |
|
4955 |
self.migration_info = migration_info = result.payload
|
4956 |
|
4957 |
if self.instance.disk_template not in constants.DTS_EXT_MIRROR: |
4958 |
# Then switch the disks to master/master mode
|
4959 |
self._EnsureSecondary(target_node)
|
4960 |
self._GoStandalone()
|
4961 |
self._GoReconnect(True) |
4962 |
self._WaitUntilSync()
|
4963 |
|
4964 |
self.feedback_fn("* preparing %s to accept the instance" % target_node) |
4965 |
result = self.rpc.call_accept_instance(target_node,
|
4966 |
instance, |
4967 |
migration_info, |
4968 |
self.nodes_ip[target_node])
|
4969 |
|
4970 |
msg = result.fail_msg |
4971 |
if msg:
|
4972 |
logging.error("Instance pre-migration failed, trying to revert"
|
4973 |
" disk status: %s", msg)
|
4974 |
self.feedback_fn("Pre-migration failed, aborting") |
4975 |
self._AbortMigration()
|
4976 |
self._RevertDiskStatus()
|
4977 |
raise errors.OpExecError("Could not pre-migrate instance %s: %s" % |
4978 |
(instance.name, msg)) |
4979 |
|
4980 |
self.feedback_fn("* migrating instance to %s" % target_node) |
4981 |
result = self.rpc.call_instance_migrate(source_node, instance,
|
4982 |
self.nodes_ip[target_node],
|
4983 |
self.live)
|
4984 |
msg = result.fail_msg |
4985 |
if msg:
|
4986 |
logging.error("Instance migration failed, trying to revert"
|
4987 |
" disk status: %s", msg)
|
4988 |
self.feedback_fn("Migration failed, aborting") |
4989 |
self._AbortMigration()
|
4990 |
self._RevertDiskStatus()
|
4991 |
raise errors.OpExecError("Could not migrate instance %s: %s" % |
4992 |
(instance.name, msg)) |
4993 |
|
4994 |
self.feedback_fn("* starting memory transfer") |
4995 |
last_feedback = time.time() |
4996 |
while True: |
4997 |
result = self.rpc.call_instance_get_migration_status(source_node,
|
4998 |
instance) |
4999 |
msg = result.fail_msg |
5000 |
ms = result.payload # MigrationStatus instance
|
5001 |
if msg or (ms.status in constants.HV_MIGRATION_FAILED_STATUSES): |
5002 |
logging.error("Instance migration failed, trying to revert"
|
5003 |
" disk status: %s", msg)
|
5004 |
self.feedback_fn("Migration failed, aborting") |
5005 |
self._AbortMigration()
|
5006 |
self._RevertDiskStatus()
|
5007 |
if not msg: |
5008 |
msg = "hypervisor returned failure"
|
5009 |
raise errors.OpExecError("Could not migrate instance %s: %s" % |
5010 |
(instance.name, msg)) |
5011 |
|
5012 |
if result.payload.status != constants.HV_MIGRATION_ACTIVE:
|
5013 |
self.feedback_fn("* memory transfer complete") |
5014 |
break
|
5015 |
|
5016 |
if (utils.TimeoutExpired(last_feedback,
|
5017 |
self._MIGRATION_FEEDBACK_INTERVAL) and |
5018 |
ms.transferred_ram is not None): |
5019 |
mem_progress = 100 * float(ms.transferred_ram) / float(ms.total_ram) |
5020 |
self.feedback_fn("* memory transfer progress: %.2f %%" % mem_progress) |
5021 |
last_feedback = time.time() |
5022 |
|
5023 |
time.sleep(self._MIGRATION_POLL_INTERVAL)
|
5024 |
|
5025 |
result = self.rpc.call_instance_finalize_migration_src(source_node,
|
5026 |
instance, |
5027 |
True,
|
5028 |
self.live)
|
5029 |
msg = result.fail_msg |
5030 |
if msg:
|
5031 |
logging.error("Instance migration succeeded, but finalization failed"
|
5032 |
" on the source node: %s", msg)
|
5033 |
raise errors.OpExecError("Could not finalize instance migration: %s" % |
5034 |
msg) |
5035 |
|
5036 |
instance.primary_node = target_node |
5037 |
|
5038 |
# distribute new instance config to the other nodes
|
5039 |
self.cfg.Update(instance, self.feedback_fn) |
5040 |
|
5041 |
result = self.rpc.call_instance_finalize_migration_dst(target_node,
|
5042 |
instance, |
5043 |
migration_info, |
5044 |
True)
|
5045 |
msg = result.fail_msg |
5046 |
if msg:
|
5047 |
logging.error("Instance migration succeeded, but finalization failed"
|
5048 |
" on the target node: %s", msg)
|
5049 |
raise errors.OpExecError("Could not finalize instance migration: %s" % |
5050 |
msg) |
5051 |
|
5052 |
if self.instance.disk_template not in constants.DTS_EXT_MIRROR: |
5053 |
self._EnsureSecondary(source_node)
|
5054 |
self._WaitUntilSync()
|
5055 |
self._GoStandalone()
|
5056 |
self._GoReconnect(False) |
5057 |
self._WaitUntilSync()
|
5058 |
|
5059 |
# If the instance's disk template is `rbd' or `ext' and there was a
|
5060 |
# successful migration, unmap the device from the source node.
|
5061 |
if self.instance.disk_template in (constants.DT_RBD, constants.DT_EXT): |
5062 |
disks = _ExpandCheckDisks(instance, instance.disks) |
5063 |
self.feedback_fn("* unmapping instance's disks from %s" % source_node) |
5064 |
for disk in disks: |
5065 |
result = self.rpc.call_blockdev_shutdown(source_node, (disk, instance))
|
5066 |
msg = result.fail_msg |
5067 |
if msg:
|
5068 |
logging.error("Migration was successful, but couldn't unmap the"
|
5069 |
" block device %s on source node %s: %s",
|
5070 |
disk.iv_name, source_node, msg) |
5071 |
logging.error("You need to unmap the device %s manually on %s",
|
5072 |
disk.iv_name, source_node) |
5073 |
|
5074 |
self.feedback_fn("* done") |
5075 |
|
5076 |
def _ExecFailover(self): |
5077 |
"""Failover an instance.
|
5078 |
|
5079 |
The failover is done by shutting it down on its present node and
|
5080 |
starting it on the secondary.
|
5081 |
|
5082 |
"""
|
5083 |
instance = self.instance
|
5084 |
primary_node = self.cfg.GetNodeInfo(instance.primary_node)
|
5085 |
|
5086 |
source_node = instance.primary_node |
5087 |
target_node = self.target_node
|
5088 |
|
5089 |
if instance.admin_state == constants.ADMINST_UP:
|
5090 |
self.feedback_fn("* checking disk consistency between source and target") |
5091 |
for (idx, dev) in enumerate(instance.disks): |
5092 |
# for drbd, these are drbd over lvm
|
5093 |
if not _CheckDiskConsistency(self.lu, instance, dev, target_node, |
5094 |
False):
|
5095 |
if primary_node.offline:
|
5096 |
self.feedback_fn("Node %s is offline, ignoring degraded disk %s on" |
5097 |
" target node %s" %
|
5098 |
(primary_node.name, idx, target_node)) |
5099 |
elif not self.ignore_consistency: |
5100 |
raise errors.OpExecError("Disk %s is degraded on target node," |
5101 |
" aborting failover" % idx)
|
5102 |
else:
|
5103 |
self.feedback_fn("* not checking disk consistency as instance is not" |
5104 |
" running")
|
5105 |
|
5106 |
self.feedback_fn("* shutting down instance on source node") |
5107 |
logging.info("Shutting down instance %s on node %s",
|
5108 |
instance.name, source_node) |
5109 |
|
5110 |
result = self.rpc.call_instance_shutdown(source_node, instance,
|
5111 |
self.shutdown_timeout,
|
5112 |
self.lu.op.reason)
|
5113 |
msg = result.fail_msg |
5114 |
if msg:
|
5115 |
if self.ignore_consistency or primary_node.offline: |
5116 |
self.lu.LogWarning("Could not shutdown instance %s on node %s," |
5117 |
" proceeding anyway; please make sure node"
|
5118 |
" %s is down; error details: %s",
|
5119 |
instance.name, source_node, source_node, msg) |
5120 |
else:
|
5121 |
raise errors.OpExecError("Could not shutdown instance %s on" |
5122 |
" node %s: %s" %
|
5123 |
(instance.name, source_node, msg)) |
5124 |
|
5125 |
self.feedback_fn("* deactivating the instance's disks on source node") |
5126 |
if not _ShutdownInstanceDisks(self.lu, instance, ignore_primary=True): |
5127 |
raise errors.OpExecError("Can't shut down the instance's disks") |
5128 |
|
5129 |
instance.primary_node = target_node |
5130 |
# distribute new instance config to the other nodes
|
5131 |
self.cfg.Update(instance, self.feedback_fn) |
5132 |
|
5133 |
# Only start the instance if it's marked as up
|
5134 |
if instance.admin_state == constants.ADMINST_UP:
|
5135 |
self.feedback_fn("* activating the instance's disks on target node %s" % |
5136 |
target_node) |
5137 |
logging.info("Starting instance %s on node %s",
|
5138 |
instance.name, target_node) |
5139 |
|
5140 |
disks_ok, _ = _AssembleInstanceDisks(self.lu, instance,
|
5141 |
ignore_secondaries=True)
|
5142 |
if not disks_ok: |
5143 |
_ShutdownInstanceDisks(self.lu, instance)
|
5144 |
raise errors.OpExecError("Can't activate the instance's disks") |
5145 |
|
5146 |
self.feedback_fn("* starting the instance on the target node %s" % |
5147 |
target_node) |
5148 |
result = self.rpc.call_instance_start(target_node, (instance, None, None), |
5149 |
False, self.lu.op.reason) |
5150 |
msg = result.fail_msg |
5151 |
if msg:
|
5152 |
_ShutdownInstanceDisks(self.lu, instance)
|
5153 |
raise errors.OpExecError("Could not start instance %s on node %s: %s" % |
5154 |
(instance.name, target_node, msg)) |
5155 |
|
5156 |
def Exec(self, feedback_fn): |
5157 |
"""Perform the migration.
|
5158 |
|
5159 |
"""
|
5160 |
self.feedback_fn = feedback_fn
|
5161 |
self.source_node = self.instance.primary_node |
5162 |
|
5163 |
# FIXME: if we implement migrate-to-any in DRBD, this needs fixing
|
5164 |
if self.instance.disk_template in constants.DTS_INT_MIRROR: |
5165 |
self.target_node = self.instance.secondary_nodes[0] |
5166 |
# Otherwise self.target_node has been populated either
|
5167 |
# directly, or through an iallocator.
|
5168 |
|
5169 |
self.all_nodes = [self.source_node, self.target_node] |
5170 |
self.nodes_ip = dict((name, node.secondary_ip) for (name, node) |
5171 |
in self.cfg.GetMultiNodeInfo(self.all_nodes)) |
5172 |
|
5173 |
if self.failover: |
5174 |
feedback_fn("Failover instance %s" % self.instance.name) |
5175 |
self._ExecFailover()
|
5176 |
else:
|
5177 |
feedback_fn("Migrating instance %s" % self.instance.name) |
5178 |
|
5179 |
if self.cleanup: |
5180 |
return self._ExecCleanup() |
5181 |
else:
|
5182 |
return self._ExecMigration() |
5183 |
|
5184 |
|
5185 |
def _CreateBlockDev(lu, node, instance, device, force_create, info, |
5186 |
force_open): |
5187 |
"""Wrapper around L{_CreateBlockDevInner}.
|
5188 |
|
5189 |
This method annotates the root device first.
|
5190 |
|
5191 |
"""
|
5192 |
(disk,) = _AnnotateDiskParams(instance, [device], lu.cfg) |
5193 |
excl_stor = _IsExclusiveStorageEnabledNodeName(lu.cfg, node) |
5194 |
return _CreateBlockDevInner(lu, node, instance, disk, force_create, info,
|
5195 |
force_open, excl_stor) |
5196 |
|
5197 |
|
5198 |
def _CreateBlockDevInner(lu, node, instance, device, force_create, |
5199 |
info, force_open, excl_stor): |
5200 |
"""Create a tree of block devices on a given node.
|
5201 |
|
5202 |
If this device type has to be created on secondaries, create it and
|
5203 |
all its children.
|
5204 |
|
5205 |
If not, just recurse to children keeping the same 'force' value.
|
5206 |
|
5207 |
@attention: The device has to be annotated already.
|
5208 |
|
5209 |
@param lu: the lu on whose behalf we execute
|
5210 |
@param node: the node on which to create the device
|
5211 |
@type instance: L{objects.Instance}
|
5212 |
@param instance: the instance which owns the device
|
5213 |
@type device: L{objects.Disk}
|
5214 |
@param device: the device to create
|
5215 |
@type force_create: boolean
|
5216 |
@param force_create: whether to force creation of this device; this
|
5217 |
will be change to True whenever we find a device which has
|
5218 |
CreateOnSecondary() attribute
|
5219 |
@param info: the extra 'metadata' we should attach to the device
|
5220 |
(this will be represented as a LVM tag)
|
5221 |
@type force_open: boolean
|
5222 |
@param force_open: this parameter will be passes to the
|
5223 |
L{backend.BlockdevCreate} function where it specifies
|
5224 |
whether we run on primary or not, and it affects both
|
5225 |
the child assembly and the device own Open() execution
|
5226 |
@type excl_stor: boolean
|
5227 |
@param excl_stor: Whether exclusive_storage is active for the node
|
5228 |
|
5229 |
@return: list of created devices
|
5230 |
"""
|
5231 |
created_devices = [] |
5232 |
try:
|
5233 |
if device.CreateOnSecondary():
|
5234 |
force_create = True
|
5235 |
|
5236 |
if device.children:
|
5237 |
for child in device.children: |
5238 |
devs = _CreateBlockDevInner(lu, node, instance, child, force_create, |
5239 |
info, force_open, excl_stor) |
5240 |
created_devices.extend(devs) |
5241 |
|
5242 |
if not force_create: |
5243 |
return created_devices
|
5244 |
|
5245 |
_CreateSingleBlockDev(lu, node, instance, device, info, force_open, |
5246 |
excl_stor) |
5247 |
# The device has been completely created, so there is no point in keeping
|
5248 |
# its subdevices in the list. We just add the device itself instead.
|
5249 |
created_devices = [(node, device)] |
5250 |
return created_devices
|
5251 |
|
5252 |
except errors.DeviceCreationError, e:
|
5253 |
e.created_devices.extend(created_devices) |
5254 |
raise e
|
5255 |
except errors.OpExecError, e:
|
5256 |
raise errors.DeviceCreationError(str(e), created_devices) |
5257 |
|
5258 |
|
5259 |
def _CreateSingleBlockDev(lu, node, instance, device, info, force_open, |
5260 |
excl_stor): |
5261 |
"""Create a single block device on a given node.
|
5262 |
|
5263 |
This will not recurse over children of the device, so they must be
|
5264 |
created in advance.
|
5265 |
|
5266 |
@param lu: the lu on whose behalf we execute
|
5267 |
@param node: the node on which to create the device
|
5268 |
@type instance: L{objects.Instance}
|
5269 |
@param instance: the instance which owns the device
|
5270 |
@type device: L{objects.Disk}
|
5271 |
@param device: the device to create
|
5272 |
@param info: the extra 'metadata' we should attach to the device
|
5273 |
(this will be represented as a LVM tag)
|
5274 |
@type force_open: boolean
|
5275 |
@param force_open: this parameter will be passes to the
|
5276 |
L{backend.BlockdevCreate} function where it specifies
|
5277 |
whether we run on primary or not, and it affects both
|
5278 |
the child assembly and the device own Open() execution
|
5279 |
@type excl_stor: boolean
|
5280 |
@param excl_stor: Whether exclusive_storage is active for the node
|
5281 |
|
5282 |
"""
|
5283 |
lu.cfg.SetDiskID(device, node) |
5284 |
result = lu.rpc.call_blockdev_create(node, device, device.size, |
5285 |
instance.name, force_open, info, |
5286 |
excl_stor) |
5287 |
result.Raise("Can't create block device %s on"
|
5288 |
" node %s for instance %s" % (device, node, instance.name))
|
5289 |
if device.physical_id is None: |
5290 |
device.physical_id = result.payload |
5291 |
|
5292 |
|
5293 |
def _GenerateUniqueNames(lu, exts): |
5294 |
"""Generate a suitable LV name.
|
5295 |
|
5296 |
This will generate a logical volume name for the given instance.
|
5297 |
|
5298 |
"""
|
5299 |
results = [] |
5300 |
for val in exts: |
5301 |
new_id = lu.cfg.GenerateUniqueID(lu.proc.GetECId()) |
5302 |
results.append("%s%s" % (new_id, val))
|
5303 |
return results
|
5304 |
|
5305 |
|
5306 |
def _GenerateDRBD8Branch(lu, primary, secondary, size, vgnames, names, |
5307 |
iv_name, p_minor, s_minor): |
5308 |
"""Generate a drbd8 device complete with its children.
|
5309 |
|
5310 |
"""
|
5311 |
assert len(vgnames) == len(names) == 2 |
5312 |
port = lu.cfg.AllocatePort() |
5313 |
shared_secret = lu.cfg.GenerateDRBDSecret(lu.proc.GetECId()) |
5314 |
|
5315 |
dev_data = objects.Disk(dev_type=constants.LD_LV, size=size, |
5316 |
logical_id=(vgnames[0], names[0]), |
5317 |
params={}) |
5318 |
dev_data.uuid = lu.cfg.GenerateUniqueID(lu.proc.GetECId()) |
5319 |
dev_meta = objects.Disk(dev_type=constants.LD_LV, |
5320 |
size=constants.DRBD_META_SIZE, |
5321 |
logical_id=(vgnames[1], names[1]), |
5322 |
params={}) |
5323 |
dev_meta.uuid = lu.cfg.GenerateUniqueID(lu.proc.GetECId()) |
5324 |
drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size, |
5325 |
logical_id=(primary, secondary, port, |
5326 |
p_minor, s_minor, |
5327 |
shared_secret), |
5328 |
children=[dev_data, dev_meta], |
5329 |
iv_name=iv_name, params={}) |
5330 |
drbd_dev.uuid = lu.cfg.GenerateUniqueID(lu.proc.GetECId()) |
5331 |
return drbd_dev
|
5332 |
|
5333 |
|
5334 |
_DISK_TEMPLATE_NAME_PREFIX = { |
5335 |
constants.DT_PLAIN: "",
|
5336 |
constants.DT_RBD: ".rbd",
|
5337 |
constants.DT_EXT: ".ext",
|
5338 |
} |
5339 |
|
5340 |
|
5341 |
_DISK_TEMPLATE_DEVICE_TYPE = { |
5342 |
constants.DT_PLAIN: constants.LD_LV, |
5343 |
constants.DT_FILE: constants.LD_FILE, |
5344 |
constants.DT_SHARED_FILE: constants.LD_FILE, |
5345 |
constants.DT_BLOCK: constants.LD_BLOCKDEV, |
5346 |
constants.DT_RBD: constants.LD_RBD, |
5347 |
constants.DT_EXT: constants.LD_EXT, |
5348 |
} |
5349 |
|
5350 |
|
5351 |
def _GenerateDiskTemplate( |
5352 |
lu, template_name, instance_name, primary_node, secondary_nodes, |
5353 |
disk_info, file_storage_dir, file_driver, base_index, |
5354 |
feedback_fn, full_disk_params, _req_file_storage=opcodes.RequireFileStorage, |
5355 |
_req_shr_file_storage=opcodes.RequireSharedFileStorage): |
5356 |
"""Generate the entire disk layout for a given template type.
|
5357 |
|
5358 |
"""
|
5359 |
vgname = lu.cfg.GetVGName() |
5360 |
disk_count = len(disk_info)
|
5361 |
disks = [] |
5362 |
|
5363 |
if template_name == constants.DT_DISKLESS:
|
5364 |
pass
|
5365 |
elif template_name == constants.DT_DRBD8:
|
5366 |
if len(secondary_nodes) != 1: |
5367 |
raise errors.ProgrammerError("Wrong template configuration") |
5368 |
remote_node = secondary_nodes[0]
|
5369 |
minors = lu.cfg.AllocateDRBDMinor( |
5370 |
[primary_node, remote_node] * len(disk_info), instance_name)
|
5371 |
|
5372 |
(drbd_params, _, _) = objects.Disk.ComputeLDParams(template_name, |
5373 |
full_disk_params) |
5374 |
drbd_default_metavg = drbd_params[constants.LDP_DEFAULT_METAVG] |
5375 |
|
5376 |
names = [] |
5377 |
for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % (base_index + i) |
5378 |
for i in range(disk_count)]): |
5379 |
names.append(lv_prefix + "_data")
|
5380 |
names.append(lv_prefix + "_meta")
|
5381 |
for idx, disk in enumerate(disk_info): |
5382 |
disk_index = idx + base_index |
5383 |
data_vg = disk.get(constants.IDISK_VG, vgname) |
5384 |
meta_vg = disk.get(constants.IDISK_METAVG, drbd_default_metavg) |
5385 |
disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node, |
5386 |
disk[constants.IDISK_SIZE], |
5387 |
[data_vg, meta_vg], |
5388 |
names[idx * 2:idx * 2 + 2], |
5389 |
"disk/%d" % disk_index,
|
5390 |
minors[idx * 2], minors[idx * 2 + 1]) |
5391 |
disk_dev.mode = disk[constants.IDISK_MODE] |
5392 |
disk_dev.name = disk.get(constants.IDISK_NAME, None)
|
5393 |
disks.append(disk_dev) |
5394 |
else:
|
5395 |
if secondary_nodes:
|
5396 |
raise errors.ProgrammerError("Wrong template configuration") |
5397 |
|
5398 |
if template_name == constants.DT_FILE:
|
5399 |
_req_file_storage() |
5400 |
elif template_name == constants.DT_SHARED_FILE:
|
5401 |
_req_shr_file_storage() |
5402 |
|
5403 |
name_prefix = _DISK_TEMPLATE_NAME_PREFIX.get(template_name, None)
|
5404 |
if name_prefix is None: |
5405 |
names = None
|
5406 |
else:
|
5407 |
names = _GenerateUniqueNames(lu, ["%s.disk%s" %
|
5408 |
(name_prefix, base_index + i) |
5409 |
for i in range(disk_count)]) |
5410 |
|
5411 |
if template_name == constants.DT_PLAIN:
|
5412 |
|
5413 |
def logical_id_fn(idx, _, disk): |
5414 |
vg = disk.get(constants.IDISK_VG, vgname) |
5415 |
return (vg, names[idx])
|
5416 |
|
5417 |
elif template_name in (constants.DT_FILE, constants.DT_SHARED_FILE): |
5418 |
logical_id_fn = \ |
5419 |
lambda _, disk_index, disk: (file_driver,
|
5420 |
"%s/disk%d" % (file_storage_dir,
|
5421 |
disk_index)) |
5422 |
elif template_name == constants.DT_BLOCK:
|
5423 |
logical_id_fn = \ |
5424 |
lambda idx, disk_index, disk: (constants.BLOCKDEV_DRIVER_MANUAL,
|
5425 |
disk[constants.IDISK_ADOPT]) |
5426 |
elif template_name == constants.DT_RBD:
|
5427 |
logical_id_fn = lambda idx, _, disk: ("rbd", names[idx]) |
5428 |
elif template_name == constants.DT_EXT:
|
5429 |
def logical_id_fn(idx, _, disk): |
5430 |
provider = disk.get(constants.IDISK_PROVIDER, None)
|
5431 |
if provider is None: |
5432 |
raise errors.ProgrammerError("Disk template is %s, but '%s' is" |
5433 |
" not found", constants.DT_EXT,
|
5434 |
constants.IDISK_PROVIDER) |
5435 |
return (provider, names[idx])
|
5436 |
else:
|
5437 |
raise errors.ProgrammerError("Unknown disk template '%s'" % template_name) |
5438 |
|
5439 |
dev_type = _DISK_TEMPLATE_DEVICE_TYPE[template_name] |
5440 |
|
5441 |
for idx, disk in enumerate(disk_info): |
5442 |
params = {} |
5443 |
# Only for the Ext template add disk_info to params
|
5444 |
if template_name == constants.DT_EXT:
|
5445 |
params[constants.IDISK_PROVIDER] = disk[constants.IDISK_PROVIDER] |
5446 |
for key in disk: |
5447 |
if key not in constants.IDISK_PARAMS: |
5448 |
params[key] = disk[key] |
5449 |
disk_index = idx + base_index |
5450 |
size = disk[constants.IDISK_SIZE] |
5451 |
feedback_fn("* disk %s, size %s" %
|
5452 |
(disk_index, utils.FormatUnit(size, "h")))
|
5453 |
disk_dev = objects.Disk(dev_type=dev_type, size=size, |
5454 |
logical_id=logical_id_fn(idx, disk_index, disk), |
5455 |
iv_name="disk/%d" % disk_index,
|
5456 |
mode=disk[constants.IDISK_MODE], |
5457 |
params=params) |
5458 |
disk_dev.name = disk.get(constants.IDISK_NAME, None)
|
5459 |
disk_dev.uuid = lu.cfg.GenerateUniqueID(lu.proc.GetECId()) |
5460 |
disks.append(disk_dev) |
5461 |
|
5462 |
return disks
|
5463 |
|
5464 |
|
5465 |
def _GetInstanceInfoText(instance): |
5466 |
"""Compute that text that should be added to the disk's metadata.
|
5467 |
|
5468 |
"""
|
5469 |
return "originstname+%s" % instance.name |
5470 |
|
5471 |
|
5472 |
def _CalcEta(time_taken, written, total_size): |
5473 |
"""Calculates the ETA based on size written and total size.
|
5474 |
|
5475 |
@param time_taken: The time taken so far
|
5476 |
@param written: amount written so far
|
5477 |
@param total_size: The total size of data to be written
|
5478 |
@return: The remaining time in seconds
|
5479 |
|
5480 |
"""
|
5481 |
avg_time = time_taken / float(written)
|
5482 |
return (total_size - written) * avg_time
|
5483 |
|
5484 |
|
5485 |
def _WipeDisks(lu, instance, disks=None): |
5486 |
"""Wipes instance disks.
|
5487 |
|
5488 |
@type lu: L{LogicalUnit}
|
5489 |
@param lu: the logical unit on whose behalf we execute
|
5490 |
@type instance: L{objects.Instance}
|
5491 |
@param instance: the instance whose disks we should create
|
5492 |
@type disks: None or list of tuple of (number, L{objects.Disk}, number)
|
5493 |
@param disks: Disk details; tuple contains disk index, disk object and the
|
5494 |
start offset
|
5495 |
|
5496 |
"""
|
5497 |
node = instance.primary_node |
5498 |
|
5499 |
if disks is None: |
5500 |
disks = [(idx, disk, 0)
|
5501 |
for (idx, disk) in enumerate(instance.disks)] |
5502 |
|
5503 |
for (_, device, _) in disks: |
5504 |
lu.cfg.SetDiskID(device, node) |
5505 |
|
5506 |
logging.info("Pausing synchronization of disks of instance '%s'",
|
5507 |
instance.name) |
5508 |
result = lu.rpc.call_blockdev_pause_resume_sync(node, |
5509 |
(map(compat.snd, disks),
|
5510 |
instance), |
5511 |
True)
|
5512 |
result.Raise("Failed to pause disk synchronization on node '%s'" % node)
|
5513 |
|
5514 |
for idx, success in enumerate(result.payload): |
5515 |
if not success: |
5516 |
logging.warn("Pausing synchronization of disk %s of instance '%s'"
|
5517 |
" failed", idx, instance.name)
|
5518 |
|
5519 |
try:
|
5520 |
for (idx, device, offset) in disks: |
5521 |
# The wipe size is MIN_WIPE_CHUNK_PERCENT % of the instance disk but
|
5522 |
# MAX_WIPE_CHUNK at max. Truncating to integer to avoid rounding errors.
|
5523 |
wipe_chunk_size = \ |
5524 |
int(min(constants.MAX_WIPE_CHUNK, |
5525 |
device.size / 100.0 * constants.MIN_WIPE_CHUNK_PERCENT))
|
5526 |
|
5527 |
size = device.size |
5528 |
last_output = 0
|
5529 |
start_time = time.time() |
5530 |
|
5531 |
if offset == 0: |
5532 |
info_text = ""
|
5533 |
else:
|
5534 |
info_text = (" (from %s to %s)" %
|
5535 |
(utils.FormatUnit(offset, "h"),
|
5536 |
utils.FormatUnit(size, "h")))
|
5537 |
|
5538 |
lu.LogInfo("* Wiping disk %s%s", idx, info_text)
|
5539 |
|
5540 |
logging.info("Wiping disk %d for instance %s on node %s using"
|
5541 |
" chunk size %s", idx, instance.name, node, wipe_chunk_size)
|
5542 |
|
5543 |
while offset < size:
|
5544 |
wipe_size = min(wipe_chunk_size, size - offset)
|
5545 |
|
5546 |
logging.debug("Wiping disk %d, offset %s, chunk %s",
|
5547 |
idx, offset, wipe_size) |
5548 |
|
5549 |
result = lu.rpc.call_blockdev_wipe(node, (device, instance), offset, |
5550 |
wipe_size) |
5551 |
result.Raise("Could not wipe disk %d at offset %d for size %d" %
|
5552 |
(idx, offset, wipe_size)) |
5553 |
|
5554 |
now = time.time() |
5555 |
offset += wipe_size |
5556 |
if now - last_output >= 60: |
5557 |
eta = _CalcEta(now - start_time, offset, size) |
5558 |
lu.LogInfo(" - done: %.1f%% ETA: %s",
|
5559 |
offset / float(size) * 100, utils.FormatSeconds(eta)) |
5560 |
last_output = now |
5561 |
finally:
|
5562 |
logging.info("Resuming synchronization of disks for instance '%s'",
|
5563 |
instance.name) |
5564 |
|
5565 |
result = lu.rpc.call_blockdev_pause_resume_sync(node, |
5566 |
(map(compat.snd, disks),
|
5567 |
instance), |
5568 |
False)
|
5569 |
|
5570 |
if result.fail_msg:
|
5571 |
lu.LogWarning("Failed to resume disk synchronization on node '%s': %s",
|
5572 |
node, result.fail_msg) |
5573 |
else:
|
5574 |
for idx, success in enumerate(result.payload): |
5575 |
if not success: |
5576 |
lu.LogWarning("Resuming synchronization of disk %s of instance '%s'"
|
5577 |
" failed", idx, instance.name)
|
5578 |
|
5579 |
|
5580 |
def _CreateDisks(lu, instance, to_skip=None, target_node=None): |
5581 |
"""Create all disks for an instance.
|
5582 |
|
5583 |
This abstracts away some work from AddInstance.
|
5584 |
|
5585 |
@type lu: L{LogicalUnit}
|
5586 |
@param lu: the logical unit on whose behalf we execute
|
5587 |
@type instance: L{objects.Instance}
|
5588 |
@param instance: the instance whose disks we should create
|
5589 |
@type to_skip: list
|
5590 |
@param to_skip: list of indices to skip
|
5591 |
@type target_node: string
|
5592 |
@param target_node: if passed, overrides the target node for creation
|
5593 |
@rtype: boolean
|
5594 |
@return: the success of the creation
|
5595 |
|
5596 |
"""
|
5597 |
info = _GetInstanceInfoText(instance) |
5598 |
if target_node is None: |
5599 |
pnode = instance.primary_node |
5600 |
all_nodes = instance.all_nodes |
5601 |
else:
|
5602 |
pnode = target_node |
5603 |
all_nodes = [pnode] |
5604 |
|
5605 |
if instance.disk_template in constants.DTS_FILEBASED: |
5606 |
file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1]) |
5607 |
result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir) |
5608 |
|
5609 |
result.Raise("Failed to create directory '%s' on"
|
5610 |
" node %s" % (file_storage_dir, pnode))
|
5611 |
|
5612 |
disks_created = [] |
5613 |
# Note: this needs to be kept in sync with adding of disks in
|
5614 |
# LUInstanceSetParams
|
5615 |
for idx, device in enumerate(instance.disks): |
5616 |
if to_skip and idx in to_skip: |
5617 |
continue
|
5618 |
logging.info("Creating disk %s for instance '%s'", idx, instance.name)
|
5619 |
#HARDCODE
|
5620 |
for node in all_nodes: |
5621 |
f_create = node == pnode |
5622 |
try:
|
5623 |
_CreateBlockDev(lu, node, instance, device, f_create, info, f_create) |
5624 |
disks_created.append((node, device)) |
5625 |
except errors.OpExecError:
|
5626 |
logging.warning("Creating disk %s for instance '%s' failed",
|
5627 |
idx, instance.name) |
5628 |
except errors.DeviceCreationError, e:
|
5629 |
logging.warning("Creating disk %s for instance '%s' failed",
|
5630 |
idx, instance.name) |
5631 |
disks_created.extend(e.created_devices) |
5632 |
for (node, disk) in disks_created: |
5633 |
lu.cfg.SetDiskID(disk, node) |
5634 |
result = lu.rpc.call_blockdev_remove(node, disk) |
5635 |
if result.fail_msg:
|
5636 |
logging.warning("Failed to remove newly-created disk %s on node %s:"
|
5637 |
" %s", device, node, result.fail_msg)
|
5638 |
raise errors.OpExecError(e.message)
|
5639 |
|
5640 |
|
5641 |
def _RemoveDisks(lu, instance, target_node=None, ignore_failures=False): |
5642 |
"""Remove all disks for an instance.
|
5643 |
|
5644 |
This abstracts away some work from `AddInstance()` and
|
5645 |
`RemoveInstance()`. Note that in case some of the devices couldn't
|
5646 |
be removed, the removal will continue with the other ones.
|
5647 |
|
5648 |
@type lu: L{LogicalUnit}
|
5649 |
@param lu: the logical unit on whose behalf we execute
|
5650 |
@type instance: L{objects.Instance}
|
5651 |
@param instance: the instance whose disks we should remove
|
5652 |
@type target_node: string
|
5653 |
@param target_node: used to override the node on which to remove the disks
|
5654 |
@rtype: boolean
|
5655 |
@return: the success of the removal
|
5656 |
|
5657 |
"""
|
5658 |
logging.info("Removing block devices for instance %s", instance.name)
|
5659 |
|
5660 |
all_result = True
|
5661 |
ports_to_release = set()
|
5662 |
anno_disks = _AnnotateDiskParams(instance, instance.disks, lu.cfg) |
5663 |
for (idx, device) in enumerate(anno_disks): |
5664 |
if target_node:
|
5665 |
edata = [(target_node, device)] |
5666 |
else:
|
5667 |
edata = device.ComputeNodeTree(instance.primary_node) |
5668 |
for node, disk in edata: |
5669 |
lu.cfg.SetDiskID(disk, node) |
5670 |
result = lu.rpc.call_blockdev_remove(node, disk) |
5671 |
if result.fail_msg:
|
5672 |
lu.LogWarning("Could not remove disk %s on node %s,"
|
5673 |
" continuing anyway: %s", idx, node, result.fail_msg)
|
5674 |
if not (result.offline and node != instance.primary_node): |
5675 |
all_result = False
|
5676 |
|
5677 |
# if this is a DRBD disk, return its port to the pool
|
5678 |
if device.dev_type in constants.LDS_DRBD: |
5679 |
ports_to_release.add(device.logical_id[2])
|
5680 |
|
5681 |
if all_result or ignore_failures: |
5682 |
for port in ports_to_release: |
5683 |
lu.cfg.AddTcpUdpPort(port) |
5684 |
|
5685 |
if instance.disk_template in constants.DTS_FILEBASED: |
5686 |
file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1]) |
5687 |
if target_node:
|
5688 |
tgt = target_node |
5689 |
else:
|
5690 |
tgt = instance.primary_node |
5691 |
result = lu.rpc.call_file_storage_dir_remove(tgt, file_storage_dir) |
5692 |
if result.fail_msg:
|
5693 |
lu.LogWarning("Could not remove directory '%s' on node %s: %s",
|
5694 |
file_storage_dir, instance.primary_node, result.fail_msg) |
5695 |
all_result = False
|
5696 |
|
5697 |
return all_result
|
5698 |
|
5699 |
|
5700 |
def _ComputeDiskSizePerVG(disk_template, disks): |
5701 |
"""Compute disk size requirements in the volume group
|
5702 |
|
5703 |
"""
|
5704 |
def _compute(disks, payload): |
5705 |
"""Universal algorithm.
|
5706 |
|
5707 |
"""
|
5708 |
vgs = {} |
5709 |
for disk in disks: |
5710 |
vgs[disk[constants.IDISK_VG]] = \ |
5711 |
vgs.get(constants.IDISK_VG, 0) + disk[constants.IDISK_SIZE] + payload
|
5712 |
|
5713 |
return vgs
|
5714 |
|
5715 |
# Required free disk space as a function of disk and swap space
|
5716 |
req_size_dict = { |
5717 |
constants.DT_DISKLESS: {}, |
5718 |
constants.DT_PLAIN: _compute(disks, 0),
|
5719 |
# 128 MB are added for drbd metadata for each disk
|
5720 |
constants.DT_DRBD8: _compute(disks, constants.DRBD_META_SIZE), |
5721 |
constants.DT_FILE: {}, |
5722 |
constants.DT_SHARED_FILE: {}, |
5723 |
} |
5724 |
|
5725 |
if disk_template not in req_size_dict: |
5726 |
raise errors.ProgrammerError("Disk template '%s' size requirement" |
5727 |
" is unknown" % disk_template)
|
5728 |
|
5729 |
return req_size_dict[disk_template]
|
5730 |
|
5731 |
|
5732 |
def _CreateInstanceAllocRequest(op, disks, nics, beparams, node_whitelist): |
5733 |
"""Wrapper around IAReqInstanceAlloc.
|
5734 |
|
5735 |
@param op: The instance opcode
|
5736 |
@param disks: The computed disks
|
5737 |
@param nics: The computed nics
|
5738 |
@param beparams: The full filled beparams
|
5739 |
@param node_whitelist: List of nodes which should appear as online to the
|
5740 |
allocator (unless the node is already marked offline)
|
5741 |
|
5742 |
@returns: A filled L{iallocator.IAReqInstanceAlloc}
|
5743 |
|
5744 |
"""
|
5745 |
spindle_use = beparams[constants.BE_SPINDLE_USE] |
5746 |
return iallocator.IAReqInstanceAlloc(name=op.instance_name,
|
5747 |
disk_template=op.disk_template, |
5748 |
tags=op.tags, |
5749 |
os=op.os_type, |
5750 |
vcpus=beparams[constants.BE_VCPUS], |
5751 |
memory=beparams[constants.BE_MAXMEM], |
5752 |
spindle_use=spindle_use, |
5753 |
disks=disks, |
5754 |
nics=[n.ToDict() for n in nics], |
5755 |
hypervisor=op.hypervisor, |
5756 |
node_whitelist=node_whitelist) |
5757 |
|
5758 |
|
5759 |
def _ComputeNics(op, cluster, default_ip, cfg, ec_id): |
5760 |
"""Computes the nics.
|
5761 |
|
5762 |
@param op: The instance opcode
|
5763 |
@param cluster: Cluster configuration object
|
5764 |
@param default_ip: The default ip to assign
|
5765 |
@param cfg: An instance of the configuration object
|
5766 |
@param ec_id: Execution context ID
|
5767 |
|
5768 |
@returns: The build up nics
|
5769 |
|
5770 |
"""
|
5771 |
nics = [] |
5772 |
for nic in op.nics: |
5773 |
nic_mode_req = nic.get(constants.INIC_MODE, None)
|
5774 |
nic_mode = nic_mode_req |
5775 |
if nic_mode is None or nic_mode == constants.VALUE_AUTO: |
5776 |
nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE] |
5777 |
|
5778 |
net = nic.get(constants.INIC_NETWORK, None)
|
5779 |
link = nic.get(constants.NIC_LINK, None)
|
5780 |
ip = nic.get(constants.INIC_IP, None)
|
5781 |
|
5782 |
if net is None or net.lower() == constants.VALUE_NONE: |
5783 |
net = None
|
5784 |
else:
|
5785 |
if nic_mode_req is not None or link is not None: |
5786 |
raise errors.OpPrereqError("If network is given, no mode or link" |
5787 |
" is allowed to be passed",
|
5788 |
errors.ECODE_INVAL) |
5789 |
|
5790 |
# ip validity checks
|
5791 |
if ip is None or ip.lower() == constants.VALUE_NONE: |
5792 |
nic_ip = None
|
5793 |
elif ip.lower() == constants.VALUE_AUTO:
|
5794 |
if not op.name_check: |
5795 |
raise errors.OpPrereqError("IP address set to auto but name checks" |
5796 |
" have been skipped",
|
5797 |
errors.ECODE_INVAL) |
5798 |
nic_ip = default_ip |
5799 |
else:
|
5800 |
# We defer pool operations until later, so that the iallocator has
|
5801 |
# filled in the instance's node(s) dimara
|
5802 |
if ip.lower() == constants.NIC_IP_POOL:
|
5803 |
if net is None: |
5804 |
raise errors.OpPrereqError("if ip=pool, parameter network" |
5805 |
" must be passed too",
|
5806 |
errors.ECODE_INVAL) |
5807 |
|
5808 |
elif not netutils.IPAddress.IsValid(ip): |
5809 |
raise errors.OpPrereqError("Invalid IP address '%s'" % ip, |
5810 |
errors.ECODE_INVAL) |
5811 |
|
5812 |
nic_ip = ip |
5813 |
|
5814 |
# TODO: check the ip address for uniqueness
|
5815 |
if nic_mode == constants.NIC_MODE_ROUTED and not nic_ip: |
5816 |
raise errors.OpPrereqError("Routed nic mode requires an ip address", |
5817 |
errors.ECODE_INVAL) |
5818 |
|
5819 |
# MAC address verification
|
5820 |
mac = nic.get(constants.INIC_MAC, constants.VALUE_AUTO) |
5821 |
if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE): |
5822 |
mac = utils.NormalizeAndValidateMac(mac) |
5823 |
|
5824 |
try:
|
5825 |
# TODO: We need to factor this out
|
5826 |
cfg.ReserveMAC(mac, ec_id) |
5827 |
except errors.ReservationError:
|
5828 |
raise errors.OpPrereqError("MAC address %s already in use" |
5829 |
" in cluster" % mac,
|
5830 |
errors.ECODE_NOTUNIQUE) |
5831 |
|
5832 |
# Build nic parameters
|
5833 |
nicparams = {} |
5834 |
if nic_mode_req:
|
5835 |
nicparams[constants.NIC_MODE] = nic_mode |
5836 |
if link:
|
5837 |
nicparams[constants.NIC_LINK] = link |
5838 |
|
5839 |
check_params = cluster.SimpleFillNIC(nicparams) |
5840 |
objects.NIC.CheckParameterSyntax(check_params) |
5841 |
net_uuid = cfg.LookupNetwork(net) |
5842 |
name = nic.get(constants.INIC_NAME, None)
|
5843 |
if name is not None and name.lower() == constants.VALUE_NONE: |
5844 |
name = None
|
5845 |
nic_obj = objects.NIC(mac=mac, ip=nic_ip, name=name, |
5846 |
network=net_uuid, nicparams=nicparams) |
5847 |
nic_obj.uuid = cfg.GenerateUniqueID(ec_id) |
5848 |
nics.append(nic_obj) |
5849 |
|
5850 |
return nics
|
5851 |
|
5852 |
|
5853 |
def _ComputeDisks(op, default_vg): |
5854 |
"""Computes the instance disks.
|
5855 |
|
5856 |
@param op: The instance opcode
|
5857 |
@param default_vg: The default_vg to assume
|
5858 |
|
5859 |
@return: The computed disks
|
5860 |
|
5861 |
"""
|
5862 |
disks = [] |
5863 |
for disk in op.disks: |
5864 |
mode = disk.get(constants.IDISK_MODE, constants.DISK_RDWR) |
5865 |
if mode not in constants.DISK_ACCESS_SET: |
5866 |
raise errors.OpPrereqError("Invalid disk access mode '%s'" % |
5867 |
mode, errors.ECODE_INVAL) |
5868 |
size = disk.get(constants.IDISK_SIZE, None)
|
5869 |
if size is None: |
5870 |
raise errors.OpPrereqError("Missing disk size", errors.ECODE_INVAL) |
5871 |
try:
|
5872 |
size = int(size)
|
5873 |
except (TypeError, ValueError): |
5874 |
raise errors.OpPrereqError("Invalid disk size '%s'" % size, |
5875 |
errors.ECODE_INVAL) |
5876 |
|
5877 |
ext_provider = disk.get(constants.IDISK_PROVIDER, None)
|
5878 |
if ext_provider and op.disk_template != constants.DT_EXT: |
5879 |
raise errors.OpPrereqError("The '%s' option is only valid for the %s" |
5880 |
" disk template, not %s" %
|
5881 |
(constants.IDISK_PROVIDER, constants.DT_EXT, |
5882 |
op.disk_template), errors.ECODE_INVAL) |
5883 |
|
5884 |
data_vg = disk.get(constants.IDISK_VG, default_vg) |
5885 |
name = disk.get(constants.IDISK_NAME, None)
|
5886 |
if name is not None and name.lower() == constants.VALUE_NONE: |
5887 |
name = None
|
5888 |
new_disk = { |
5889 |
constants.IDISK_SIZE: size, |
5890 |
constants.IDISK_MODE: mode, |
5891 |
constants.IDISK_VG: data_vg, |
5892 |
constants.IDISK_NAME: name, |
5893 |
} |
5894 |
|
5895 |
if constants.IDISK_METAVG in disk: |
5896 |
new_disk[constants.IDISK_METAVG] = disk[constants.IDISK_METAVG] |
5897 |
if constants.IDISK_ADOPT in disk: |
5898 |
new_disk[constants.IDISK_ADOPT] = disk[constants.IDISK_ADOPT] |
5899 |
|
5900 |
# For extstorage, demand the `provider' option and add any
|
5901 |
# additional parameters (ext-params) to the dict
|
5902 |
if op.disk_template == constants.DT_EXT:
|
5903 |
if ext_provider:
|
5904 |
new_disk[constants.IDISK_PROVIDER] = ext_provider |
5905 |
for key in disk: |
5906 |
if key not in constants.IDISK_PARAMS: |
5907 |
new_disk[key] = disk[key] |
5908 |
else:
|
5909 |
raise errors.OpPrereqError("Missing provider for template '%s'" % |
5910 |
constants.DT_EXT, errors.ECODE_INVAL) |
5911 |
|
5912 |
disks.append(new_disk) |
5913 |
|
5914 |
return disks
|
5915 |
|
5916 |
|
5917 |
def _ComputeFullBeParams(op, cluster): |
5918 |
"""Computes the full beparams.
|
5919 |
|
5920 |
@param op: The instance opcode
|
5921 |
@param cluster: The cluster config object
|
5922 |
|
5923 |
@return: The fully filled beparams
|
5924 |
|
5925 |
"""
|
5926 |
default_beparams = cluster.beparams[constants.PP_DEFAULT] |
5927 |
for param, value in op.beparams.iteritems(): |
5928 |
if value == constants.VALUE_AUTO:
|
5929 |
op.beparams[param] = default_beparams[param] |
5930 |
objects.UpgradeBeParams(op.beparams) |
5931 |
utils.ForceDictType(op.beparams, constants.BES_PARAMETER_TYPES) |
5932 |
return cluster.SimpleFillBE(op.beparams)
|
5933 |
|
5934 |
|
5935 |
def _CheckOpportunisticLocking(op): |
5936 |
"""Generate error if opportunistic locking is not possible.
|
5937 |
|
5938 |
"""
|
5939 |
if op.opportunistic_locking and not op.iallocator: |
5940 |
raise errors.OpPrereqError("Opportunistic locking is only available in" |
5941 |
" combination with an instance allocator",
|
5942 |
errors.ECODE_INVAL) |
5943 |
|
5944 |
|
5945 |
class LUInstanceCreate(LogicalUnit): |
5946 |
"""Create an instance.
|
5947 |
|
5948 |
"""
|
5949 |
HPATH = "instance-add"
|
5950 |
HTYPE = constants.HTYPE_INSTANCE |
5951 |
REQ_BGL = False
|
5952 |
|
5953 |
def CheckArguments(self): |
5954 |
"""Check arguments.
|
5955 |
|
5956 |
"""
|
5957 |
# do not require name_check to ease forward/backward compatibility
|
5958 |
# for tools
|
5959 |
if self.op.no_install and self.op.start: |
5960 |
self.LogInfo("No-installation mode selected, disabling startup") |
5961 |
self.op.start = False |
5962 |
# validate/normalize the instance name
|
5963 |
self.op.instance_name = \
|
5964 |
netutils.Hostname.GetNormalizedName(self.op.instance_name)
|
5965 |
|
5966 |
if self.op.ip_check and not self.op.name_check: |
5967 |
# TODO: make the ip check more flexible and not depend on the name check
|
5968 |
raise errors.OpPrereqError("Cannot do IP address check without a name" |
5969 |
" check", errors.ECODE_INVAL)
|
5970 |
|
5971 |
# check nics' parameter names
|
5972 |
for nic in self.op.nics: |
5973 |
utils.ForceDictType(nic, constants.INIC_PARAMS_TYPES) |
5974 |
# check that NIC's parameters names are unique and valid
|
5975 |
utils.ValidateDeviceNames("NIC", self.op.nics) |
5976 |
|
5977 |
# check that disk's names are unique and valid
|
5978 |
utils.ValidateDeviceNames("disk", self.op.disks) |
5979 |
|
5980 |
cluster = self.cfg.GetClusterInfo()
|
5981 |
if not self.op.disk_template in cluster.enabled_disk_templates: |
5982 |
raise errors.OpPrereqError("Cannot create an instance with disk template" |
5983 |
" '%s', because it is not enabled in the"
|
5984 |
" cluster. Enabled disk templates are: %s." %
|
5985 |
(self.op.disk_template,
|
5986 |
",".join(cluster.enabled_disk_templates)))
|
5987 |
|
5988 |
# check disks. parameter names and consistent adopt/no-adopt strategy
|
5989 |
has_adopt = has_no_adopt = False
|
5990 |
for disk in self.op.disks: |
5991 |
if self.op.disk_template != constants.DT_EXT: |
5992 |
utils.ForceDictType(disk, constants.IDISK_PARAMS_TYPES) |
5993 |
if constants.IDISK_ADOPT in disk: |
5994 |
has_adopt = True
|
5995 |
else:
|
5996 |
has_no_adopt = True
|
5997 |
if has_adopt and has_no_adopt: |
5998 |
raise errors.OpPrereqError("Either all disks are adopted or none is", |
5999 |
errors.ECODE_INVAL) |
6000 |
if has_adopt:
|
6001 |
if self.op.disk_template not in constants.DTS_MAY_ADOPT: |
6002 |
raise errors.OpPrereqError("Disk adoption is not supported for the" |
6003 |
" '%s' disk template" %
|
6004 |
self.op.disk_template,
|
6005 |
errors.ECODE_INVAL) |
6006 |
if self.op.iallocator is not None: |
6007 |
raise errors.OpPrereqError("Disk adoption not allowed with an" |
6008 |
" iallocator script", errors.ECODE_INVAL)
|
6009 |
if self.op.mode == constants.INSTANCE_IMPORT: |
6010 |
raise errors.OpPrereqError("Disk adoption not allowed for" |
6011 |
" instance import", errors.ECODE_INVAL)
|
6012 |
else:
|
6013 |
if self.op.disk_template in constants.DTS_MUST_ADOPT: |
6014 |
raise errors.OpPrereqError("Disk template %s requires disk adoption," |
6015 |
" but no 'adopt' parameter given" %
|
6016 |
self.op.disk_template,
|
6017 |
errors.ECODE_INVAL) |
6018 |
|
6019 |
self.adopt_disks = has_adopt
|
6020 |
|
6021 |
# instance name verification
|
6022 |
if self.op.name_check: |
6023 |
self.hostname1 = _CheckHostnameSane(self, self.op.instance_name) |
6024 |
self.op.instance_name = self.hostname1.name |
6025 |
# used in CheckPrereq for ip ping check
|
6026 |
self.check_ip = self.hostname1.ip |
6027 |
else:
|
6028 |
self.check_ip = None |
6029 |
|
6030 |
# file storage checks
|
6031 |
if (self.op.file_driver and |
6032 |
not self.op.file_driver in constants.FILE_DRIVER): |
6033 |
raise errors.OpPrereqError("Invalid file driver name '%s'" % |
6034 |
self.op.file_driver, errors.ECODE_INVAL)
|
6035 |
|
6036 |
if self.op.disk_template == constants.DT_FILE: |
6037 |
opcodes.RequireFileStorage() |
6038 |
elif self.op.disk_template == constants.DT_SHARED_FILE: |
6039 |
opcodes.RequireSharedFileStorage() |
6040 |
|
6041 |
### Node/iallocator related checks
|
6042 |
_CheckIAllocatorOrNode(self, "iallocator", "pnode") |
6043 |
|
6044 |
if self.op.pnode is not None: |
6045 |
if self.op.disk_template in constants.DTS_INT_MIRROR: |
6046 |
if self.op.snode is None: |
6047 |
raise errors.OpPrereqError("The networked disk templates need" |
6048 |
" a mirror node", errors.ECODE_INVAL)
|
6049 |
elif self.op.snode: |
6050 |
self.LogWarning("Secondary node will be ignored on non-mirrored disk" |
6051 |
" template")
|
6052 |
self.op.snode = None |
6053 |
|
6054 |
_CheckOpportunisticLocking(self.op)
|
6055 |
|
6056 |
self._cds = _GetClusterDomainSecret()
|
6057 |
|
6058 |
if self.op.mode == constants.INSTANCE_IMPORT: |
6059 |
# On import force_variant must be True, because if we forced it at
|
6060 |
# initial install, our only chance when importing it back is that it
|
6061 |
# works again!
|
6062 |
self.op.force_variant = True |
6063 |
|
6064 |
if self.op.no_install: |
6065 |
self.LogInfo("No-installation mode has no effect during import") |
6066 |
|
6067 |
elif self.op.mode == constants.INSTANCE_CREATE: |
6068 |
if self.op.os_type is None: |
6069 |
raise errors.OpPrereqError("No guest OS specified", |
6070 |
errors.ECODE_INVAL) |
6071 |
if self.op.os_type in self.cfg.GetClusterInfo().blacklisted_os: |
6072 |
raise errors.OpPrereqError("Guest OS '%s' is not allowed for" |
6073 |
" installation" % self.op.os_type, |
6074 |
errors.ECODE_STATE) |
6075 |
if self.op.disk_template is None: |
6076 |
raise errors.OpPrereqError("No disk template specified", |
6077 |
errors.ECODE_INVAL) |
6078 |
|
6079 |
elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT: |
6080 |
# Check handshake to ensure both clusters have the same domain secret
|
6081 |
src_handshake = self.op.source_handshake
|
6082 |
if not src_handshake: |
6083 |
raise errors.OpPrereqError("Missing source handshake", |
6084 |
errors.ECODE_INVAL) |
6085 |
|
6086 |
errmsg = masterd.instance.CheckRemoteExportHandshake(self._cds,
|
6087 |
src_handshake) |
6088 |
if errmsg:
|
6089 |
raise errors.OpPrereqError("Invalid handshake: %s" % errmsg, |
6090 |
errors.ECODE_INVAL) |
6091 |
|
6092 |
# Load and check source CA
|
6093 |
self.source_x509_ca_pem = self.op.source_x509_ca |
6094 |
if not self.source_x509_ca_pem: |
6095 |
raise errors.OpPrereqError("Missing source X509 CA", |
6096 |
errors.ECODE_INVAL) |
6097 |
|
6098 |
try:
|
6099 |
(cert, _) = utils.LoadSignedX509Certificate(self.source_x509_ca_pem,
|
6100 |
self._cds)
|
6101 |
except OpenSSL.crypto.Error, err:
|
6102 |
raise errors.OpPrereqError("Unable to load source X509 CA (%s)" % |
6103 |
(err, ), errors.ECODE_INVAL) |
6104 |
|
6105 |
(errcode, msg) = utils.VerifyX509Certificate(cert, None, None) |
6106 |
if errcode is not None: |
6107 |
raise errors.OpPrereqError("Invalid source X509 CA (%s)" % (msg, ), |
6108 |
errors.ECODE_INVAL) |
6109 |
|
6110 |
self.source_x509_ca = cert
|
6111 |
|
6112 |
src_instance_name = self.op.source_instance_name
|
6113 |
if not src_instance_name: |
6114 |
raise errors.OpPrereqError("Missing source instance name", |
6115 |
errors.ECODE_INVAL) |
6116 |
|
6117 |
self.source_instance_name = \
|
6118 |
netutils.GetHostname(name=src_instance_name).name |
6119 |
|
6120 |
else:
|
6121 |
raise errors.OpPrereqError("Invalid instance creation mode %r" % |
6122 |
self.op.mode, errors.ECODE_INVAL)
|
6123 |
|
6124 |
def ExpandNames(self): |
6125 |
"""ExpandNames for CreateInstance.
|
6126 |
|
6127 |
Figure out the right locks for instance creation.
|
6128 |
|
6129 |
"""
|
6130 |
self.needed_locks = {}
|
6131 |
|
6132 |
instance_name = self.op.instance_name
|
6133 |
# this is just a preventive check, but someone might still add this
|
6134 |
# instance in the meantime, and creation will fail at lock-add time
|
6135 |
if instance_name in self.cfg.GetInstanceList(): |
6136 |
raise errors.OpPrereqError("Instance '%s' is already in the cluster" % |
6137 |
instance_name, errors.ECODE_EXISTS) |
6138 |
|
6139 |
self.add_locks[locking.LEVEL_INSTANCE] = instance_name
|
6140 |
|
6141 |
if self.op.iallocator: |
6142 |
# TODO: Find a solution to not lock all nodes in the cluster, e.g. by
|
6143 |
# specifying a group on instance creation and then selecting nodes from
|
6144 |
# that group
|
6145 |
self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
|
6146 |
self.needed_locks[locking.LEVEL_NODE_ALLOC] = locking.ALL_SET
|
6147 |
|
6148 |
if self.op.opportunistic_locking: |
6149 |
self.opportunistic_locks[locking.LEVEL_NODE] = True |
6150 |
self.opportunistic_locks[locking.LEVEL_NODE_RES] = True |
6151 |
else:
|
6152 |
self.op.pnode = _ExpandNodeName(self.cfg, self.op.pnode) |
6153 |
nodelist = [self.op.pnode]
|
6154 |
if self.op.snode is not None: |
6155 |
self.op.snode = _ExpandNodeName(self.cfg, self.op.snode) |
6156 |
nodelist.append(self.op.snode)
|
6157 |
self.needed_locks[locking.LEVEL_NODE] = nodelist
|
6158 |
|
6159 |
# in case of import lock the source node too
|
6160 |
if self.op.mode == constants.INSTANCE_IMPORT: |
6161 |
src_node = self.op.src_node
|
6162 |
src_path = self.op.src_path
|
6163 |
|
6164 |
if src_path is None: |
6165 |
self.op.src_path = src_path = self.op.instance_name |
6166 |
|
6167 |
if src_node is None: |
6168 |
self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
|
6169 |
self.needed_locks[locking.LEVEL_NODE_ALLOC] = locking.ALL_SET
|
6170 |
self.op.src_node = None |
6171 |
if os.path.isabs(src_path):
|
6172 |
raise errors.OpPrereqError("Importing an instance from a path" |
6173 |
" requires a source node option",
|
6174 |
errors.ECODE_INVAL) |
6175 |
else:
|
6176 |
self.op.src_node = src_node = _ExpandNodeName(self.cfg, src_node) |
6177 |
if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET: |
6178 |
self.needed_locks[locking.LEVEL_NODE].append(src_node)
|
6179 |
if not os.path.isabs(src_path): |
6180 |
self.op.src_path = src_path = \
|
6181 |
utils.PathJoin(pathutils.EXPORT_DIR, src_path) |
6182 |
|
6183 |
self.needed_locks[locking.LEVEL_NODE_RES] = \
|
6184 |
_CopyLockList(self.needed_locks[locking.LEVEL_NODE])
|
6185 |
|
6186 |
def _RunAllocator(self): |
6187 |
"""Run the allocator based on input opcode.
|
6188 |
|
6189 |
"""
|
6190 |
if self.op.opportunistic_locking: |
6191 |
# Only consider nodes for which a lock is held
|
6192 |
node_whitelist = list(self.owned_locks(locking.LEVEL_NODE)) |
6193 |
else:
|
6194 |
node_whitelist = None
|
6195 |
|
6196 |
#TODO Export network to iallocator so that it chooses a pnode
|
6197 |
# in a nodegroup that has the desired network connected to
|
6198 |
req = _CreateInstanceAllocRequest(self.op, self.disks, |
6199 |
self.nics, self.be_full, |
6200 |
node_whitelist) |
6201 |
ial = iallocator.IAllocator(self.cfg, self.rpc, req) |
6202 |
|
6203 |
ial.Run(self.op.iallocator)
|
6204 |
|
6205 |
if not ial.success: |
6206 |
# When opportunistic locks are used only a temporary failure is generated
|
6207 |
if self.op.opportunistic_locking: |
6208 |
ecode = errors.ECODE_TEMP_NORES |
6209 |
else:
|
6210 |
ecode = errors.ECODE_NORES |
6211 |
|
6212 |
raise errors.OpPrereqError("Can't compute nodes using" |
6213 |
" iallocator '%s': %s" %
|
6214 |
(self.op.iallocator, ial.info),
|
6215 |
ecode) |
6216 |
|
6217 |
self.op.pnode = ial.result[0] |
6218 |
self.LogInfo("Selected nodes for instance %s via iallocator %s: %s", |
6219 |
self.op.instance_name, self.op.iallocator, |
6220 |
utils.CommaJoin(ial.result)) |
6221 |
|
6222 |
assert req.RequiredNodes() in (1, 2), "Wrong node count from iallocator" |
6223 |
|
6224 |
if req.RequiredNodes() == 2: |
6225 |
self.op.snode = ial.result[1] |
6226 |
|
6227 |
def BuildHooksEnv(self): |
6228 |
"""Build hooks env.
|
6229 |
|
6230 |
This runs on master, primary and secondary nodes of the instance.
|
6231 |
|
6232 |
"""
|
6233 |
env = { |
6234 |
"ADD_MODE": self.op.mode, |
6235 |
} |
6236 |
if self.op.mode == constants.INSTANCE_IMPORT: |
6237 |
env["SRC_NODE"] = self.op.src_node |
6238 |
env["SRC_PATH"] = self.op.src_path |
6239 |
env["SRC_IMAGES"] = self.src_images |
6240 |
|
6241 |
env.update(_BuildInstanceHookEnv( |
6242 |
name=self.op.instance_name,
|
6243 |
primary_node=self.op.pnode,
|
6244 |
secondary_nodes=self.secondaries,
|
6245 |
status=self.op.start,
|
6246 |
os_type=self.op.os_type,
|
6247 |
minmem=self.be_full[constants.BE_MINMEM],
|
6248 |
maxmem=self.be_full[constants.BE_MAXMEM],
|
6249 |
vcpus=self.be_full[constants.BE_VCPUS],
|
6250 |
nics=_NICListToTuple(self, self.nics), |
6251 |
disk_template=self.op.disk_template,
|
6252 |
disks=[(d[constants.IDISK_NAME], d[constants.IDISK_SIZE], |
6253 |
d[constants.IDISK_MODE]) for d in self.disks], |
6254 |
bep=self.be_full,
|
6255 |
hvp=self.hv_full,
|
6256 |
hypervisor_name=self.op.hypervisor,
|
6257 |
tags=self.op.tags,
|
6258 |
)) |
6259 |
|
6260 |
return env
|
6261 |
|
6262 |
def BuildHooksNodes(self): |
6263 |
"""Build hooks nodes.
|
6264 |
|
6265 |
"""
|
6266 |
nl = [self.cfg.GetMasterNode(), self.op.pnode] + self.secondaries |
6267 |
return nl, nl
|
6268 |
|
6269 |
def _ReadExportInfo(self): |
6270 |
"""Reads the export information from disk.
|
6271 |
|
6272 |
It will override the opcode source node and path with the actual
|
6273 |
information, if these two were not specified before.
|
6274 |
|
6275 |
@return: the export information
|
6276 |
|
6277 |
"""
|
6278 |
assert self.op.mode == constants.INSTANCE_IMPORT |
6279 |
|
6280 |
src_node = self.op.src_node
|
6281 |
src_path = self.op.src_path
|
6282 |
|
6283 |
if src_node is None: |
6284 |
locked_nodes = self.owned_locks(locking.LEVEL_NODE)
|
6285 |
exp_list = self.rpc.call_export_list(locked_nodes)
|
6286 |
found = False
|
6287 |
for node in exp_list: |
6288 |
if exp_list[node].fail_msg:
|
6289 |
continue
|
6290 |
if src_path in exp_list[node].payload: |
6291 |
found = True
|
6292 |
self.op.src_node = src_node = node
|
6293 |
self.op.src_path = src_path = utils.PathJoin(pathutils.EXPORT_DIR,
|
6294 |
src_path) |
6295 |
break
|
6296 |
if not found: |
6297 |
raise errors.OpPrereqError("No export found for relative path %s" % |
6298 |
src_path, errors.ECODE_INVAL) |
6299 |
|
6300 |
_CheckNodeOnline(self, src_node)
|
6301 |
result = self.rpc.call_export_info(src_node, src_path)
|
6302 |
result.Raise("No export or invalid export found in dir %s" % src_path)
|
6303 |
|
6304 |
export_info = objects.SerializableConfigParser.Loads(str(result.payload))
|
6305 |
if not export_info.has_section(constants.INISECT_EXP): |
6306 |
raise errors.ProgrammerError("Corrupted export config", |
6307 |
errors.ECODE_ENVIRON) |
6308 |
|
6309 |
ei_version = export_info.get(constants.INISECT_EXP, "version")
|
6310 |
if (int(ei_version) != constants.EXPORT_VERSION): |
6311 |
raise errors.OpPrereqError("Wrong export version %s (wanted %d)" % |
6312 |
(ei_version, constants.EXPORT_VERSION), |
6313 |
errors.ECODE_ENVIRON) |
6314 |
return export_info
|
6315 |
|
6316 |
def _ReadExportParams(self, einfo): |
6317 |
"""Use export parameters as defaults.
|
6318 |
|
6319 |
In case the opcode doesn't specify (as in override) some instance
|
6320 |
parameters, then try to use them from the export information, if
|
6321 |
that declares them.
|
6322 |
|
6323 |
"""
|
6324 |
self.op.os_type = einfo.get(constants.INISECT_EXP, "os") |
6325 |
|
6326 |
if self.op.disk_template is None: |
6327 |
if einfo.has_option(constants.INISECT_INS, "disk_template"): |
6328 |
self.op.disk_template = einfo.get(constants.INISECT_INS,
|
6329 |
"disk_template")
|
6330 |
if self.op.disk_template not in constants.DISK_TEMPLATES: |
6331 |
raise errors.OpPrereqError("Disk template specified in configuration" |
6332 |
" file is not one of the allowed values:"
|
6333 |
" %s" %
|
6334 |
" ".join(constants.DISK_TEMPLATES),
|
6335 |
errors.ECODE_INVAL) |
6336 |
else:
|
6337 |
raise errors.OpPrereqError("No disk template specified and the export" |
6338 |
" is missing the disk_template information",
|
6339 |
errors.ECODE_INVAL) |
6340 |
|
6341 |
if not self.op.disks: |
6342 |
disks = [] |
6343 |
# TODO: import the disk iv_name too
|
6344 |
for idx in range(constants.MAX_DISKS): |
6345 |
if einfo.has_option(constants.INISECT_INS, "disk%d_size" % idx): |
6346 |
disk_sz = einfo.getint(constants.INISECT_INS, "disk%d_size" % idx)
|
6347 |
disks.append({constants.IDISK_SIZE: disk_sz}) |
6348 |
self.op.disks = disks
|
6349 |
if not disks and self.op.disk_template != constants.DT_DISKLESS: |
6350 |
raise errors.OpPrereqError("No disk info specified and the export" |
6351 |
" is missing the disk information",
|
6352 |
errors.ECODE_INVAL) |
6353 |
|
6354 |
if not self.op.nics: |
6355 |
nics = [] |
6356 |
for idx in range(constants.MAX_NICS): |
6357 |
if einfo.has_option(constants.INISECT_INS, "nic%d_mac" % idx): |
6358 |
ndict = {} |
6359 |
for name in list(constants.NICS_PARAMETERS) + ["ip", "mac"]: |
6360 |
v = einfo.get(constants.INISECT_INS, "nic%d_%s" % (idx, name))
|
6361 |
ndict[name] = v |
6362 |
nics.append(ndict) |
6363 |
else:
|
6364 |
break
|
6365 |
self.op.nics = nics
|
6366 |
|
6367 |
if not self.op.tags and einfo.has_option(constants.INISECT_INS, "tags"): |
6368 |
self.op.tags = einfo.get(constants.INISECT_INS, "tags").split() |
6369 |
|
6370 |
if (self.op.hypervisor is None and |
6371 |
einfo.has_option(constants.INISECT_INS, "hypervisor")):
|
6372 |
self.op.hypervisor = einfo.get(constants.INISECT_INS, "hypervisor") |
6373 |
|
6374 |
if einfo.has_section(constants.INISECT_HYP):
|
6375 |
# use the export parameters but do not override the ones
|
6376 |
# specified by the user
|
6377 |
for name, value in einfo.items(constants.INISECT_HYP): |
6378 |
if name not in self.op.hvparams: |
6379 |
self.op.hvparams[name] = value
|
6380 |
|
6381 |
if einfo.has_section(constants.INISECT_BEP):
|
6382 |
# use the parameters, without overriding
|
6383 |
for name, value in einfo.items(constants.INISECT_BEP): |
6384 |
if name not in self.op.beparams: |
6385 |
self.op.beparams[name] = value
|
6386 |
# Compatibility for the old "memory" be param
|
6387 |
if name == constants.BE_MEMORY:
|
6388 |
if constants.BE_MAXMEM not in self.op.beparams: |
6389 |
self.op.beparams[constants.BE_MAXMEM] = value
|
6390 |
if constants.BE_MINMEM not in self.op.beparams: |
6391 |
self.op.beparams[constants.BE_MINMEM] = value
|
6392 |
else:
|
6393 |
# try to read the parameters old style, from the main section
|
6394 |
for name in constants.BES_PARAMETERS: |
6395 |
if (name not in self.op.beparams and |
6396 |
einfo.has_option(constants.INISECT_INS, name)): |
6397 |
self.op.beparams[name] = einfo.get(constants.INISECT_INS, name)
|
6398 |
|
6399 |
if einfo.has_section(constants.INISECT_OSP):
|
6400 |
# use the parameters, without overriding
|
6401 |
for name, value in einfo.items(constants.INISECT_OSP): |
6402 |
if name not in self.op.osparams: |
6403 |
self.op.osparams[name] = value
|
6404 |
|
6405 |
def _RevertToDefaults(self, cluster): |
6406 |
"""Revert the instance parameters to the default values.
|
6407 |
|
6408 |
"""
|
6409 |
# hvparams
|
6410 |
hv_defs = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type, {}) |
6411 |
for name in self.op.hvparams.keys(): |
6412 |
if name in hv_defs and hv_defs[name] == self.op.hvparams[name]: |
6413 |
del self.op.hvparams[name] |
6414 |
# beparams
|
6415 |
be_defs = cluster.SimpleFillBE({}) |
6416 |
for name in self.op.beparams.keys(): |
6417 |
if name in be_defs and be_defs[name] == self.op.beparams[name]: |
6418 |
del self.op.beparams[name] |
6419 |
# nic params
|
6420 |
nic_defs = cluster.SimpleFillNIC({}) |
6421 |
for nic in self.op.nics: |
6422 |
for name in constants.NICS_PARAMETERS: |
6423 |
if name in nic and name in nic_defs and nic[name] == nic_defs[name]: |
6424 |
del nic[name]
|
6425 |
# osparams
|
6426 |
os_defs = cluster.SimpleFillOS(self.op.os_type, {})
|
6427 |
for name in self.op.osparams.keys(): |
6428 |
if name in os_defs and os_defs[name] == self.op.osparams[name]: |
6429 |
del self.op.osparams[name] |
6430 |
|
6431 |
def _CalculateFileStorageDir(self): |
6432 |
"""Calculate final instance file storage dir.
|
6433 |
|
6434 |
"""
|
6435 |
# file storage dir calculation/check
|
6436 |
self.instance_file_storage_dir = None |
6437 |
if self.op.disk_template in constants.DTS_FILEBASED: |
6438 |
# build the full file storage dir path
|
6439 |
joinargs = [] |
6440 |
|
6441 |
if self.op.disk_template == constants.DT_SHARED_FILE: |
6442 |
get_fsd_fn = self.cfg.GetSharedFileStorageDir
|
6443 |
else:
|
6444 |
get_fsd_fn = self.cfg.GetFileStorageDir
|
6445 |
|
6446 |
cfg_storagedir = get_fsd_fn() |
6447 |
if not cfg_storagedir: |
6448 |
raise errors.OpPrereqError("Cluster file storage dir not defined", |
6449 |
errors.ECODE_STATE) |
6450 |
joinargs.append(cfg_storagedir) |
6451 |
|
6452 |
if self.op.file_storage_dir is not None: |
6453 |
joinargs.append(self.op.file_storage_dir)
|
6454 |
|
6455 |
joinargs.append(self.op.instance_name)
|
6456 |
|
6457 |
# pylint: disable=W0142
|
6458 |
self.instance_file_storage_dir = utils.PathJoin(*joinargs)
|
6459 |
|
6460 |
def CheckPrereq(self): # pylint: disable=R0914 |
6461 |
"""Check prerequisites.
|
6462 |
|
6463 |
"""
|
6464 |
self._CalculateFileStorageDir()
|
6465 |
|
6466 |
if self.op.mode == constants.INSTANCE_IMPORT: |
6467 |
export_info = self._ReadExportInfo()
|
6468 |
self._ReadExportParams(export_info)
|
6469 |
self._old_instance_name = export_info.get(constants.INISECT_INS, "name") |
6470 |
else:
|
6471 |
self._old_instance_name = None |
6472 |
|
6473 |
if (not self.cfg.GetVGName() and |
6474 |
self.op.disk_template not in constants.DTS_NOT_LVM): |
6475 |
raise errors.OpPrereqError("Cluster does not support lvm-based" |
6476 |
" instances", errors.ECODE_STATE)
|
6477 |
|
6478 |
if (self.op.hypervisor is None or |
6479 |
self.op.hypervisor == constants.VALUE_AUTO):
|
6480 |
self.op.hypervisor = self.cfg.GetHypervisorType() |
6481 |
|
6482 |
cluster = self.cfg.GetClusterInfo()
|
6483 |
enabled_hvs = cluster.enabled_hypervisors |
6484 |
if self.op.hypervisor not in enabled_hvs: |
6485 |
raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the" |
6486 |
" cluster (%s)" %
|
6487 |
(self.op.hypervisor, ",".join(enabled_hvs)), |
6488 |
errors.ECODE_STATE) |
6489 |
|
6490 |
# Check tag validity
|
6491 |
for tag in self.op.tags: |
6492 |
objects.TaggableObject.ValidateTag(tag) |
6493 |
|
6494 |
# check hypervisor parameter syntax (locally)
|
6495 |
utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
|
6496 |
filled_hvp = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type, |
6497 |
self.op.hvparams)
|
6498 |
hv_type = hypervisor.GetHypervisorClass(self.op.hypervisor)
|
6499 |
hv_type.CheckParameterSyntax(filled_hvp) |
6500 |
self.hv_full = filled_hvp
|
6501 |
# check that we don't specify global parameters on an instance
|
6502 |
_CheckParamsNotGlobal(self.op.hvparams, constants.HVC_GLOBALS, "hypervisor", |
6503 |
"instance", "cluster") |
6504 |
|
6505 |
# fill and remember the beparams dict
|
6506 |
self.be_full = _ComputeFullBeParams(self.op, cluster) |
6507 |
|
6508 |
# build os parameters
|
6509 |
self.os_full = cluster.SimpleFillOS(self.op.os_type, self.op.osparams) |
6510 |
|
6511 |
# now that hvp/bep are in final format, let's reset to defaults,
|
6512 |
# if told to do so
|
6513 |
if self.op.identify_defaults: |
6514 |
self._RevertToDefaults(cluster)
|
6515 |
|
6516 |
# NIC buildup
|
6517 |
self.nics = _ComputeNics(self.op, cluster, self.check_ip, self.cfg, |
6518 |
self.proc.GetECId())
|
6519 |
|
6520 |
# disk checks/pre-build
|
6521 |
default_vg = self.cfg.GetVGName()
|
6522 |
self.disks = _ComputeDisks(self.op, default_vg) |
6523 |
|
6524 |
if self.op.mode == constants.INSTANCE_IMPORT: |
6525 |
disk_images = [] |
6526 |
for idx in range(len(self.disks)): |
6527 |
option = "disk%d_dump" % idx
|
6528 |
if export_info.has_option(constants.INISECT_INS, option):
|
6529 |
# FIXME: are the old os-es, disk sizes, etc. useful?
|
6530 |
export_name = export_info.get(constants.INISECT_INS, option) |
6531 |
image = utils.PathJoin(self.op.src_path, export_name)
|
6532 |
disk_images.append(image) |
6533 |
else:
|
6534 |
disk_images.append(False)
|
6535 |
|
6536 |
self.src_images = disk_images
|
6537 |
|
6538 |
if self.op.instance_name == self._old_instance_name: |
6539 |
for idx, nic in enumerate(self.nics): |
6540 |
if nic.mac == constants.VALUE_AUTO:
|
6541 |
nic_mac_ini = "nic%d_mac" % idx
|
6542 |
nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini) |
6543 |
|
6544 |
# ENDIF: self.op.mode == constants.INSTANCE_IMPORT
|
6545 |
|
6546 |
# ip ping checks (we use the same ip that was resolved in ExpandNames)
|
6547 |
if self.op.ip_check: |
6548 |
if netutils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT): |
6549 |
raise errors.OpPrereqError("IP %s of instance %s already in use" % |
6550 |
(self.check_ip, self.op.instance_name), |
6551 |
errors.ECODE_NOTUNIQUE) |
6552 |
|
6553 |
#### mac address generation
|
6554 |
# By generating here the mac address both the allocator and the hooks get
|
6555 |
# the real final mac address rather than the 'auto' or 'generate' value.
|
6556 |
# There is a race condition between the generation and the instance object
|
6557 |
# creation, which means that we know the mac is valid now, but we're not
|
6558 |
# sure it will be when we actually add the instance. If things go bad
|
6559 |
# adding the instance will abort because of a duplicate mac, and the
|
6560 |
# creation job will fail.
|
6561 |
for nic in self.nics: |
6562 |
if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE): |
6563 |
nic.mac = self.cfg.GenerateMAC(nic.network, self.proc.GetECId()) |
6564 |
|
6565 |
#### allocator run
|
6566 |
|
6567 |
if self.op.iallocator is not None: |
6568 |
self._RunAllocator()
|
6569 |
|
6570 |
# Release all unneeded node locks
|
6571 |
keep_locks = filter(None, [self.op.pnode, self.op.snode, self.op.src_node]) |
6572 |
_ReleaseLocks(self, locking.LEVEL_NODE, keep=keep_locks)
|
6573 |
_ReleaseLocks(self, locking.LEVEL_NODE_RES, keep=keep_locks)
|
6574 |
_ReleaseLocks(self, locking.LEVEL_NODE_ALLOC)
|
6575 |
|
6576 |
assert (self.owned_locks(locking.LEVEL_NODE) == |
6577 |
self.owned_locks(locking.LEVEL_NODE_RES)), \
|
6578 |
"Node locks differ from node resource locks"
|
6579 |
|
6580 |
#### node related checks
|
6581 |
|
6582 |
# check primary node
|
6583 |
self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode) |
6584 |
assert self.pnode is not None, \ |
6585 |
"Cannot retrieve locked node %s" % self.op.pnode |
6586 |
if pnode.offline:
|
6587 |
raise errors.OpPrereqError("Cannot use offline primary node '%s'" % |
6588 |
pnode.name, errors.ECODE_STATE) |
6589 |
if pnode.drained:
|
6590 |
raise errors.OpPrereqError("Cannot use drained primary node '%s'" % |
6591 |
pnode.name, errors.ECODE_STATE) |
6592 |
if not pnode.vm_capable: |
6593 |
raise errors.OpPrereqError("Cannot use non-vm_capable primary node" |
6594 |
" '%s'" % pnode.name, errors.ECODE_STATE)
|
6595 |
|
6596 |
self.secondaries = []
|
6597 |
|
6598 |
# Fill in any IPs from IP pools. This must happen here, because we need to
|
6599 |
# know the nic's primary node, as specified by the iallocator
|
6600 |
for idx, nic in enumerate(self.nics): |
6601 |
net_uuid = nic.network |
6602 |
if net_uuid is not None: |
6603 |
nobj = self.cfg.GetNetwork(net_uuid)
|
6604 |
netparams = self.cfg.GetGroupNetParams(net_uuid, self.pnode.name) |
6605 |
if netparams is None: |
6606 |
raise errors.OpPrereqError("No netparams found for network" |
6607 |
" %s. Propably not connected to"
|
6608 |
" node's %s nodegroup" %
|
6609 |
(nobj.name, self.pnode.name),
|
6610 |
errors.ECODE_INVAL) |
6611 |
self.LogInfo("NIC/%d inherits netparams %s" % |
6612 |
(idx, netparams.values())) |
6613 |
nic.nicparams = dict(netparams)
|
6614 |
if nic.ip is not None: |
6615 |
if nic.ip.lower() == constants.NIC_IP_POOL:
|
6616 |
try:
|
6617 |
nic.ip = self.cfg.GenerateIp(net_uuid, self.proc.GetECId()) |
6618 |
except errors.ReservationError:
|
6619 |
raise errors.OpPrereqError("Unable to get a free IP for NIC %d" |
6620 |
" from the address pool" % idx,
|
6621 |
errors.ECODE_STATE) |
6622 |
self.LogInfo("Chose IP %s from network %s", nic.ip, nobj.name) |
6623 |
else:
|
6624 |
try:
|
6625 |
self.cfg.ReserveIp(net_uuid, nic.ip, self.proc.GetECId()) |
6626 |
except errors.ReservationError:
|
6627 |
raise errors.OpPrereqError("IP address %s already in use" |
6628 |
" or does not belong to network %s" %
|
6629 |
(nic.ip, nobj.name), |
6630 |
errors.ECODE_NOTUNIQUE) |
6631 |
|
6632 |
# net is None, ip None or given
|
6633 |
elif self.op.conflicts_check: |
6634 |
_CheckForConflictingIp(self, nic.ip, self.pnode.name) |
6635 |
|
6636 |
# mirror node verification
|
6637 |
if self.op.disk_template in constants.DTS_INT_MIRROR: |
6638 |
if self.op.snode == pnode.name: |
6639 |
raise errors.OpPrereqError("The secondary node cannot be the" |
6640 |
" primary node", errors.ECODE_INVAL)
|
6641 |
_CheckNodeOnline(self, self.op.snode) |
6642 |
_CheckNodeNotDrained(self, self.op.snode) |
6643 |
_CheckNodeVmCapable(self, self.op.snode) |
6644 |
self.secondaries.append(self.op.snode) |
6645 |
|
6646 |
snode = self.cfg.GetNodeInfo(self.op.snode) |
6647 |
if pnode.group != snode.group:
|
6648 |
self.LogWarning("The primary and secondary nodes are in two" |
6649 |
" different node groups; the disk parameters"
|
6650 |
" from the first disk's node group will be"
|
6651 |
" used")
|
6652 |
|
6653 |
if not self.op.disk_template in constants.DTS_EXCL_STORAGE: |
6654 |
nodes = [pnode] |
6655 |
if self.op.disk_template in constants.DTS_INT_MIRROR: |
6656 |
nodes.append(snode) |
6657 |
has_es = lambda n: _IsExclusiveStorageEnabledNode(self.cfg, n) |
6658 |
if compat.any(map(has_es, nodes)): |
6659 |
raise errors.OpPrereqError("Disk template %s not supported with" |
6660 |
" exclusive storage" % self.op.disk_template, |
6661 |
errors.ECODE_STATE) |
6662 |
|
6663 |
nodenames = [pnode.name] + self.secondaries
|
6664 |
|
6665 |
if not self.adopt_disks: |
6666 |
if self.op.disk_template == constants.DT_RBD: |
6667 |
# _CheckRADOSFreeSpace() is just a placeholder.
|
6668 |
# Any function that checks prerequisites can be placed here.
|
6669 |
# Check if there is enough space on the RADOS cluster.
|
6670 |
_CheckRADOSFreeSpace() |
6671 |
elif self.op.disk_template == constants.DT_EXT: |
6672 |
# FIXME: Function that checks prereqs if needed
|
6673 |
pass
|
6674 |
else:
|
6675 |
# Check lv size requirements, if not adopting
|
6676 |
req_sizes = _ComputeDiskSizePerVG(self.op.disk_template, self.disks) |
6677 |
_CheckNodesFreeDiskPerVG(self, nodenames, req_sizes)
|
6678 |
|
6679 |
elif self.op.disk_template == constants.DT_PLAIN: # Check the adoption data |
6680 |
all_lvs = set(["%s/%s" % (disk[constants.IDISK_VG], |
6681 |
disk[constants.IDISK_ADOPT]) |
6682 |
for disk in self.disks]) |
6683 |
if len(all_lvs) != len(self.disks): |
6684 |
raise errors.OpPrereqError("Duplicate volume names given for adoption", |
6685 |
errors.ECODE_INVAL) |
6686 |
for lv_name in all_lvs: |
6687 |
try:
|
6688 |
# FIXME: lv_name here is "vg/lv" need to ensure that other calls
|
6689 |
# to ReserveLV uses the same syntax
|
6690 |
self.cfg.ReserveLV(lv_name, self.proc.GetECId()) |
6691 |
except errors.ReservationError:
|
6692 |
raise errors.OpPrereqError("LV named %s used by another instance" % |
6693 |
lv_name, errors.ECODE_NOTUNIQUE) |
6694 |
|
6695 |
vg_names = self.rpc.call_vg_list([pnode.name])[pnode.name]
|
6696 |
vg_names.Raise("Cannot get VG information from node %s" % pnode.name)
|
6697 |
|
6698 |
node_lvs = self.rpc.call_lv_list([pnode.name],
|
6699 |
vg_names.payload.keys())[pnode.name] |
6700 |
node_lvs.Raise("Cannot get LV information from node %s" % pnode.name)
|
6701 |
node_lvs = node_lvs.payload |
6702 |
|
6703 |
delta = all_lvs.difference(node_lvs.keys()) |
6704 |
if delta:
|
6705 |
raise errors.OpPrereqError("Missing logical volume(s): %s" % |
6706 |
utils.CommaJoin(delta), |
6707 |
errors.ECODE_INVAL) |
6708 |
online_lvs = [lv for lv in all_lvs if node_lvs[lv][2]] |
6709 |
if online_lvs:
|
6710 |
raise errors.OpPrereqError("Online logical volumes found, cannot" |
6711 |
" adopt: %s" % utils.CommaJoin(online_lvs),
|
6712 |
errors.ECODE_STATE) |
6713 |
# update the size of disk based on what is found
|
6714 |
for dsk in self.disks: |
6715 |
dsk[constants.IDISK_SIZE] = \ |
6716 |
int(float(node_lvs["%s/%s" % (dsk[constants.IDISK_VG], |
6717 |
dsk[constants.IDISK_ADOPT])][0]))
|
6718 |
|
6719 |
elif self.op.disk_template == constants.DT_BLOCK: |
6720 |
# Normalize and de-duplicate device paths
|
6721 |
all_disks = set([os.path.abspath(disk[constants.IDISK_ADOPT])
|
6722 |
for disk in self.disks]) |
6723 |
if len(all_disks) != len(self.disks): |
6724 |
raise errors.OpPrereqError("Duplicate disk names given for adoption", |
6725 |
errors.ECODE_INVAL) |
6726 |
baddisks = [d for d in all_disks |
6727 |
if not d.startswith(constants.ADOPTABLE_BLOCKDEV_ROOT)] |
6728 |
if baddisks:
|
6729 |
raise errors.OpPrereqError("Device node(s) %s lie outside %s and" |
6730 |
" cannot be adopted" %
|
6731 |
(utils.CommaJoin(baddisks), |
6732 |
constants.ADOPTABLE_BLOCKDEV_ROOT), |
6733 |
errors.ECODE_INVAL) |
6734 |
|
6735 |
node_disks = self.rpc.call_bdev_sizes([pnode.name],
|
6736 |
list(all_disks))[pnode.name]
|
6737 |
node_disks.Raise("Cannot get block device information from node %s" %
|
6738 |
pnode.name) |
6739 |
node_disks = node_disks.payload |
6740 |
delta = all_disks.difference(node_disks.keys()) |
6741 |
if delta:
|
6742 |
raise errors.OpPrereqError("Missing block device(s): %s" % |
6743 |
utils.CommaJoin(delta), |
6744 |
errors.ECODE_INVAL) |
6745 |
for dsk in self.disks: |
6746 |
dsk[constants.IDISK_SIZE] = \ |
6747 |
int(float(node_disks[dsk[constants.IDISK_ADOPT]])) |
6748 |
|
6749 |
# Verify instance specs
|
6750 |
spindle_use = self.be_full.get(constants.BE_SPINDLE_USE, None) |
6751 |
ispec = { |
6752 |
constants.ISPEC_MEM_SIZE: self.be_full.get(constants.BE_MAXMEM, None), |
6753 |
constants.ISPEC_CPU_COUNT: self.be_full.get(constants.BE_VCPUS, None), |
6754 |
constants.ISPEC_DISK_COUNT: len(self.disks), |
6755 |
constants.ISPEC_DISK_SIZE: [disk[constants.IDISK_SIZE] |
6756 |
for disk in self.disks], |
6757 |
constants.ISPEC_NIC_COUNT: len(self.nics), |
6758 |
constants.ISPEC_SPINDLE_USE: spindle_use, |
6759 |
} |
6760 |
|
6761 |
group_info = self.cfg.GetNodeGroup(pnode.group)
|
6762 |
ipolicy = ganeti.masterd.instance.CalculateGroupIPolicy(cluster, group_info) |
6763 |
res = _ComputeIPolicyInstanceSpecViolation(ipolicy, ispec, |
6764 |
self.op.disk_template)
|
6765 |
if not self.op.ignore_ipolicy and res: |
6766 |
msg = ("Instance allocation to group %s (%s) violates policy: %s" %
|
6767 |
(pnode.group, group_info.name, utils.CommaJoin(res))) |
6768 |
raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
|
6769 |
|
6770 |
_CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams) |
6771 |
|
6772 |
_CheckNodeHasOS(self, pnode.name, self.op.os_type, self.op.force_variant) |
6773 |
# check OS parameters (remotely)
|
6774 |
_CheckOSParams(self, True, nodenames, self.op.os_type, self.os_full) |
6775 |
|
6776 |
_CheckNicsBridgesExist(self, self.nics, self.pnode.name) |
6777 |
|
6778 |
#TODO: _CheckExtParams (remotely)
|
6779 |
# Check parameters for extstorage
|
6780 |
|
6781 |
# memory check on primary node
|
6782 |
#TODO(dynmem): use MINMEM for checking
|
6783 |
if self.op.start: |
6784 |
_CheckNodeFreeMemory(self, self.pnode.name, |
6785 |
"creating instance %s" % self.op.instance_name, |
6786 |
self.be_full[constants.BE_MAXMEM],
|
6787 |
self.op.hypervisor)
|
6788 |
|
6789 |
self.dry_run_result = list(nodenames) |
6790 |
|
6791 |
def Exec(self, feedback_fn): |
6792 |
"""Create and add the instance to the cluster.
|
6793 |
|
6794 |
"""
|
6795 |
instance = self.op.instance_name
|
6796 |
pnode_name = self.pnode.name
|
6797 |
|
6798 |
assert not (self.owned_locks(locking.LEVEL_NODE_RES) - |
6799 |
self.owned_locks(locking.LEVEL_NODE)), \
|
6800 |
"Node locks differ from node resource locks"
|
6801 |
assert not self.glm.is_owned(locking.LEVEL_NODE_ALLOC) |
6802 |
|
6803 |
ht_kind = self.op.hypervisor
|
6804 |
if ht_kind in constants.HTS_REQ_PORT: |
6805 |
network_port = self.cfg.AllocatePort()
|
6806 |
else:
|
6807 |
network_port = None
|
6808 |
|
6809 |
# This is ugly but we got a chicken-egg problem here
|
6810 |
# We can only take the group disk parameters, as the instance
|
6811 |
# has no disks yet (we are generating them right here).
|
6812 |
node = self.cfg.GetNodeInfo(pnode_name)
|
6813 |
nodegroup = self.cfg.GetNodeGroup(node.group)
|
6814 |
disks = _GenerateDiskTemplate(self,
|
6815 |
self.op.disk_template,
|
6816 |
instance, pnode_name, |
6817 |
self.secondaries,
|
6818 |
self.disks,
|
6819 |
self.instance_file_storage_dir,
|
6820 |
self.op.file_driver,
|
6821 |
0,
|
6822 |
feedback_fn, |
6823 |
self.cfg.GetGroupDiskParams(nodegroup))
|
6824 |
|
6825 |
iobj = objects.Instance(name=instance, os=self.op.os_type,
|
6826 |
primary_node=pnode_name, |
6827 |
nics=self.nics, disks=disks,
|
6828 |
disk_template=self.op.disk_template,
|
6829 |
admin_state=constants.ADMINST_DOWN, |
6830 |
network_port=network_port, |
6831 |
beparams=self.op.beparams,
|
6832 |
hvparams=self.op.hvparams,
|
6833 |
hypervisor=self.op.hypervisor,
|
6834 |
osparams=self.op.osparams,
|
6835 |
) |
6836 |
|
6837 |
if self.op.tags: |
6838 |
for tag in self.op.tags: |
6839 |
iobj.AddTag(tag) |
6840 |
|
6841 |
if self.adopt_disks: |
6842 |
if self.op.disk_template == constants.DT_PLAIN: |
6843 |
# rename LVs to the newly-generated names; we need to construct
|
6844 |
# 'fake' LV disks with the old data, plus the new unique_id
|
6845 |
tmp_disks = [objects.Disk.FromDict(v.ToDict()) for v in disks] |
6846 |
rename_to = [] |
6847 |
for t_dsk, a_dsk in zip(tmp_disks, self.disks): |
6848 |
rename_to.append(t_dsk.logical_id) |
6849 |
t_dsk.logical_id = (t_dsk.logical_id[0], a_dsk[constants.IDISK_ADOPT])
|
6850 |
self.cfg.SetDiskID(t_dsk, pnode_name)
|
6851 |
result = self.rpc.call_blockdev_rename(pnode_name,
|
6852 |
zip(tmp_disks, rename_to))
|
6853 |
result.Raise("Failed to rename adoped LVs")
|
6854 |
else:
|
6855 |
feedback_fn("* creating instance disks...")
|
6856 |
try:
|
6857 |
_CreateDisks(self, iobj)
|
6858 |
except errors.OpExecError:
|
6859 |
self.LogWarning("Device creation failed") |
6860 |
self.cfg.ReleaseDRBDMinors(instance)
|
6861 |
raise
|
6862 |
|
6863 |
feedback_fn("adding instance %s to cluster config" % instance)
|
6864 |
|
6865 |
self.cfg.AddInstance(iobj, self.proc.GetECId()) |
6866 |
|
6867 |
# Declare that we don't want to remove the instance lock anymore, as we've
|
6868 |
# added the instance to the config
|
6869 |
del self.remove_locks[locking.LEVEL_INSTANCE] |
6870 |
|
6871 |
if self.op.mode == constants.INSTANCE_IMPORT: |
6872 |
# Release unused nodes
|
6873 |
_ReleaseLocks(self, locking.LEVEL_NODE, keep=[self.op.src_node]) |
6874 |
else:
|
6875 |
# Release all nodes
|
6876 |
_ReleaseLocks(self, locking.LEVEL_NODE)
|
6877 |
|
6878 |
disk_abort = False
|
6879 |
if not self.adopt_disks and self.cfg.GetClusterInfo().prealloc_wipe_disks: |
6880 |
feedback_fn("* wiping instance disks...")
|
6881 |
try:
|
6882 |
_WipeDisks(self, iobj)
|
6883 |
except errors.OpExecError, err:
|
6884 |
logging.exception("Wiping disks failed")
|
6885 |
self.LogWarning("Wiping instance disks failed (%s)", err) |
6886 |
disk_abort = True
|
6887 |
|
6888 |
if disk_abort:
|
6889 |
# Something is already wrong with the disks, don't do anything else
|
6890 |
pass
|
6891 |
elif self.op.wait_for_sync: |
6892 |
disk_abort = not _WaitForSync(self, iobj) |
6893 |
elif iobj.disk_template in constants.DTS_INT_MIRROR: |
6894 |
# make sure the disks are not degraded (still sync-ing is ok)
|
6895 |
feedback_fn("* checking mirrors status")
|
6896 |
disk_abort = not _WaitForSync(self, iobj, oneshot=True) |
6897 |
else:
|
6898 |
disk_abort = False
|
6899 |
|
6900 |
if disk_abort:
|
6901 |
_RemoveDisks(self, iobj)
|
6902 |
self.cfg.RemoveInstance(iobj.name)
|
6903 |
# Make sure the instance lock gets removed
|
6904 |
self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
|
6905 |
raise errors.OpExecError("There are some degraded disks for" |
6906 |
" this instance")
|
6907 |
|
6908 |
# Release all node resource locks
|
6909 |
_ReleaseLocks(self, locking.LEVEL_NODE_RES)
|
6910 |
|
6911 |
if iobj.disk_template != constants.DT_DISKLESS and not self.adopt_disks: |
6912 |
# we need to set the disks ID to the primary node, since the
|
6913 |
# preceding code might or might have not done it, depending on
|
6914 |
# disk template and other options
|
6915 |
for disk in iobj.disks: |
6916 |
self.cfg.SetDiskID(disk, pnode_name)
|
6917 |
if self.op.mode == constants.INSTANCE_CREATE: |
6918 |
if not self.op.no_install: |
6919 |
pause_sync = (iobj.disk_template in constants.DTS_INT_MIRROR and |
6920 |
not self.op.wait_for_sync) |
6921 |
if pause_sync:
|
6922 |
feedback_fn("* pausing disk sync to install instance OS")
|
6923 |
result = self.rpc.call_blockdev_pause_resume_sync(pnode_name,
|
6924 |
(iobj.disks, |
6925 |
iobj), True)
|
6926 |
for idx, success in enumerate(result.payload): |
6927 |
if not success: |
6928 |
logging.warn("pause-sync of instance %s for disk %d failed",
|
6929 |
instance, idx) |
6930 |
|
6931 |
feedback_fn("* running the instance OS create scripts...")
|
6932 |
# FIXME: pass debug option from opcode to backend
|
6933 |
os_add_result = \ |
6934 |
self.rpc.call_instance_os_add(pnode_name, (iobj, None), False, |
6935 |
self.op.debug_level)
|
6936 |
if pause_sync:
|
6937 |
feedback_fn("* resuming disk sync")
|
6938 |
result = self.rpc.call_blockdev_pause_resume_sync(pnode_name,
|
6939 |
(iobj.disks, |
6940 |
iobj), False)
|
6941 |
for idx, success in enumerate(result.payload): |
6942 |
if not success: |
6943 |
logging.warn("resume-sync of instance %s for disk %d failed",
|
6944 |
instance, idx) |
6945 |
|
6946 |
os_add_result.Raise("Could not add os for instance %s"
|
6947 |
" on node %s" % (instance, pnode_name))
|
6948 |
|
6949 |
else:
|
6950 |
if self.op.mode == constants.INSTANCE_IMPORT: |
6951 |
feedback_fn("* running the instance OS import scripts...")
|
6952 |
|
6953 |
transfers = [] |
6954 |
|
6955 |
for idx, image in enumerate(self.src_images): |
6956 |
if not image: |
6957 |
continue
|
6958 |
|
6959 |
# FIXME: pass debug option from opcode to backend
|
6960 |
dt = masterd.instance.DiskTransfer("disk/%s" % idx,
|
6961 |
constants.IEIO_FILE, (image, ), |
6962 |
constants.IEIO_SCRIPT, |
6963 |
(iobj.disks[idx], idx), |
6964 |
None)
|
6965 |
transfers.append(dt) |
6966 |
|
6967 |
import_result = \ |
6968 |
masterd.instance.TransferInstanceData(self, feedback_fn,
|
6969 |
self.op.src_node, pnode_name,
|
6970 |
self.pnode.secondary_ip,
|
6971 |
iobj, transfers) |
6972 |
if not compat.all(import_result): |
6973 |
self.LogWarning("Some disks for instance %s on node %s were not" |
6974 |
" imported successfully" % (instance, pnode_name))
|
6975 |
|
6976 |
rename_from = self._old_instance_name
|
6977 |
|
6978 |
elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT: |
6979 |
feedback_fn("* preparing remote import...")
|
6980 |
# The source cluster will stop the instance before attempting to make
|
6981 |
# a connection. In some cases stopping an instance can take a long
|
6982 |
# time, hence the shutdown timeout is added to the connection
|
6983 |
# timeout.
|
6984 |
connect_timeout = (constants.RIE_CONNECT_TIMEOUT + |
6985 |
self.op.source_shutdown_timeout)
|
6986 |
timeouts = masterd.instance.ImportExportTimeouts(connect_timeout) |
6987 |
|
6988 |
assert iobj.primary_node == self.pnode.name |
6989 |
disk_results = \ |
6990 |
masterd.instance.RemoteImport(self, feedback_fn, iobj, self.pnode, |
6991 |
self.source_x509_ca,
|
6992 |
self._cds, timeouts)
|
6993 |
if not compat.all(disk_results): |
6994 |
# TODO: Should the instance still be started, even if some disks
|
6995 |
# failed to import (valid for local imports, too)?
|
6996 |
self.LogWarning("Some disks for instance %s on node %s were not" |
6997 |
" imported successfully" % (instance, pnode_name))
|
6998 |
|
6999 |
rename_from = self.source_instance_name
|
7000 |
|
7001 |
else:
|
7002 |
# also checked in the prereq part
|
7003 |
raise errors.ProgrammerError("Unknown OS initialization mode '%s'" |
7004 |
% self.op.mode)
|
7005 |
|
7006 |
# Run rename script on newly imported instance
|
7007 |
assert iobj.name == instance
|
7008 |
feedback_fn("Running rename script for %s" % instance)
|
7009 |
result = self.rpc.call_instance_run_rename(pnode_name, iobj,
|
7010 |
rename_from, |
7011 |
self.op.debug_level)
|
7012 |
if result.fail_msg:
|
7013 |
self.LogWarning("Failed to run rename script for %s on node" |
7014 |
" %s: %s" % (instance, pnode_name, result.fail_msg))
|
7015 |
|
7016 |
assert not self.owned_locks(locking.LEVEL_NODE_RES) |
7017 |
|
7018 |
if self.op.start: |
7019 |
iobj.admin_state = constants.ADMINST_UP |
7020 |
self.cfg.Update(iobj, feedback_fn)
|
7021 |
logging.info("Starting instance %s on node %s", instance, pnode_name)
|
7022 |
feedback_fn("* starting instance...")
|
7023 |
result = self.rpc.call_instance_start(pnode_name, (iobj, None, None), |
7024 |
False, self.op.reason) |
7025 |
result.Raise("Could not start instance")
|
7026 |
|
7027 |
return list(iobj.all_nodes) |
7028 |
|
7029 |
|
7030 |
class LUInstanceMultiAlloc(NoHooksLU): |
7031 |
"""Allocates multiple instances at the same time.
|
7032 |
|
7033 |
"""
|
7034 |
REQ_BGL = False
|
7035 |
|
7036 |
def CheckArguments(self): |
7037 |
"""Check arguments.
|
7038 |
|
7039 |
"""
|
7040 |
nodes = [] |
7041 |
for inst in self.op.instances: |
7042 |
if inst.iallocator is not None: |
7043 |
raise errors.OpPrereqError("iallocator are not allowed to be set on" |
7044 |
" instance objects", errors.ECODE_INVAL)
|
7045 |
nodes.append(bool(inst.pnode))
|
7046 |
if inst.disk_template in constants.DTS_INT_MIRROR: |
7047 |
nodes.append(bool(inst.snode))
|
7048 |
|
7049 |
has_nodes = compat.any(nodes) |
7050 |
if compat.all(nodes) ^ has_nodes:
|
7051 |
raise errors.OpPrereqError("There are instance objects providing" |
7052 |
" pnode/snode while others do not",
|
7053 |
errors.ECODE_INVAL) |
7054 |
|
7055 |
if self.op.iallocator is None: |
7056 |
default_iallocator = self.cfg.GetDefaultIAllocator()
|
7057 |
if default_iallocator and has_nodes: |
7058 |
self.op.iallocator = default_iallocator
|
7059 |
else:
|
7060 |
raise errors.OpPrereqError("No iallocator or nodes on the instances" |
7061 |
" given and no cluster-wide default"
|
7062 |
" iallocator found; please specify either"
|
7063 |
" an iallocator or nodes on the instances"
|
7064 |
" or set a cluster-wide default iallocator",
|
7065 |
errors.ECODE_INVAL) |
7066 |
|
7067 |
_CheckOpportunisticLocking(self.op)
|
7068 |
|
7069 |
dups = utils.FindDuplicates([op.instance_name for op in self.op.instances]) |
7070 |
if dups:
|
7071 |
raise errors.OpPrereqError("There are duplicate instance names: %s" % |
7072 |
utils.CommaJoin(dups), errors.ECODE_INVAL) |
7073 |
|
7074 |
def ExpandNames(self): |
7075 |
"""Calculate the locks.
|
7076 |
|
7077 |
"""
|
7078 |
self.share_locks = _ShareAll()
|
7079 |
self.needed_locks = {
|
7080 |
# iallocator will select nodes and even if no iallocator is used,
|
7081 |
# collisions with LUInstanceCreate should be avoided
|
7082 |
locking.LEVEL_NODE_ALLOC: locking.ALL_SET, |
7083 |
} |
7084 |
|
7085 |
if self.op.iallocator: |
7086 |
self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
|
7087 |
self.needed_locks[locking.LEVEL_NODE_RES] = locking.ALL_SET
|
7088 |
|
7089 |
if self.op.opportunistic_locking: |
7090 |
self.opportunistic_locks[locking.LEVEL_NODE] = True |
7091 |
self.opportunistic_locks[locking.LEVEL_NODE_RES] = True |
7092 |
else:
|
7093 |
nodeslist = [] |
7094 |
for inst in self.op.instances: |
7095 |
inst.pnode = _ExpandNodeName(self.cfg, inst.pnode)
|
7096 |
nodeslist.append(inst.pnode) |
7097 |
if inst.snode is not None: |
7098 |
inst.snode = _ExpandNodeName(self.cfg, inst.snode)
|
7099 |
nodeslist.append(inst.snode) |
7100 |
|
7101 |
self.needed_locks[locking.LEVEL_NODE] = nodeslist
|
7102 |
# Lock resources of instance's primary and secondary nodes (copy to
|
7103 |
# prevent accidential modification)
|
7104 |
self.needed_locks[locking.LEVEL_NODE_RES] = list(nodeslist) |
7105 |
|
7106 |
def CheckPrereq(self): |
7107 |
"""Check prerequisite.
|
7108 |
|
7109 |
"""
|
7110 |
cluster = self.cfg.GetClusterInfo()
|
7111 |
default_vg = self.cfg.GetVGName()
|
7112 |
ec_id = self.proc.GetECId()
|
7113 |
|
7114 |
if self.op.opportunistic_locking: |
7115 |
# Only consider nodes for which a lock is held
|
7116 |
node_whitelist = list(self.owned_locks(locking.LEVEL_NODE)) |
7117 |
else:
|
7118 |
node_whitelist = None
|
7119 |
|
7120 |
insts = [_CreateInstanceAllocRequest(op, _ComputeDisks(op, default_vg), |
7121 |
_ComputeNics(op, cluster, None,
|
7122 |
self.cfg, ec_id),
|
7123 |
_ComputeFullBeParams(op, cluster), |
7124 |
node_whitelist) |
7125 |
for op in self.op.instances] |
7126 |
|
7127 |
req = iallocator.IAReqMultiInstanceAlloc(instances=insts) |
7128 |
ial = iallocator.IAllocator(self.cfg, self.rpc, req) |
7129 |
|
7130 |
ial.Run(self.op.iallocator)
|
7131 |
|
7132 |
if not ial.success: |
7133 |
raise errors.OpPrereqError("Can't compute nodes using" |
7134 |
" iallocator '%s': %s" %
|
7135 |
(self.op.iallocator, ial.info),
|
7136 |
errors.ECODE_NORES) |
7137 |
|
7138 |
self.ia_result = ial.result
|
7139 |
|
7140 |
if self.op.dry_run: |
7141 |
self.dry_run_result = objects.FillDict(self._ConstructPartialResult(), { |
7142 |
constants.JOB_IDS_KEY: [], |
7143 |
}) |
7144 |
|
7145 |
def _ConstructPartialResult(self): |
7146 |
"""Contructs the partial result.
|
7147 |
|
7148 |
"""
|
7149 |
(allocatable, failed) = self.ia_result
|
7150 |
return {
|
7151 |
opcodes.OpInstanceMultiAlloc.ALLOCATABLE_KEY: |
7152 |
map(compat.fst, allocatable),
|
7153 |
opcodes.OpInstanceMultiAlloc.FAILED_KEY: failed, |
7154 |
} |
7155 |
|
7156 |
def Exec(self, feedback_fn): |
7157 |
"""Executes the opcode.
|
7158 |
|
7159 |
"""
|
7160 |
op2inst = dict((op.instance_name, op) for op in self.op.instances) |
7161 |
(allocatable, failed) = self.ia_result
|
7162 |
|
7163 |
jobs = [] |
7164 |
for (name, nodes) in allocatable: |
7165 |
op = op2inst.pop(name) |
7166 |
|
7167 |
if len(nodes) > 1: |
7168 |
(op.pnode, op.snode) = nodes |
7169 |
else:
|
7170 |
(op.pnode,) = nodes |
7171 |
|
7172 |
jobs.append([op]) |
7173 |
|
7174 |
missing = set(op2inst.keys()) - set(failed) |
7175 |
assert not missing, \ |
7176 |
"Iallocator did return incomplete result: %s" % utils.CommaJoin(missing)
|
7177 |
|
7178 |
return ResultWithJobs(jobs, **self._ConstructPartialResult()) |
7179 |
|
7180 |
|
7181 |
def _CheckRADOSFreeSpace(): |
7182 |
"""Compute disk size requirements inside the RADOS cluster.
|
7183 |
|
7184 |
"""
|
7185 |
# For the RADOS cluster we assume there is always enough space.
|
7186 |
pass
|
7187 |
|
7188 |
|
7189 |
class LUInstanceConsole(NoHooksLU): |
7190 |
"""Connect to an instance's console.
|
7191 |
|
7192 |
This is somewhat special in that it returns the command line that
|
7193 |
you need to run on the master node in order to connect to the
|
7194 |
console.
|
7195 |
|
7196 |
"""
|
7197 |
REQ_BGL = False
|
7198 |
|
7199 |
def ExpandNames(self): |
7200 |
self.share_locks = _ShareAll()
|
7201 |
self._ExpandAndLockInstance()
|
7202 |
|
7203 |
def CheckPrereq(self): |
7204 |
"""Check prerequisites.
|
7205 |
|
7206 |
This checks that the instance is in the cluster.
|
7207 |
|
7208 |
"""
|
7209 |
self.instance = self.cfg.GetInstanceInfo(self.op.instance_name) |
7210 |
assert self.instance is not None, \ |
7211 |
"Cannot retrieve locked instance %s" % self.op.instance_name |
7212 |
_CheckNodeOnline(self, self.instance.primary_node) |
7213 |
|
7214 |
def Exec(self, feedback_fn): |
7215 |
"""Connect to the console of an instance
|
7216 |
|
7217 |
"""
|
7218 |
instance = self.instance
|
7219 |
node = instance.primary_node |
7220 |
|
7221 |
node_insts = self.rpc.call_instance_list([node],
|
7222 |
[instance.hypervisor])[node] |
7223 |
node_insts.Raise("Can't get node information from %s" % node)
|
7224 |
|
7225 |
if instance.name not in node_insts.payload: |
7226 |
if instance.admin_state == constants.ADMINST_UP:
|
7227 |
state = constants.INSTST_ERRORDOWN |
7228 |
elif instance.admin_state == constants.ADMINST_DOWN:
|
7229 |
state = constants.INSTST_ADMINDOWN |
7230 |
else:
|
7231 |
state = constants.INSTST_ADMINOFFLINE |
7232 |
raise errors.OpExecError("Instance %s is not running (state %s)" % |
7233 |
(instance.name, state)) |
7234 |
|
7235 |
logging.debug("Connecting to console of %s on %s", instance.name, node)
|
7236 |
|
7237 |
return _GetInstanceConsole(self.cfg.GetClusterInfo(), instance) |
7238 |
|
7239 |
|
7240 |
def _GetInstanceConsole(cluster, instance): |
7241 |
"""Returns console information for an instance.
|
7242 |
|
7243 |
@type cluster: L{objects.Cluster}
|
7244 |
@type instance: L{objects.Instance}
|
7245 |
@rtype: dict
|
7246 |
|
7247 |
"""
|
7248 |
hyper = hypervisor.GetHypervisorClass(instance.hypervisor) |
7249 |
# beparams and hvparams are passed separately, to avoid editing the
|
7250 |
# instance and then saving the defaults in the instance itself.
|
7251 |
hvparams = cluster.FillHV(instance) |
7252 |
beparams = cluster.FillBE(instance) |
7253 |
console = hyper.GetInstanceConsole(instance, hvparams, beparams) |
7254 |
|
7255 |
assert console.instance == instance.name
|
7256 |
assert console.Validate()
|
7257 |
|
7258 |
return console.ToDict()
|
7259 |
|
7260 |
|
7261 |
class LUInstanceReplaceDisks(LogicalUnit): |
7262 |
"""Replace the disks of an instance.
|
7263 |
|
7264 |
"""
|
7265 |
HPATH = "mirrors-replace"
|
7266 |
HTYPE = constants.HTYPE_INSTANCE |
7267 |
REQ_BGL = False
|
7268 |
|
7269 |
def CheckArguments(self): |
7270 |
"""Check arguments.
|
7271 |
|
7272 |
"""
|
7273 |
remote_node = self.op.remote_node
|
7274 |
ialloc = self.op.iallocator
|
7275 |
if self.op.mode == constants.REPLACE_DISK_CHG: |
7276 |
if remote_node is None and ialloc is None: |
7277 |
raise errors.OpPrereqError("When changing the secondary either an" |
7278 |
" iallocator script must be used or the"
|
7279 |
" new node given", errors.ECODE_INVAL)
|
7280 |
else:
|
7281 |
_CheckIAllocatorOrNode(self, "iallocator", "remote_node") |
7282 |
|
7283 |
elif remote_node is not None or ialloc is not None: |
7284 |
# Not replacing the secondary
|
7285 |
raise errors.OpPrereqError("The iallocator and new node options can" |
7286 |
" only be used when changing the"
|
7287 |
" secondary node", errors.ECODE_INVAL)
|
7288 |
|
7289 |
def ExpandNames(self): |
7290 |
self._ExpandAndLockInstance()
|
7291 |
|
7292 |
assert locking.LEVEL_NODE not in self.needed_locks |
7293 |
assert locking.LEVEL_NODE_RES not in self.needed_locks |
7294 |
assert locking.LEVEL_NODEGROUP not in self.needed_locks |
7295 |
|
7296 |
assert self.op.iallocator is None or self.op.remote_node is None, \ |
7297 |
"Conflicting options"
|
7298 |
|
7299 |
if self.op.remote_node is not None: |
7300 |
self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node) |
7301 |
|
7302 |
# Warning: do not remove the locking of the new secondary here
|
7303 |
# unless DRBD8.AddChildren is changed to work in parallel;
|
7304 |
# currently it doesn't since parallel invocations of
|
7305 |
# FindUnusedMinor will conflict
|
7306 |
self.needed_locks[locking.LEVEL_NODE] = [self.op.remote_node] |
7307 |
self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
|
7308 |
else:
|
7309 |
self.needed_locks[locking.LEVEL_NODE] = []
|
7310 |
self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
|
7311 |
|
7312 |
if self.op.iallocator is not None: |
7313 |
# iallocator will select a new node in the same group
|
7314 |
self.needed_locks[locking.LEVEL_NODEGROUP] = []
|
7315 |
self.needed_locks[locking.LEVEL_NODE_ALLOC] = locking.ALL_SET
|
7316 |
|
7317 |
self.needed_locks[locking.LEVEL_NODE_RES] = []
|
7318 |
|
7319 |
self.replacer = TLReplaceDisks(self, self.op.instance_name, self.op.mode, |
7320 |
self.op.iallocator, self.op.remote_node, |
7321 |
self.op.disks, self.op.early_release, |
7322 |
self.op.ignore_ipolicy)
|
7323 |
|
7324 |
self.tasklets = [self.replacer] |
7325 |
|
7326 |
def DeclareLocks(self, level): |
7327 |
if level == locking.LEVEL_NODEGROUP:
|
7328 |
assert self.op.remote_node is None |
7329 |
assert self.op.iallocator is not None |
7330 |
assert not self.needed_locks[locking.LEVEL_NODEGROUP] |
7331 |
|
7332 |
self.share_locks[locking.LEVEL_NODEGROUP] = 1 |
7333 |
# Lock all groups used by instance optimistically; this requires going
|
7334 |
# via the node before it's locked, requiring verification later on
|
7335 |
self.needed_locks[locking.LEVEL_NODEGROUP] = \
|
7336 |
self.cfg.GetInstanceNodeGroups(self.op.instance_name) |
7337 |
|
7338 |
elif level == locking.LEVEL_NODE:
|
7339 |
if self.op.iallocator is not None: |
7340 |
assert self.op.remote_node is None |
7341 |
assert not self.needed_locks[locking.LEVEL_NODE] |
7342 |
assert locking.NAL in self.owned_locks(locking.LEVEL_NODE_ALLOC) |
7343 |
|
7344 |
# Lock member nodes of all locked groups
|
7345 |
self.needed_locks[locking.LEVEL_NODE] = \
|
7346 |
[node_name |
7347 |
for group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP) |
7348 |
for node_name in self.cfg.GetNodeGroup(group_uuid).members] |
7349 |
else:
|
7350 |
assert not self.glm.is_owned(locking.LEVEL_NODE_ALLOC) |
7351 |
|
7352 |
self._LockInstancesNodes()
|
7353 |
|
7354 |
elif level == locking.LEVEL_NODE_RES:
|
7355 |
# Reuse node locks
|
7356 |
self.needed_locks[locking.LEVEL_NODE_RES] = \
|
7357 |
self.needed_locks[locking.LEVEL_NODE]
|
7358 |
|
7359 |
def BuildHooksEnv(self): |
7360 |
"""Build hooks env.
|
7361 |
|
7362 |
This runs on the master, the primary and all the secondaries.
|
7363 |
|
7364 |
"""
|
7365 |
instance = self.replacer.instance
|
7366 |
env = { |
7367 |
"MODE": self.op.mode, |
7368 |
"NEW_SECONDARY": self.op.remote_node, |
7369 |
"OLD_SECONDARY": instance.secondary_nodes[0], |
7370 |
} |
7371 |
env.update(_BuildInstanceHookEnvByObject(self, instance))
|
7372 |
return env
|
7373 |
|
7374 |
def BuildHooksNodes(self): |
7375 |
"""Build hooks nodes.
|
7376 |
|
7377 |
"""
|
7378 |
instance = self.replacer.instance
|
7379 |
nl = [ |
7380 |
self.cfg.GetMasterNode(),
|
7381 |
instance.primary_node, |
7382 |
] |
7383 |
if self.op.remote_node is not None: |
7384 |
nl.append(self.op.remote_node)
|
7385 |
return nl, nl
|
7386 |
|
7387 |
def CheckPrereq(self): |
7388 |
"""Check prerequisites.
|
7389 |
|
7390 |
"""
|
7391 |
assert (self.glm.is_owned(locking.LEVEL_NODEGROUP) or |
7392 |
self.op.iallocator is None) |
7393 |
|
7394 |
# Verify if node group locks are still correct
|
7395 |
owned_groups = self.owned_locks(locking.LEVEL_NODEGROUP)
|
7396 |
if owned_groups:
|
7397 |
_CheckInstanceNodeGroups(self.cfg, self.op.instance_name, owned_groups) |
7398 |
|
7399 |
return LogicalUnit.CheckPrereq(self) |
7400 |
|
7401 |
|
7402 |
class TLReplaceDisks(Tasklet): |
7403 |
"""Replaces disks for an instance.
|
7404 |
|
7405 |
Note: Locking is not within the scope of this class.
|
7406 |
|
7407 |
"""
|
7408 |
def __init__(self, lu, instance_name, mode, iallocator_name, remote_node, |
7409 |
disks, early_release, ignore_ipolicy): |
7410 |
"""Initializes this class.
|
7411 |
|
7412 |
"""
|
7413 |
Tasklet.__init__(self, lu)
|
7414 |
|
7415 |
# Parameters
|
7416 |
self.instance_name = instance_name
|
7417 |
self.mode = mode
|
7418 |
self.iallocator_name = iallocator_name
|
7419 |
self.remote_node = remote_node
|
7420 |
self.disks = disks
|
7421 |
self.early_release = early_release
|
7422 |
self.ignore_ipolicy = ignore_ipolicy
|
7423 |
|
7424 |
# Runtime data
|
7425 |
self.instance = None |
7426 |
self.new_node = None |
7427 |
self.target_node = None |
7428 |
self.other_node = None |
7429 |
self.remote_node_info = None |
7430 |
self.node_secondary_ip = None |
7431 |
|
7432 |
@staticmethod
|
7433 |
def _RunAllocator(lu, iallocator_name, instance_name, relocate_from): |
7434 |
"""Compute a new secondary node using an IAllocator.
|
7435 |
|
7436 |
"""
|
7437 |
req = iallocator.IAReqRelocate(name=instance_name, |
7438 |
relocate_from=list(relocate_from))
|
7439 |
ial = iallocator.IAllocator(lu.cfg, lu.rpc, req) |
7440 |
|
7441 |
ial.Run(iallocator_name) |
7442 |
|
7443 |
if not ial.success: |
7444 |
raise errors.OpPrereqError("Can't compute nodes using iallocator '%s':" |
7445 |
" %s" % (iallocator_name, ial.info),
|
7446 |
errors.ECODE_NORES) |
7447 |
|
7448 |
remote_node_name = ial.result[0]
|
7449 |
|
7450 |
lu.LogInfo("Selected new secondary for instance '%s': %s",
|
7451 |
instance_name, remote_node_name) |
7452 |
|
7453 |
return remote_node_name
|
7454 |
|
7455 |
def _FindFaultyDisks(self, node_name): |
7456 |
"""Wrapper for L{_FindFaultyInstanceDisks}.
|
7457 |
|
7458 |
"""
|
7459 |
return _FindFaultyInstanceDisks(self.cfg, self.rpc, self.instance, |
7460 |
node_name, True)
|
7461 |
|
7462 |
def _CheckDisksActivated(self, instance): |
7463 |
"""Checks if the instance disks are activated.
|
7464 |
|
7465 |
@param instance: The instance to check disks
|
7466 |
@return: True if they are activated, False otherwise
|
7467 |
|
7468 |
"""
|
7469 |
nodes = instance.all_nodes |
7470 |
|
7471 |
for idx, dev in enumerate(instance.disks): |
7472 |
for node in nodes: |
7473 |
self.lu.LogInfo("Checking disk/%d on %s", idx, node) |
7474 |
self.cfg.SetDiskID(dev, node)
|
7475 |
|
7476 |
result = _BlockdevFind(self, node, dev, instance)
|
7477 |
|
7478 |
if result.offline:
|
7479 |
continue
|
7480 |
elif result.fail_msg or not result.payload: |
7481 |
return False |
7482 |
|
7483 |
return True |
7484 |
|
7485 |
def CheckPrereq(self): |
7486 |
"""Check prerequisites.
|
7487 |
|
7488 |
This checks that the instance is in the cluster.
|
7489 |
|
7490 |
"""
|
7491 |
self.instance = instance = self.cfg.GetInstanceInfo(self.instance_name) |
7492 |
assert instance is not None, \ |
7493 |
"Cannot retrieve locked instance %s" % self.instance_name |
7494 |
|
7495 |
if instance.disk_template != constants.DT_DRBD8:
|
7496 |
raise errors.OpPrereqError("Can only run replace disks for DRBD8-based" |
7497 |
" instances", errors.ECODE_INVAL)
|
7498 |
|
7499 |
if len(instance.secondary_nodes) != 1: |
7500 |
raise errors.OpPrereqError("The instance has a strange layout," |
7501 |
" expected one secondary but found %d" %
|
7502 |
len(instance.secondary_nodes),
|
7503 |
errors.ECODE_FAULT) |
7504 |
|
7505 |
instance = self.instance
|
7506 |
secondary_node = instance.secondary_nodes[0]
|
7507 |
|
7508 |
if self.iallocator_name is None: |
7509 |
remote_node = self.remote_node
|
7510 |
else:
|
7511 |
remote_node = self._RunAllocator(self.lu, self.iallocator_name, |
7512 |
instance.name, instance.secondary_nodes) |
7513 |
|
7514 |
if remote_node is None: |
7515 |
self.remote_node_info = None |
7516 |
else:
|
7517 |
assert remote_node in self.lu.owned_locks(locking.LEVEL_NODE), \ |
7518 |
"Remote node '%s' is not locked" % remote_node
|
7519 |
|
7520 |
self.remote_node_info = self.cfg.GetNodeInfo(remote_node) |
7521 |
assert self.remote_node_info is not None, \ |
7522 |
"Cannot retrieve locked node %s" % remote_node
|
7523 |
|
7524 |
if remote_node == self.instance.primary_node: |
7525 |
raise errors.OpPrereqError("The specified node is the primary node of" |
7526 |
" the instance", errors.ECODE_INVAL)
|
7527 |
|
7528 |
if remote_node == secondary_node:
|
7529 |
raise errors.OpPrereqError("The specified node is already the" |
7530 |
" secondary node of the instance",
|
7531 |
errors.ECODE_INVAL) |
7532 |
|
7533 |
if self.disks and self.mode in (constants.REPLACE_DISK_AUTO, |
7534 |
constants.REPLACE_DISK_CHG): |
7535 |
raise errors.OpPrereqError("Cannot specify disks to be replaced", |
7536 |
errors.ECODE_INVAL) |
7537 |
|
7538 |
if self.mode == constants.REPLACE_DISK_AUTO: |
7539 |
if not self._CheckDisksActivated(instance): |
7540 |
raise errors.OpPrereqError("Please run activate-disks on instance %s" |
7541 |
" first" % self.instance_name, |
7542 |
errors.ECODE_STATE) |
7543 |
faulty_primary = self._FindFaultyDisks(instance.primary_node)
|
7544 |
faulty_secondary = self._FindFaultyDisks(secondary_node)
|
7545 |
|
7546 |
if faulty_primary and faulty_secondary: |
7547 |
raise errors.OpPrereqError("Instance %s has faulty disks on more than" |
7548 |
" one node and can not be repaired"
|
7549 |
" automatically" % self.instance_name, |
7550 |
errors.ECODE_STATE) |
7551 |
|
7552 |
if faulty_primary:
|
7553 |
self.disks = faulty_primary
|
7554 |
self.target_node = instance.primary_node
|
7555 |
self.other_node = secondary_node
|
7556 |
check_nodes = [self.target_node, self.other_node] |
7557 |
elif faulty_secondary:
|
7558 |
self.disks = faulty_secondary
|
7559 |
self.target_node = secondary_node
|
7560 |
self.other_node = instance.primary_node
|
7561 |
check_nodes = [self.target_node, self.other_node] |
7562 |
else:
|
7563 |
self.disks = []
|
7564 |
check_nodes = [] |
7565 |
|
7566 |
else:
|
7567 |
# Non-automatic modes
|
7568 |
if self.mode == constants.REPLACE_DISK_PRI: |
7569 |
self.target_node = instance.primary_node
|
7570 |
self.other_node = secondary_node
|
7571 |
check_nodes = [self.target_node, self.other_node] |
7572 |
|
7573 |
elif self.mode == constants.REPLACE_DISK_SEC: |
7574 |
self.target_node = secondary_node
|
7575 |
self.other_node = instance.primary_node
|
7576 |
check_nodes = [self.target_node, self.other_node] |
7577 |
|
7578 |
elif self.mode == constants.REPLACE_DISK_CHG: |
7579 |
self.new_node = remote_node
|
7580 |
self.other_node = instance.primary_node
|
7581 |
self.target_node = secondary_node
|
7582 |
check_nodes = [self.new_node, self.other_node] |
7583 |
|
7584 |
_CheckNodeNotDrained(self.lu, remote_node)
|
7585 |
_CheckNodeVmCapable(self.lu, remote_node)
|
7586 |
|
7587 |
old_node_info = self.cfg.GetNodeInfo(secondary_node)
|
7588 |
assert old_node_info is not None |
7589 |
if old_node_info.offline and not self.early_release: |
7590 |
# doesn't make sense to delay the release
|
7591 |
self.early_release = True |
7592 |
self.lu.LogInfo("Old secondary %s is offline, automatically enabling" |
7593 |
" early-release mode", secondary_node)
|
7594 |
|
7595 |
else:
|
7596 |
raise errors.ProgrammerError("Unhandled disk replace mode (%s)" % |
7597 |
self.mode)
|
7598 |
|
7599 |
# If not specified all disks should be replaced
|
7600 |
if not self.disks: |
7601 |
self.disks = range(len(self.instance.disks)) |
7602 |
|
7603 |
# TODO: This is ugly, but right now we can't distinguish between internal
|
7604 |
# submitted opcode and external one. We should fix that.
|
7605 |
if self.remote_node_info: |
7606 |
# We change the node, lets verify it still meets instance policy
|
7607 |
new_group_info = self.cfg.GetNodeGroup(self.remote_node_info.group) |
7608 |
cluster = self.cfg.GetClusterInfo()
|
7609 |
ipolicy = ganeti.masterd.instance.CalculateGroupIPolicy(cluster, |
7610 |
new_group_info) |
7611 |
_CheckTargetNodeIPolicy(self, ipolicy, instance, self.remote_node_info, |
7612 |
self.cfg, ignore=self.ignore_ipolicy) |
7613 |
|
7614 |
for node in check_nodes: |
7615 |
_CheckNodeOnline(self.lu, node)
|
7616 |
|
7617 |
touched_nodes = frozenset(node_name for node_name in [self.new_node, |
7618 |
self.other_node,
|
7619 |
self.target_node]
|
7620 |
if node_name is not None) |
7621 |
|
7622 |
# Release unneeded node and node resource locks
|
7623 |
_ReleaseLocks(self.lu, locking.LEVEL_NODE, keep=touched_nodes)
|
7624 |
_ReleaseLocks(self.lu, locking.LEVEL_NODE_RES, keep=touched_nodes)
|
7625 |
_ReleaseLocks(self.lu, locking.LEVEL_NODE_ALLOC)
|
7626 |
|
7627 |
# Release any owned node group
|
7628 |
_ReleaseLocks(self.lu, locking.LEVEL_NODEGROUP)
|
7629 |
|
7630 |
# Check whether disks are valid
|
7631 |
for disk_idx in self.disks: |
7632 |
instance.FindDisk(disk_idx) |
7633 |
|
7634 |
# Get secondary node IP addresses
|
7635 |
self.node_secondary_ip = dict((name, node.secondary_ip) for (name, node) |
7636 |
in self.cfg.GetMultiNodeInfo(touched_nodes)) |
7637 |
|
7638 |
def Exec(self, feedback_fn): |
7639 |
"""Execute disk replacement.
|
7640 |
|
7641 |
This dispatches the disk replacement to the appropriate handler.
|
7642 |
|
7643 |
"""
|
7644 |
if __debug__:
|
7645 |
# Verify owned locks before starting operation
|
7646 |
owned_nodes = self.lu.owned_locks(locking.LEVEL_NODE)
|
7647 |
assert set(owned_nodes) == set(self.node_secondary_ip), \ |
7648 |
("Incorrect node locks, owning %s, expected %s" %
|
7649 |
(owned_nodes, self.node_secondary_ip.keys()))
|
7650 |
assert (self.lu.owned_locks(locking.LEVEL_NODE) == |
7651 |
self.lu.owned_locks(locking.LEVEL_NODE_RES))
|
7652 |
assert not self.lu.glm.is_owned(locking.LEVEL_NODE_ALLOC) |
7653 |
|
7654 |
owned_instances = self.lu.owned_locks(locking.LEVEL_INSTANCE)
|
7655 |
assert list(owned_instances) == [self.instance_name], \ |
7656 |
"Instance '%s' not locked" % self.instance_name |
7657 |
|
7658 |
assert not self.lu.glm.is_owned(locking.LEVEL_NODEGROUP), \ |
7659 |
"Should not own any node group lock at this point"
|
7660 |
|
7661 |
if not self.disks: |
7662 |
feedback_fn("No disks need replacement for instance '%s'" %
|
7663 |
self.instance.name)
|
7664 |
return
|
7665 |
|
7666 |
feedback_fn("Replacing disk(s) %s for instance '%s'" %
|
7667 |
(utils.CommaJoin(self.disks), self.instance.name)) |
7668 |
feedback_fn("Current primary node: %s" % self.instance.primary_node) |
7669 |
feedback_fn("Current seconary node: %s" %
|
7670 |
utils.CommaJoin(self.instance.secondary_nodes))
|
7671 |
|
7672 |
activate_disks = (self.instance.admin_state != constants.ADMINST_UP)
|
7673 |
|
7674 |
# Activate the instance disks if we're replacing them on a down instance
|
7675 |
if activate_disks:
|
7676 |
_StartInstanceDisks(self.lu, self.instance, True) |
7677 |
|
7678 |
try:
|
7679 |
# Should we replace the secondary node?
|
7680 |
if self.new_node is not None: |
7681 |
fn = self._ExecDrbd8Secondary
|
7682 |
else:
|
7683 |
fn = self._ExecDrbd8DiskOnly
|
7684 |
|
7685 |
result = fn(feedback_fn) |
7686 |
finally:
|
7687 |
# Deactivate the instance disks if we're replacing them on a
|
7688 |
# down instance
|
7689 |
if activate_disks:
|
7690 |
_SafeShutdownInstanceDisks(self.lu, self.instance) |
7691 |
|
7692 |
assert not self.lu.owned_locks(locking.LEVEL_NODE) |
7693 |
|
7694 |
if __debug__:
|
7695 |
# Verify owned locks
|
7696 |
owned_nodes = self.lu.owned_locks(locking.LEVEL_NODE_RES)
|
7697 |
nodes = frozenset(self.node_secondary_ip) |
7698 |
assert ((self.early_release and not owned_nodes) or |
7699 |
(not self.early_release and not (set(owned_nodes) - nodes))), \ |
7700 |
("Not owning the correct locks, early_release=%s, owned=%r,"
|
7701 |
" nodes=%r" % (self.early_release, owned_nodes, nodes)) |
7702 |
|
7703 |
return result
|
7704 |
|
7705 |
def _CheckVolumeGroup(self, nodes): |
7706 |
self.lu.LogInfo("Checking volume groups") |
7707 |
|
7708 |
vgname = self.cfg.GetVGName()
|
7709 |
|
7710 |
# Make sure volume group exists on all involved nodes
|
7711 |
results = self.rpc.call_vg_list(nodes)
|
7712 |
if not results: |
7713 |
raise errors.OpExecError("Can't list volume groups on the nodes") |
7714 |
|
7715 |
for node in nodes: |
7716 |
res = results[node] |
7717 |
res.Raise("Error checking node %s" % node)
|
7718 |
if vgname not in res.payload: |
7719 |
raise errors.OpExecError("Volume group '%s' not found on node %s" % |
7720 |
(vgname, node)) |
7721 |
|
7722 |
def _CheckDisksExistence(self, nodes): |
7723 |
# Check disk existence
|
7724 |
for idx, dev in enumerate(self.instance.disks): |
7725 |
if idx not in self.disks: |
7726 |
continue
|
7727 |
|
7728 |
for node in nodes: |
7729 |
self.lu.LogInfo("Checking disk/%d on %s", idx, node) |
7730 |
self.cfg.SetDiskID(dev, node)
|
7731 |
|
7732 |
result = _BlockdevFind(self, node, dev, self.instance) |
7733 |
|
7734 |
msg = result.fail_msg |
7735 |
if msg or not result.payload: |
7736 |
if not msg: |
7737 |
msg = "disk not found"
|
7738 |
raise errors.OpExecError("Can't find disk/%d on node %s: %s" % |
7739 |
(idx, node, msg)) |
7740 |
|
7741 |
def _CheckDisksConsistency(self, node_name, on_primary, ldisk): |
7742 |
for idx, dev in enumerate(self.instance.disks): |
7743 |
if idx not in self.disks: |
7744 |
continue
|
7745 |
|
7746 |
self.lu.LogInfo("Checking disk/%d consistency on node %s" % |
7747 |
(idx, node_name)) |
7748 |
|
7749 |
if not _CheckDiskConsistency(self.lu, self.instance, dev, node_name, |
7750 |
on_primary, ldisk=ldisk): |
7751 |
raise errors.OpExecError("Node %s has degraded storage, unsafe to" |
7752 |
" replace disks for instance %s" %
|
7753 |
(node_name, self.instance.name))
|
7754 |
|
7755 |
def _CreateNewStorage(self, node_name): |
7756 |
"""Create new storage on the primary or secondary node.
|
7757 |
|
7758 |
This is only used for same-node replaces, not for changing the
|
7759 |
secondary node, hence we don't want to modify the existing disk.
|
7760 |
|
7761 |
"""
|
7762 |
iv_names = {} |
7763 |
|
7764 |
disks = _AnnotateDiskParams(self.instance, self.instance.disks, self.cfg) |
7765 |
for idx, dev in enumerate(disks): |
7766 |
if idx not in self.disks: |
7767 |
continue
|
7768 |
|
7769 |
self.lu.LogInfo("Adding storage on %s for disk/%d", node_name, idx) |
7770 |
|
7771 |
self.cfg.SetDiskID(dev, node_name)
|
7772 |
|
7773 |
lv_names = [".disk%d_%s" % (idx, suffix) for suffix in ["data", "meta"]] |
7774 |
names = _GenerateUniqueNames(self.lu, lv_names)
|
7775 |
|
7776 |
(data_disk, meta_disk) = dev.children |
7777 |
vg_data = data_disk.logical_id[0]
|
7778 |
lv_data = objects.Disk(dev_type=constants.LD_LV, size=dev.size, |
7779 |
logical_id=(vg_data, names[0]),
|
7780 |
params=data_disk.params) |
7781 |
vg_meta = meta_disk.logical_id[0]
|
7782 |
lv_meta = objects.Disk(dev_type=constants.LD_LV, |
7783 |
size=constants.DRBD_META_SIZE, |
7784 |
logical_id=(vg_meta, names[1]),
|
7785 |
params=meta_disk.params) |
7786 |
|
7787 |
new_lvs = [lv_data, lv_meta] |
7788 |
old_lvs = [child.Copy() for child in dev.children] |
7789 |
iv_names[dev.iv_name] = (dev, old_lvs, new_lvs) |
7790 |
excl_stor = _IsExclusiveStorageEnabledNodeName(self.lu.cfg, node_name)
|
7791 |
|
7792 |
# we pass force_create=True to force the LVM creation
|
7793 |
for new_lv in new_lvs: |
7794 |
_CreateBlockDevInner(self.lu, node_name, self.instance, new_lv, True, |
7795 |
_GetInstanceInfoText(self.instance), False, |
7796 |
excl_stor) |
7797 |
|
7798 |
return iv_names
|
7799 |
|
7800 |
def _CheckDevices(self, node_name, iv_names): |
7801 |
for name, (dev, _, _) in iv_names.iteritems(): |
7802 |
self.cfg.SetDiskID(dev, node_name)
|
7803 |
|
7804 |
result = _BlockdevFind(self, node_name, dev, self.instance) |
7805 |
|
7806 |
msg = result.fail_msg |
7807 |
if msg or not result.payload: |
7808 |
if not msg: |
7809 |
msg = "disk not found"
|
7810 |
raise errors.OpExecError("Can't find DRBD device %s: %s" % |
7811 |
(name, msg)) |
7812 |
|
7813 |
if result.payload.is_degraded:
|
7814 |
raise errors.OpExecError("DRBD device %s is degraded!" % name) |
7815 |
|
7816 |
def _RemoveOldStorage(self, node_name, iv_names): |
7817 |
for name, (_, old_lvs, _) in iv_names.iteritems(): |
7818 |
self.lu.LogInfo("Remove logical volumes for %s", name) |
7819 |
|
7820 |
for lv in old_lvs: |
7821 |
self.cfg.SetDiskID(lv, node_name)
|
7822 |
|
7823 |
msg = self.rpc.call_blockdev_remove(node_name, lv).fail_msg
|
7824 |
if msg:
|
7825 |
self.lu.LogWarning("Can't remove old LV: %s", msg, |
7826 |
hint="remove unused LVs manually")
|
7827 |
|
7828 |
def _ExecDrbd8DiskOnly(self, feedback_fn): # pylint: disable=W0613 |
7829 |
"""Replace a disk on the primary or secondary for DRBD 8.
|
7830 |
|
7831 |
The algorithm for replace is quite complicated:
|
7832 |
|
7833 |
1. for each disk to be replaced:
|
7834 |
|
7835 |
1. create new LVs on the target node with unique names
|
7836 |
1. detach old LVs from the drbd device
|
7837 |
1. rename old LVs to name_replaced.<time_t>
|
7838 |
1. rename new LVs to old LVs
|
7839 |
1. attach the new LVs (with the old names now) to the drbd device
|
7840 |
|
7841 |
1. wait for sync across all devices
|
7842 |
|
7843 |
1. for each modified disk:
|
7844 |
|
7845 |
1. remove old LVs (which have the name name_replaces.<time_t>)
|
7846 |
|
7847 |
Failures are not very well handled.
|
7848 |
|
7849 |
"""
|
7850 |
steps_total = 6
|
7851 |
|
7852 |
# Step: check device activation
|
7853 |
self.lu.LogStep(1, steps_total, "Check device existence") |
7854 |
self._CheckDisksExistence([self.other_node, self.target_node]) |
7855 |
self._CheckVolumeGroup([self.target_node, self.other_node]) |
7856 |
|
7857 |
# Step: check other node consistency
|
7858 |
self.lu.LogStep(2, steps_total, "Check peer consistency") |
7859 |
self._CheckDisksConsistency(self.other_node, |
7860 |
self.other_node == self.instance.primary_node, |
7861 |
False)
|
7862 |
|
7863 |
# Step: create new storage
|
7864 |
self.lu.LogStep(3, steps_total, "Allocate new storage") |
7865 |
iv_names = self._CreateNewStorage(self.target_node) |
7866 |
|
7867 |
# Step: for each lv, detach+rename*2+attach
|
7868 |
self.lu.LogStep(4, steps_total, "Changing drbd configuration") |
7869 |
for dev, old_lvs, new_lvs in iv_names.itervalues(): |
7870 |
self.lu.LogInfo("Detaching %s drbd from local storage", dev.iv_name) |
7871 |
|
7872 |
result = self.rpc.call_blockdev_removechildren(self.target_node, dev, |
7873 |
old_lvs) |
7874 |
result.Raise("Can't detach drbd from local storage on node"
|
7875 |
" %s for device %s" % (self.target_node, dev.iv_name)) |
7876 |
#dev.children = []
|
7877 |
#cfg.Update(instance)
|
7878 |
|
7879 |
# ok, we created the new LVs, so now we know we have the needed
|
7880 |
# storage; as such, we proceed on the target node to rename
|
7881 |
# old_lv to _old, and new_lv to old_lv; note that we rename LVs
|
7882 |
# using the assumption that logical_id == physical_id (which in
|
7883 |
# turn is the unique_id on that node)
|
7884 |
|
7885 |
# FIXME(iustin): use a better name for the replaced LVs
|
7886 |
temp_suffix = int(time.time())
|
7887 |
ren_fn = lambda d, suff: (d.physical_id[0], |
7888 |
d.physical_id[1] + "_replaced-%s" % suff) |
7889 |
|
7890 |
# Build the rename list based on what LVs exist on the node
|
7891 |
rename_old_to_new = [] |
7892 |
for to_ren in old_lvs: |
7893 |
result = self.rpc.call_blockdev_find(self.target_node, to_ren) |
7894 |
if not result.fail_msg and result.payload: |
7895 |
# device exists
|
7896 |
rename_old_to_new.append((to_ren, ren_fn(to_ren, temp_suffix))) |
7897 |
|
7898 |
self.lu.LogInfo("Renaming the old LVs on the target node") |
7899 |
result = self.rpc.call_blockdev_rename(self.target_node, |
7900 |
rename_old_to_new) |
7901 |
result.Raise("Can't rename old LVs on node %s" % self.target_node) |
7902 |
|
7903 |
# Now we rename the new LVs to the old LVs
|
7904 |
self.lu.LogInfo("Renaming the new LVs on the target node") |
7905 |
rename_new_to_old = [(new, old.physical_id) |
7906 |
for old, new in zip(old_lvs, new_lvs)] |
7907 |
result = self.rpc.call_blockdev_rename(self.target_node, |
7908 |
rename_new_to_old) |
7909 |
result.Raise("Can't rename new LVs on node %s" % self.target_node) |
7910 |
|
7911 |
# Intermediate steps of in memory modifications
|
7912 |
for old, new in zip(old_lvs, new_lvs): |
7913 |
new.logical_id = old.logical_id |
7914 |
self.cfg.SetDiskID(new, self.target_node) |
7915 |
|
7916 |
# We need to modify old_lvs so that removal later removes the
|
7917 |
# right LVs, not the newly added ones; note that old_lvs is a
|
7918 |
# copy here
|
7919 |
for disk in old_lvs: |
7920 |
disk.logical_id = ren_fn(disk, temp_suffix) |
7921 |
self.cfg.SetDiskID(disk, self.target_node) |
7922 |
|
7923 |
# Now that the new lvs have the old name, we can add them to the device
|
7924 |
self.lu.LogInfo("Adding new mirror component on %s", self.target_node) |
7925 |
result = self.rpc.call_blockdev_addchildren(self.target_node, |
7926 |
(dev, self.instance), new_lvs)
|
7927 |
msg = result.fail_msg |
7928 |
if msg:
|
7929 |
for new_lv in new_lvs: |
7930 |
msg2 = self.rpc.call_blockdev_remove(self.target_node, |
7931 |
new_lv).fail_msg |
7932 |
if msg2:
|
7933 |
self.lu.LogWarning("Can't rollback device %s: %s", dev, msg2, |
7934 |
hint=("cleanup manually the unused logical"
|
7935 |
"volumes"))
|
7936 |
raise errors.OpExecError("Can't add local storage to drbd: %s" % msg) |
7937 |
|
7938 |
cstep = itertools.count(5)
|
7939 |
|
7940 |
if self.early_release: |
7941 |
self.lu.LogStep(cstep.next(), steps_total, "Removing old storage") |
7942 |
self._RemoveOldStorage(self.target_node, iv_names) |
7943 |
# TODO: Check if releasing locks early still makes sense
|
7944 |
_ReleaseLocks(self.lu, locking.LEVEL_NODE_RES)
|
7945 |
else:
|
7946 |
# Release all resource locks except those used by the instance
|
7947 |
_ReleaseLocks(self.lu, locking.LEVEL_NODE_RES,
|
7948 |
keep=self.node_secondary_ip.keys())
|
7949 |
|
7950 |
# Release all node locks while waiting for sync
|
7951 |
_ReleaseLocks(self.lu, locking.LEVEL_NODE)
|
7952 |
|
7953 |
# TODO: Can the instance lock be downgraded here? Take the optional disk
|
7954 |
# shutdown in the caller into consideration.
|
7955 |
|
7956 |
# Wait for sync
|
7957 |
# This can fail as the old devices are degraded and _WaitForSync
|
7958 |
# does a combined result over all disks, so we don't check its return value
|
7959 |
self.lu.LogStep(cstep.next(), steps_total, "Sync devices") |
7960 |
_WaitForSync(self.lu, self.instance) |
7961 |
|
7962 |
# Check all devices manually
|
7963 |
self._CheckDevices(self.instance.primary_node, iv_names) |
7964 |
|
7965 |
# Step: remove old storage
|
7966 |
if not self.early_release: |
7967 |
self.lu.LogStep(cstep.next(), steps_total, "Removing old storage") |
7968 |
self._RemoveOldStorage(self.target_node, iv_names) |
7969 |
|
7970 |
def _ExecDrbd8Secondary(self, feedback_fn): |
7971 |
"""Replace the secondary node for DRBD 8.
|
7972 |
|
7973 |
The algorithm for replace is quite complicated:
|
7974 |
- for all disks of the instance:
|
7975 |
- create new LVs on the new node with same names
|
7976 |
- shutdown the drbd device on the old secondary
|
7977 |
- disconnect the drbd network on the primary
|
7978 |
- create the drbd device on the new secondary
|
7979 |
- network attach the drbd on the primary, using an artifice:
|
7980 |
the drbd code for Attach() will connect to the network if it
|
7981 |
finds a device which is connected to the good local disks but
|
7982 |
not network enabled
|
7983 |
- wait for sync across all devices
|
7984 |
- remove all disks from the old secondary
|
7985 |
|
7986 |
Failures are not very well handled.
|
7987 |
|
7988 |
"""
|
7989 |
steps_total = 6
|
7990 |
|
7991 |
pnode = self.instance.primary_node
|
7992 |
|
7993 |
# Step: check device activation
|
7994 |
self.lu.LogStep(1, steps_total, "Check device existence") |
7995 |
self._CheckDisksExistence([self.instance.primary_node]) |
7996 |
self._CheckVolumeGroup([self.instance.primary_node]) |
7997 |
|
7998 |
# Step: check other node consistency
|
7999 |
self.lu.LogStep(2, steps_total, "Check peer consistency") |
8000 |
self._CheckDisksConsistency(self.instance.primary_node, True, True) |
8001 |
|
8002 |
# Step: create new storage
|
8003 |
self.lu.LogStep(3, steps_total, "Allocate new storage") |
8004 |
disks = _AnnotateDiskParams(self.instance, self.instance.disks, self.cfg) |
8005 |
excl_stor = _IsExclusiveStorageEnabledNodeName(self.lu.cfg, self.new_node) |
8006 |
for idx, dev in enumerate(disks): |
8007 |
self.lu.LogInfo("Adding new local storage on %s for disk/%d" % |
8008 |
(self.new_node, idx))
|
8009 |
# we pass force_create=True to force LVM creation
|
8010 |
for new_lv in dev.children: |
8011 |
_CreateBlockDevInner(self.lu, self.new_node, self.instance, new_lv, |
8012 |
True, _GetInstanceInfoText(self.instance), False, |
8013 |
excl_stor) |
8014 |
|
8015 |
# Step 4: dbrd minors and drbd setups changes
|
8016 |
# after this, we must manually remove the drbd minors on both the
|
8017 |
# error and the success paths
|
8018 |
self.lu.LogStep(4, steps_total, "Changing drbd configuration") |
8019 |
minors = self.cfg.AllocateDRBDMinor([self.new_node |
8020 |
for dev in self.instance.disks], |
8021 |
self.instance.name)
|
8022 |
logging.debug("Allocated minors %r", minors)
|
8023 |
|
8024 |
iv_names = {} |
8025 |
for idx, (dev, new_minor) in enumerate(zip(self.instance.disks, minors)): |
8026 |
self.lu.LogInfo("activating a new drbd on %s for disk/%d" % |
8027 |
(self.new_node, idx))
|
8028 |
# create new devices on new_node; note that we create two IDs:
|
8029 |
# one without port, so the drbd will be activated without
|
8030 |
# networking information on the new node at this stage, and one
|
8031 |
# with network, for the latter activation in step 4
|
8032 |
(o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id |
8033 |
if self.instance.primary_node == o_node1: |
8034 |
p_minor = o_minor1 |
8035 |
else:
|
8036 |
assert self.instance.primary_node == o_node2, "Three-node instance?" |
8037 |
p_minor = o_minor2 |
8038 |
|
8039 |
new_alone_id = (self.instance.primary_node, self.new_node, None, |
8040 |
p_minor, new_minor, o_secret) |
8041 |
new_net_id = (self.instance.primary_node, self.new_node, o_port, |
8042 |
p_minor, new_minor, o_secret) |
8043 |
|
8044 |
iv_names[idx] = (dev, dev.children, new_net_id) |
8045 |
logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
|
8046 |
new_net_id) |
8047 |
new_drbd = objects.Disk(dev_type=constants.LD_DRBD8, |
8048 |
logical_id=new_alone_id, |
8049 |
children=dev.children, |
8050 |
size=dev.size, |
8051 |
params={}) |
8052 |
(anno_new_drbd,) = _AnnotateDiskParams(self.instance, [new_drbd],
|
8053 |
self.cfg)
|
8054 |
try:
|
8055 |
_CreateSingleBlockDev(self.lu, self.new_node, self.instance, |
8056 |
anno_new_drbd, |
8057 |
_GetInstanceInfoText(self.instance), False, |
8058 |
excl_stor) |
8059 |
except errors.GenericError:
|
8060 |
self.cfg.ReleaseDRBDMinors(self.instance.name) |
8061 |
raise
|
8062 |
|
8063 |
# We have new devices, shutdown the drbd on the old secondary
|
8064 |
for idx, dev in enumerate(self.instance.disks): |
8065 |
self.lu.LogInfo("Shutting down drbd for disk/%d on old node", idx) |
8066 |
self.cfg.SetDiskID(dev, self.target_node) |
8067 |
msg = self.rpc.call_blockdev_shutdown(self.target_node, |
8068 |
(dev, self.instance)).fail_msg
|
8069 |
if msg:
|
8070 |
self.lu.LogWarning("Failed to shutdown drbd for disk/%d on old" |
8071 |
"node: %s" % (idx, msg),
|
8072 |
hint=("Please cleanup this device manually as"
|
8073 |
" soon as possible"))
|
8074 |
|
8075 |
self.lu.LogInfo("Detaching primary drbds from the network (=> standalone)") |
8076 |
result = self.rpc.call_drbd_disconnect_net([pnode], self.node_secondary_ip, |
8077 |
self.instance.disks)[pnode]
|
8078 |
|
8079 |
msg = result.fail_msg |
8080 |
if msg:
|
8081 |
# detaches didn't succeed (unlikely)
|
8082 |
self.cfg.ReleaseDRBDMinors(self.instance.name) |
8083 |
raise errors.OpExecError("Can't detach the disks from the network on" |
8084 |
" old node: %s" % (msg,))
|
8085 |
|
8086 |
# if we managed to detach at least one, we update all the disks of
|
8087 |
# the instance to point to the new secondary
|
8088 |
self.lu.LogInfo("Updating instance configuration") |
8089 |
for dev, _, new_logical_id in iv_names.itervalues(): |
8090 |
dev.logical_id = new_logical_id |
8091 |
self.cfg.SetDiskID(dev, self.instance.primary_node) |
8092 |
|
8093 |
self.cfg.Update(self.instance, feedback_fn) |
8094 |
|
8095 |
# Release all node locks (the configuration has been updated)
|
8096 |
_ReleaseLocks(self.lu, locking.LEVEL_NODE)
|
8097 |
|
8098 |
# and now perform the drbd attach
|
8099 |
self.lu.LogInfo("Attaching primary drbds to new secondary" |
8100 |
" (standalone => connected)")
|
8101 |
result = self.rpc.call_drbd_attach_net([self.instance.primary_node, |
8102 |
self.new_node],
|
8103 |
self.node_secondary_ip,
|
8104 |
(self.instance.disks, self.instance), |
8105 |
self.instance.name,
|
8106 |
False)
|
8107 |
for to_node, to_result in result.items(): |
8108 |
msg = to_result.fail_msg |
8109 |
if msg:
|
8110 |
self.lu.LogWarning("Can't attach drbd disks on node %s: %s", |
8111 |
to_node, msg, |
8112 |
hint=("please do a gnt-instance info to see the"
|
8113 |
" status of disks"))
|
8114 |
|
8115 |
cstep = itertools.count(5)
|
8116 |
|
8117 |
if self.early_release: |
8118 |
self.lu.LogStep(cstep.next(), steps_total, "Removing old storage") |
8119 |
self._RemoveOldStorage(self.target_node, iv_names) |
8120 |
# TODO: Check if releasing locks early still makes sense
|
8121 |
_ReleaseLocks(self.lu, locking.LEVEL_NODE_RES)
|
8122 |
else:
|
8123 |
# Release all resource locks except those used by the instance
|
8124 |
_ReleaseLocks(self.lu, locking.LEVEL_NODE_RES,
|
8125 |
keep=self.node_secondary_ip.keys())
|
8126 |
|
8127 |
# TODO: Can the instance lock be downgraded here? Take the optional disk
|
8128 |
# shutdown in the caller into consideration.
|
8129 |
|
8130 |
# Wait for sync
|
8131 |
# This can fail as the old devices are degraded and _WaitForSync
|
8132 |
# does a combined result over all disks, so we don't check its return value
|
8133 |
self.lu.LogStep(cstep.next(), steps_total, "Sync devices") |
8134 |
_WaitForSync(self.lu, self.instance) |
8135 |
|
8136 |
# Check all devices manually
|
8137 |
self._CheckDevices(self.instance.primary_node, iv_names) |
8138 |
|
8139 |
# Step: remove old storage
|
8140 |
if not self.early_release: |
8141 |
self.lu.LogStep(cstep.next(), steps_total, "Removing old storage") |
8142 |
self._RemoveOldStorage(self.target_node, iv_names) |
8143 |
|
8144 |
|
8145 |
class LURepairNodeStorage(NoHooksLU): |
8146 |
"""Repairs the volume group on a node.
|
8147 |
|
8148 |
"""
|
8149 |
REQ_BGL = False
|
8150 |
|
8151 |
def CheckArguments(self): |
8152 |
self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name) |
8153 |
|
8154 |
storage_type = self.op.storage_type
|
8155 |
|
8156 |
if (constants.SO_FIX_CONSISTENCY not in |
8157 |
constants.VALID_STORAGE_OPERATIONS.get(storage_type, [])): |
8158 |
raise errors.OpPrereqError("Storage units of type '%s' can not be" |
8159 |
" repaired" % storage_type,
|
8160 |
errors.ECODE_INVAL) |
8161 |
|
8162 |
def ExpandNames(self): |
8163 |
self.needed_locks = {
|
8164 |
locking.LEVEL_NODE: [self.op.node_name],
|
8165 |
} |
8166 |
|
8167 |
def _CheckFaultyDisks(self, instance, node_name): |
8168 |
"""Ensure faulty disks abort the opcode or at least warn."""
|
8169 |
try:
|
8170 |
if _FindFaultyInstanceDisks(self.cfg, self.rpc, instance, |
8171 |
node_name, True):
|
8172 |
raise errors.OpPrereqError("Instance '%s' has faulty disks on" |
8173 |
" node '%s'" % (instance.name, node_name),
|
8174 |
errors.ECODE_STATE) |
8175 |
except errors.OpPrereqError, err:
|
8176 |
if self.op.ignore_consistency: |
8177 |
self.LogWarning(str(err.args[0])) |
8178 |
else:
|
8179 |
raise
|
8180 |
|
8181 |
def CheckPrereq(self): |
8182 |
"""Check prerequisites.
|
8183 |
|
8184 |
"""
|
8185 |
# Check whether any instance on this node has faulty disks
|
8186 |
for inst in _GetNodeInstances(self.cfg, self.op.node_name): |
8187 |
if inst.admin_state != constants.ADMINST_UP:
|
8188 |
continue
|
8189 |
check_nodes = set(inst.all_nodes)
|
8190 |
check_nodes.discard(self.op.node_name)
|
8191 |
for inst_node_name in check_nodes: |
8192 |
self._CheckFaultyDisks(inst, inst_node_name)
|
8193 |
|
8194 |
def Exec(self, feedback_fn): |
8195 |
feedback_fn("Repairing storage unit '%s' on %s ..." %
|
8196 |
(self.op.name, self.op.node_name)) |
8197 |
|
8198 |
st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type) |
8199 |
result = self.rpc.call_storage_execute(self.op.node_name, |
8200 |
self.op.storage_type, st_args,
|
8201 |
self.op.name,
|
8202 |
constants.SO_FIX_CONSISTENCY) |
8203 |
result.Raise("Failed to repair storage unit '%s' on %s" %
|
8204 |
(self.op.name, self.op.node_name)) |
8205 |
|
8206 |
|
8207 |
class LUNodeEvacuate(NoHooksLU): |
8208 |
"""Evacuates instances off a list of nodes.
|
8209 |
|
8210 |
"""
|
8211 |
REQ_BGL = False
|
8212 |
|
8213 |
_MODE2IALLOCATOR = { |
8214 |
constants.NODE_EVAC_PRI: constants.IALLOCATOR_NEVAC_PRI, |
8215 |
constants.NODE_EVAC_SEC: constants.IALLOCATOR_NEVAC_SEC, |
8216 |
constants.NODE_EVAC_ALL: constants.IALLOCATOR_NEVAC_ALL, |
8217 |
} |
8218 |
assert frozenset(_MODE2IALLOCATOR.keys()) == constants.NODE_EVAC_MODES |
8219 |
assert (frozenset(_MODE2IALLOCATOR.values()) == |
8220 |
constants.IALLOCATOR_NEVAC_MODES) |
8221 |
|
8222 |
def CheckArguments(self): |
8223 |
_CheckIAllocatorOrNode(self, "iallocator", "remote_node") |
8224 |
|
8225 |
def ExpandNames(self): |
8226 |
self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name) |
8227 |
|
8228 |
if self.op.remote_node is not None: |
8229 |
self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node) |
8230 |
assert self.op.remote_node |
8231 |
|
8232 |
if self.op.remote_node == self.op.node_name: |
8233 |
raise errors.OpPrereqError("Can not use evacuated node as a new" |
8234 |
" secondary node", errors.ECODE_INVAL)
|
8235 |
|
8236 |
if self.op.mode != constants.NODE_EVAC_SEC: |
8237 |
raise errors.OpPrereqError("Without the use of an iallocator only" |
8238 |
" secondary instances can be evacuated",
|
8239 |
errors.ECODE_INVAL) |
8240 |
|
8241 |
# Declare locks
|
8242 |
self.share_locks = _ShareAll()
|
8243 |
self.needed_locks = {
|
8244 |
locking.LEVEL_INSTANCE: [], |
8245 |
locking.LEVEL_NODEGROUP: [], |
8246 |
locking.LEVEL_NODE: [], |
8247 |
} |
8248 |
|
8249 |
# Determine nodes (via group) optimistically, needs verification once locks
|
8250 |
# have been acquired
|
8251 |
self.lock_nodes = self._DetermineNodes() |
8252 |
|
8253 |
def _DetermineNodes(self): |
8254 |
"""Gets the list of nodes to operate on.
|
8255 |
|
8256 |
"""
|
8257 |
if self.op.remote_node is None: |
8258 |
# Iallocator will choose any node(s) in the same group
|
8259 |
group_nodes = self.cfg.GetNodeGroupMembersByNodes([self.op.node_name]) |
8260 |
else:
|
8261 |
group_nodes = frozenset([self.op.remote_node]) |
8262 |
|
8263 |
# Determine nodes to be locked
|
8264 |
return set([self.op.node_name]) | group_nodes |
8265 |
|
8266 |
def _DetermineInstances(self): |
8267 |
"""Builds list of instances to operate on.
|
8268 |
|
8269 |
"""
|
8270 |
assert self.op.mode in constants.NODE_EVAC_MODES |
8271 |
|
8272 |
if self.op.mode == constants.NODE_EVAC_PRI: |
8273 |
# Primary instances only
|
8274 |
inst_fn = _GetNodePrimaryInstances |
8275 |
assert self.op.remote_node is None, \ |
8276 |
"Evacuating primary instances requires iallocator"
|
8277 |
elif self.op.mode == constants.NODE_EVAC_SEC: |
8278 |
# Secondary instances only
|
8279 |
inst_fn = _GetNodeSecondaryInstances |
8280 |
else:
|
8281 |
# All instances
|
8282 |
assert self.op.mode == constants.NODE_EVAC_ALL |
8283 |
inst_fn = _GetNodeInstances |
8284 |
# TODO: In 2.6, change the iallocator interface to take an evacuation mode
|
8285 |
# per instance
|
8286 |
raise errors.OpPrereqError("Due to an issue with the iallocator" |
8287 |
" interface it is not possible to evacuate"
|
8288 |
" all instances at once; specify explicitly"
|
8289 |
" whether to evacuate primary or secondary"
|
8290 |
" instances",
|
8291 |
errors.ECODE_INVAL) |
8292 |
|
8293 |
return inst_fn(self.cfg, self.op.node_name) |
8294 |
|
8295 |
def DeclareLocks(self, level): |
8296 |
if level == locking.LEVEL_INSTANCE:
|
8297 |
# Lock instances optimistically, needs verification once node and group
|
8298 |
# locks have been acquired
|
8299 |
self.needed_locks[locking.LEVEL_INSTANCE] = \
|
8300 |
set(i.name for i in self._DetermineInstances()) |
8301 |
|
8302 |
elif level == locking.LEVEL_NODEGROUP:
|
8303 |
# Lock node groups for all potential target nodes optimistically, needs
|
8304 |
# verification once nodes have been acquired
|
8305 |
self.needed_locks[locking.LEVEL_NODEGROUP] = \
|
8306 |
self.cfg.GetNodeGroupsFromNodes(self.lock_nodes) |
8307 |
|
8308 |
elif level == locking.LEVEL_NODE:
|
8309 |
self.needed_locks[locking.LEVEL_NODE] = self.lock_nodes |
8310 |
|
8311 |
def CheckPrereq(self): |
8312 |
# Verify locks
|
8313 |
owned_instances = self.owned_locks(locking.LEVEL_INSTANCE)
|
8314 |
owned_nodes = self.owned_locks(locking.LEVEL_NODE)
|
8315 |
owned_groups = self.owned_locks(locking.LEVEL_NODEGROUP)
|
8316 |
|
8317 |
need_nodes = self._DetermineNodes()
|
8318 |
|
8319 |
if not owned_nodes.issuperset(need_nodes): |
8320 |
raise errors.OpPrereqError("Nodes in same group as '%s' changed since" |
8321 |
" locks were acquired, current nodes are"
|
8322 |
" are '%s', used to be '%s'; retry the"
|
8323 |
" operation" %
|
8324 |
(self.op.node_name,
|
8325 |
utils.CommaJoin(need_nodes), |
8326 |
utils.CommaJoin(owned_nodes)), |
8327 |
errors.ECODE_STATE) |
8328 |
|
8329 |
wanted_groups = self.cfg.GetNodeGroupsFromNodes(owned_nodes)
|
8330 |
if owned_groups != wanted_groups:
|
8331 |
raise errors.OpExecError("Node groups changed since locks were acquired," |
8332 |
" current groups are '%s', used to be '%s';"
|
8333 |
" retry the operation" %
|
8334 |
(utils.CommaJoin(wanted_groups), |
8335 |
utils.CommaJoin(owned_groups))) |
8336 |
|
8337 |
# Determine affected instances
|
8338 |
self.instances = self._DetermineInstances() |
8339 |
self.instance_names = [i.name for i in self.instances] |
8340 |
|
8341 |
if set(self.instance_names) != owned_instances: |
8342 |
raise errors.OpExecError("Instances on node '%s' changed since locks" |
8343 |
" were acquired, current instances are '%s',"
|
8344 |
" used to be '%s'; retry the operation" %
|
8345 |
(self.op.node_name,
|
8346 |
utils.CommaJoin(self.instance_names),
|
8347 |
utils.CommaJoin(owned_instances))) |
8348 |
|
8349 |
if self.instance_names: |
8350 |
self.LogInfo("Evacuating instances from node '%s': %s", |
8351 |
self.op.node_name,
|
8352 |
utils.CommaJoin(utils.NiceSort(self.instance_names)))
|
8353 |
else:
|
8354 |
self.LogInfo("No instances to evacuate from node '%s'", |
8355 |
self.op.node_name)
|
8356 |
|
8357 |
if self.op.remote_node is not None: |
8358 |
for i in self.instances: |
8359 |
if i.primary_node == self.op.remote_node: |
8360 |
raise errors.OpPrereqError("Node %s is the primary node of" |
8361 |
" instance %s, cannot use it as"
|
8362 |
" secondary" %
|
8363 |
(self.op.remote_node, i.name),
|
8364 |
errors.ECODE_INVAL) |
8365 |
|
8366 |
def Exec(self, feedback_fn): |
8367 |
assert (self.op.iallocator is not None) ^ (self.op.remote_node is not None) |
8368 |
|
8369 |
if not self.instance_names: |
8370 |
# No instances to evacuate
|
8371 |
jobs = [] |
8372 |
|
8373 |
elif self.op.iallocator is not None: |
8374 |
# TODO: Implement relocation to other group
|
8375 |
evac_mode = self._MODE2IALLOCATOR[self.op.mode] |
8376 |
req = iallocator.IAReqNodeEvac(evac_mode=evac_mode, |
8377 |
instances=list(self.instance_names)) |
8378 |
ial = iallocator.IAllocator(self.cfg, self.rpc, req) |
8379 |
|
8380 |
ial.Run(self.op.iallocator)
|
8381 |
|
8382 |
if not ial.success: |
8383 |
raise errors.OpPrereqError("Can't compute node evacuation using" |
8384 |
" iallocator '%s': %s" %
|
8385 |
(self.op.iallocator, ial.info),
|
8386 |
errors.ECODE_NORES) |
8387 |
|
8388 |
jobs = _LoadNodeEvacResult(self, ial.result, self.op.early_release, True) |
8389 |
|
8390 |
elif self.op.remote_node is not None: |
8391 |
assert self.op.mode == constants.NODE_EVAC_SEC |
8392 |
jobs = [ |
8393 |
[opcodes.OpInstanceReplaceDisks(instance_name=instance_name, |
8394 |
remote_node=self.op.remote_node,
|
8395 |
disks=[], |
8396 |
mode=constants.REPLACE_DISK_CHG, |
8397 |
early_release=self.op.early_release)]
|
8398 |
for instance_name in self.instance_names] |
8399 |
|
8400 |
else:
|
8401 |
raise errors.ProgrammerError("No iallocator or remote node") |
8402 |
|
8403 |
return ResultWithJobs(jobs)
|
8404 |
|
8405 |
|
8406 |
def _DiskSizeInBytesToMebibytes(lu, size): |
8407 |
"""Converts a disk size in bytes to mebibytes.
|
8408 |
|
8409 |
Warns and rounds up if the size isn't an even multiple of 1 MiB.
|
8410 |
|
8411 |
"""
|
8412 |
(mib, remainder) = divmod(size, 1024 * 1024) |
8413 |
|
8414 |
if remainder != 0: |
8415 |
lu.LogWarning("Disk size is not an even multiple of 1 MiB; rounding up"
|
8416 |
" to not overwrite existing data (%s bytes will not be"
|
8417 |
" wiped)", (1024 * 1024) - remainder) |
8418 |
mib += 1
|
8419 |
|
8420 |
return mib
|
8421 |
|
8422 |
|
8423 |
class LUInstanceGrowDisk(LogicalUnit): |
8424 |
"""Grow a disk of an instance.
|
8425 |
|
8426 |
"""
|
8427 |
HPATH = "disk-grow"
|
8428 |
HTYPE = constants.HTYPE_INSTANCE |
8429 |
REQ_BGL = False
|
8430 |
|
8431 |
def ExpandNames(self): |
8432 |
self._ExpandAndLockInstance()
|
8433 |
self.needed_locks[locking.LEVEL_NODE] = []
|
8434 |
self.needed_locks[locking.LEVEL_NODE_RES] = []
|
8435 |
self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
|
8436 |
self.recalculate_locks[locking.LEVEL_NODE_RES] = constants.LOCKS_REPLACE
|
8437 |
|
8438 |
def DeclareLocks(self, level): |
8439 |
if level == locking.LEVEL_NODE:
|
8440 |
self._LockInstancesNodes()
|
8441 |
elif level == locking.LEVEL_NODE_RES:
|
8442 |
# Copy node locks
|
8443 |
self.needed_locks[locking.LEVEL_NODE_RES] = \
|
8444 |
_CopyLockList(self.needed_locks[locking.LEVEL_NODE])
|
8445 |
|
8446 |
def BuildHooksEnv(self): |
8447 |
"""Build hooks env.
|
8448 |
|
8449 |
This runs on the master, the primary and all the secondaries.
|
8450 |
|
8451 |
"""
|
8452 |
env = { |
8453 |
"DISK": self.op.disk, |
8454 |
"AMOUNT": self.op.amount, |
8455 |
"ABSOLUTE": self.op.absolute, |
8456 |
} |
8457 |
env.update(_BuildInstanceHookEnvByObject(self, self.instance)) |
8458 |
return env
|
8459 |
|
8460 |
def BuildHooksNodes(self): |
8461 |
"""Build hooks nodes.
|
8462 |
|
8463 |
"""
|
8464 |
nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes) |
8465 |
return (nl, nl)
|
8466 |
|
8467 |
def CheckPrereq(self): |
8468 |
"""Check prerequisites.
|
8469 |
|
8470 |
This checks that the instance is in the cluster.
|
8471 |
|
8472 |
"""
|
8473 |
instance = self.cfg.GetInstanceInfo(self.op.instance_name) |
8474 |
assert instance is not None, \ |
8475 |
"Cannot retrieve locked instance %s" % self.op.instance_name |
8476 |
nodenames = list(instance.all_nodes)
|
8477 |
for node in nodenames: |
8478 |
_CheckNodeOnline(self, node)
|
8479 |
|
8480 |
self.instance = instance
|
8481 |
|
8482 |
if instance.disk_template not in constants.DTS_GROWABLE: |
8483 |
raise errors.OpPrereqError("Instance's disk layout does not support" |
8484 |
" growing", errors.ECODE_INVAL)
|
8485 |
|
8486 |
self.disk = instance.FindDisk(self.op.disk) |
8487 |
|
8488 |
if self.op.absolute: |
8489 |
self.target = self.op.amount |
8490 |
self.delta = self.target - self.disk.size |
8491 |
if self.delta < 0: |
8492 |
raise errors.OpPrereqError("Requested size (%s) is smaller than " |
8493 |
"current disk size (%s)" %
|
8494 |
(utils.FormatUnit(self.target, "h"), |
8495 |
utils.FormatUnit(self.disk.size, "h")), |
8496 |
errors.ECODE_STATE) |
8497 |
else:
|
8498 |
self.delta = self.op.amount |
8499 |
self.target = self.disk.size + self.delta |
8500 |
if self.delta < 0: |
8501 |
raise errors.OpPrereqError("Requested increment (%s) is negative" % |
8502 |
utils.FormatUnit(self.delta, "h"), |
8503 |
errors.ECODE_INVAL) |
8504 |
|
8505 |
self._CheckDiskSpace(nodenames, self.disk.ComputeGrowth(self.delta)) |
8506 |
|
8507 |
def _CheckDiskSpace(self, nodenames, req_vgspace): |
8508 |
template = self.instance.disk_template
|
8509 |
if template not in (constants.DTS_NO_FREE_SPACE_CHECK): |
8510 |
# TODO: check the free disk space for file, when that feature will be
|
8511 |
# supported
|
8512 |
nodes = map(self.cfg.GetNodeInfo, nodenames) |
8513 |
es_nodes = filter(lambda n: _IsExclusiveStorageEnabledNode(self.cfg, n), |
8514 |
nodes) |
8515 |
if es_nodes:
|
8516 |
# With exclusive storage we need to something smarter than just looking
|
8517 |
# at free space; for now, let's simply abort the operation.
|
8518 |
raise errors.OpPrereqError("Cannot grow disks when exclusive_storage" |
8519 |
" is enabled", errors.ECODE_STATE)
|
8520 |
_CheckNodesFreeDiskPerVG(self, nodenames, req_vgspace)
|
8521 |
|
8522 |
def Exec(self, feedback_fn): |
8523 |
"""Execute disk grow.
|
8524 |
|
8525 |
"""
|
8526 |
instance = self.instance
|
8527 |
disk = self.disk
|
8528 |
|
8529 |
assert set([instance.name]) == self.owned_locks(locking.LEVEL_INSTANCE) |
8530 |
assert (self.owned_locks(locking.LEVEL_NODE) == |
8531 |
self.owned_locks(locking.LEVEL_NODE_RES))
|
8532 |
|
8533 |
wipe_disks = self.cfg.GetClusterInfo().prealloc_wipe_disks
|
8534 |
|
8535 |
disks_ok, _ = _AssembleInstanceDisks(self, self.instance, disks=[disk]) |
8536 |
if not disks_ok: |
8537 |
raise errors.OpExecError("Cannot activate block device to grow") |
8538 |
|
8539 |
feedback_fn("Growing disk %s of instance '%s' by %s to %s" %
|
8540 |
(self.op.disk, instance.name,
|
8541 |
utils.FormatUnit(self.delta, "h"), |
8542 |
utils.FormatUnit(self.target, "h"))) |
8543 |
|
8544 |
# First run all grow ops in dry-run mode
|
8545 |
for node in instance.all_nodes: |
8546 |
self.cfg.SetDiskID(disk, node)
|
8547 |
result = self.rpc.call_blockdev_grow(node, (disk, instance), self.delta, |
8548 |
True, True) |
8549 |
result.Raise("Dry-run grow request failed to node %s" % node)
|
8550 |
|
8551 |
if wipe_disks:
|
8552 |
# Get disk size from primary node for wiping
|
8553 |
result = self.rpc.call_blockdev_getsize(instance.primary_node, [disk])
|
8554 |
result.Raise("Failed to retrieve disk size from node '%s'" %
|
8555 |
instance.primary_node) |
8556 |
|
8557 |
(disk_size_in_bytes, ) = result.payload |
8558 |
|
8559 |
if disk_size_in_bytes is None: |
8560 |
raise errors.OpExecError("Failed to retrieve disk size from primary" |
8561 |
" node '%s'" % instance.primary_node)
|
8562 |
|
8563 |
old_disk_size = _DiskSizeInBytesToMebibytes(self, disk_size_in_bytes)
|
8564 |
|
8565 |
assert old_disk_size >= disk.size, \
|
8566 |
("Retrieved disk size too small (got %s, should be at least %s)" %
|
8567 |
(old_disk_size, disk.size)) |
8568 |
else:
|
8569 |
old_disk_size = None
|
8570 |
|
8571 |
# We know that (as far as we can test) operations across different
|
8572 |
# nodes will succeed, time to run it for real on the backing storage
|
8573 |
for node in instance.all_nodes: |
8574 |
self.cfg.SetDiskID(disk, node)
|
8575 |
result = self.rpc.call_blockdev_grow(node, (disk, instance), self.delta, |
8576 |
False, True) |
8577 |
result.Raise("Grow request failed to node %s" % node)
|
8578 |
|
8579 |
# And now execute it for logical storage, on the primary node
|
8580 |
node = instance.primary_node |
8581 |
self.cfg.SetDiskID(disk, node)
|
8582 |
result = self.rpc.call_blockdev_grow(node, (disk, instance), self.delta, |
8583 |
False, False) |
8584 |
result.Raise("Grow request failed to node %s" % node)
|
8585 |
|
8586 |
disk.RecordGrow(self.delta)
|
8587 |
self.cfg.Update(instance, feedback_fn)
|
8588 |
|
8589 |
# Changes have been recorded, release node lock
|
8590 |
_ReleaseLocks(self, locking.LEVEL_NODE)
|
8591 |
|
8592 |
# Downgrade lock while waiting for sync
|
8593 |
self.glm.downgrade(locking.LEVEL_INSTANCE)
|
8594 |
|
8595 |
assert wipe_disks ^ (old_disk_size is None) |
8596 |
|
8597 |
if wipe_disks:
|
8598 |
assert instance.disks[self.op.disk] == disk |
8599 |
|
8600 |
# Wipe newly added disk space
|
8601 |
_WipeDisks(self, instance,
|
8602 |
disks=[(self.op.disk, disk, old_disk_size)])
|
8603 |
|
8604 |
if self.op.wait_for_sync: |
8605 |
disk_abort = not _WaitForSync(self, instance, disks=[disk]) |
8606 |
if disk_abort:
|
8607 |
self.LogWarning("Disk syncing has not returned a good status; check" |
8608 |
" the instance")
|
8609 |
if instance.admin_state != constants.ADMINST_UP:
|
8610 |
_SafeShutdownInstanceDisks(self, instance, disks=[disk])
|
8611 |
elif instance.admin_state != constants.ADMINST_UP:
|
8612 |
self.LogWarning("Not shutting down the disk even if the instance is" |
8613 |
" not supposed to be running because no wait for"
|
8614 |
" sync mode was requested")
|
8615 |
|
8616 |
assert self.owned_locks(locking.LEVEL_NODE_RES) |
8617 |
assert set([instance.name]) == self.owned_locks(locking.LEVEL_INSTANCE) |
8618 |
|
8619 |
|
8620 |
class LUInstanceQueryData(NoHooksLU): |
8621 |
"""Query runtime instance data.
|
8622 |
|
8623 |
"""
|
8624 |
REQ_BGL = False
|
8625 |
|
8626 |
def ExpandNames(self): |
8627 |
self.needed_locks = {}
|
8628 |
|
8629 |
# Use locking if requested or when non-static information is wanted
|
8630 |
if not (self.op.static or self.op.use_locking): |
8631 |
self.LogWarning("Non-static data requested, locks need to be acquired") |
8632 |
self.op.use_locking = True |
8633 |
|
8634 |
if self.op.instances or not self.op.use_locking: |
8635 |
# Expand instance names right here
|
8636 |
self.wanted_names = _GetWantedInstances(self, self.op.instances) |
8637 |
else:
|
8638 |
# Will use acquired locks
|
8639 |
self.wanted_names = None |
8640 |
|
8641 |
if self.op.use_locking: |
8642 |
self.share_locks = _ShareAll()
|
8643 |
|
8644 |
if self.wanted_names is None: |
8645 |
self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
|
8646 |
else:
|
8647 |
self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names |
8648 |
|
8649 |
self.needed_locks[locking.LEVEL_NODEGROUP] = []
|
8650 |
self.needed_locks[locking.LEVEL_NODE] = []
|
8651 |
self.needed_locks[locking.LEVEL_NETWORK] = []
|
8652 |
self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
|
8653 |
|
8654 |
def DeclareLocks(self, level): |
8655 |
if self.op.use_locking: |
8656 |
owned_instances = self.owned_locks(locking.LEVEL_INSTANCE)
|
8657 |
if level == locking.LEVEL_NODEGROUP:
|
8658 |
|
8659 |
# Lock all groups used by instances optimistically; this requires going
|
8660 |
# via the node before it's locked, requiring verification later on
|
8661 |
self.needed_locks[locking.LEVEL_NODEGROUP] = \
|
8662 |
frozenset(group_uuid
|
8663 |
for instance_name in owned_instances |
8664 |
for group_uuid in |
8665 |
self.cfg.GetInstanceNodeGroups(instance_name))
|
8666 |
|
8667 |
elif level == locking.LEVEL_NODE:
|
8668 |
self._LockInstancesNodes()
|
8669 |
|
8670 |
elif level == locking.LEVEL_NETWORK:
|
8671 |
self.needed_locks[locking.LEVEL_NETWORK] = \
|
8672 |
frozenset(net_uuid
|
8673 |
for instance_name in owned_instances |
8674 |
for net_uuid in |
8675 |
self.cfg.GetInstanceNetworks(instance_name))
|
8676 |
|
8677 |
def CheckPrereq(self): |
8678 |
"""Check prerequisites.
|
8679 |
|
8680 |
This only checks the optional instance list against the existing names.
|
8681 |
|
8682 |
"""
|
8683 |
owned_instances = frozenset(self.owned_locks(locking.LEVEL_INSTANCE)) |
8684 |
owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP)) |
8685 |
owned_nodes = frozenset(self.owned_locks(locking.LEVEL_NODE)) |
8686 |
owned_networks = frozenset(self.owned_locks(locking.LEVEL_NETWORK)) |
8687 |
|
8688 |
if self.wanted_names is None: |
8689 |
assert self.op.use_locking, "Locking was not used" |
8690 |
self.wanted_names = owned_instances
|
8691 |
|
8692 |
instances = dict(self.cfg.GetMultiInstanceInfo(self.wanted_names)) |
8693 |
|
8694 |
if self.op.use_locking: |
8695 |
_CheckInstancesNodeGroups(self.cfg, instances, owned_groups, owned_nodes,
|
8696 |
None)
|
8697 |
else:
|
8698 |
assert not (owned_instances or owned_groups or |
8699 |
owned_nodes or owned_networks)
|
8700 |
|
8701 |
self.wanted_instances = instances.values()
|
8702 |
|
8703 |
def _ComputeBlockdevStatus(self, node, instance, dev): |
8704 |
"""Returns the status of a block device
|
8705 |
|
8706 |
"""
|
8707 |
if self.op.static or not node: |
8708 |
return None |
8709 |
|
8710 |
self.cfg.SetDiskID(dev, node)
|
8711 |
|
8712 |
result = self.rpc.call_blockdev_find(node, dev)
|
8713 |
if result.offline:
|
8714 |
return None |
8715 |
|
8716 |
result.Raise("Can't compute disk status for %s" % instance.name)
|
8717 |
|
8718 |
status = result.payload |
8719 |
if status is None: |
8720 |
return None |
8721 |
|
8722 |
return (status.dev_path, status.major, status.minor,
|
8723 |
status.sync_percent, status.estimated_time, |
8724 |
status.is_degraded, status.ldisk_status) |
8725 |
|
8726 |
def _ComputeDiskStatus(self, instance, snode, dev): |
8727 |
"""Compute block device status.
|
8728 |
|
8729 |
"""
|
8730 |
(anno_dev,) = _AnnotateDiskParams(instance, [dev], self.cfg)
|
8731 |
|
8732 |
return self._ComputeDiskStatusInner(instance, snode, anno_dev) |
8733 |
|
8734 |
def _ComputeDiskStatusInner(self, instance, snode, dev): |
8735 |
"""Compute block device status.
|
8736 |
|
8737 |
@attention: The device has to be annotated already.
|
8738 |
|
8739 |
"""
|
8740 |
if dev.dev_type in constants.LDS_DRBD: |
8741 |
# we change the snode then (otherwise we use the one passed in)
|
8742 |
if dev.logical_id[0] == instance.primary_node: |
8743 |
snode = dev.logical_id[1]
|
8744 |
else:
|
8745 |
snode = dev.logical_id[0]
|
8746 |
|
8747 |
dev_pstatus = self._ComputeBlockdevStatus(instance.primary_node,
|
8748 |
instance, dev) |
8749 |
dev_sstatus = self._ComputeBlockdevStatus(snode, instance, dev)
|
8750 |
|
8751 |
if dev.children:
|
8752 |
dev_children = map(compat.partial(self._ComputeDiskStatusInner, |
8753 |
instance, snode), |
8754 |
dev.children) |
8755 |
else:
|
8756 |
dev_children = [] |
8757 |
|
8758 |
return {
|
8759 |
"iv_name": dev.iv_name,
|
8760 |
"dev_type": dev.dev_type,
|
8761 |
"logical_id": dev.logical_id,
|
8762 |
"physical_id": dev.physical_id,
|
8763 |
"pstatus": dev_pstatus,
|
8764 |
"sstatus": dev_sstatus,
|
8765 |
"children": dev_children,
|
8766 |
"mode": dev.mode,
|
8767 |
"size": dev.size,
|
8768 |
"name": dev.name,
|
8769 |
"uuid": dev.uuid,
|
8770 |
} |
8771 |
|
8772 |
def Exec(self, feedback_fn): |
8773 |
"""Gather and return data"""
|
8774 |
result = {} |
8775 |
|
8776 |
cluster = self.cfg.GetClusterInfo()
|
8777 |
|
8778 |
node_names = itertools.chain(*(i.all_nodes for i in self.wanted_instances)) |
8779 |
nodes = dict(self.cfg.GetMultiNodeInfo(node_names)) |
8780 |
|
8781 |
groups = dict(self.cfg.GetMultiNodeGroupInfo(node.group |
8782 |
for node in nodes.values())) |
8783 |
|
8784 |
group2name_fn = lambda uuid: groups[uuid].name
|
8785 |
for instance in self.wanted_instances: |
8786 |
pnode = nodes[instance.primary_node] |
8787 |
|
8788 |
if self.op.static or pnode.offline: |
8789 |
remote_state = None
|
8790 |
if pnode.offline:
|
8791 |
self.LogWarning("Primary node %s is marked offline, returning static" |
8792 |
" information only for instance %s" %
|
8793 |
(pnode.name, instance.name)) |
8794 |
else:
|
8795 |
remote_info = self.rpc.call_instance_info(instance.primary_node,
|
8796 |
instance.name, |
8797 |
instance.hypervisor) |
8798 |
remote_info.Raise("Error checking node %s" % instance.primary_node)
|
8799 |
remote_info = remote_info.payload |
8800 |
if remote_info and "state" in remote_info: |
8801 |
remote_state = "up"
|
8802 |
else:
|
8803 |
if instance.admin_state == constants.ADMINST_UP:
|
8804 |
remote_state = "down"
|
8805 |
else:
|
8806 |
remote_state = instance.admin_state |
8807 |
|
8808 |
disks = map(compat.partial(self._ComputeDiskStatus, instance, None), |
8809 |
instance.disks) |
8810 |
|
8811 |
snodes_group_uuids = [nodes[snode_name].group |
8812 |
for snode_name in instance.secondary_nodes] |
8813 |
|
8814 |
result[instance.name] = { |
8815 |
"name": instance.name,
|
8816 |
"config_state": instance.admin_state,
|
8817 |
"run_state": remote_state,
|
8818 |
"pnode": instance.primary_node,
|
8819 |
"pnode_group_uuid": pnode.group,
|
8820 |
"pnode_group_name": group2name_fn(pnode.group),
|
8821 |
"snodes": instance.secondary_nodes,
|
8822 |
"snodes_group_uuids": snodes_group_uuids,
|
8823 |
"snodes_group_names": map(group2name_fn, snodes_group_uuids), |
8824 |
"os": instance.os,
|
8825 |
# this happens to be the same format used for hooks
|
8826 |
"nics": _NICListToTuple(self, instance.nics), |
8827 |
"disk_template": instance.disk_template,
|
8828 |
"disks": disks,
|
8829 |
"hypervisor": instance.hypervisor,
|
8830 |
"network_port": instance.network_port,
|
8831 |
"hv_instance": instance.hvparams,
|
8832 |
"hv_actual": cluster.FillHV(instance, skip_globals=True), |
8833 |
"be_instance": instance.beparams,
|
8834 |
"be_actual": cluster.FillBE(instance),
|
8835 |
"os_instance": instance.osparams,
|
8836 |
"os_actual": cluster.SimpleFillOS(instance.os, instance.osparams),
|
8837 |
"serial_no": instance.serial_no,
|
8838 |
"mtime": instance.mtime,
|
8839 |
"ctime": instance.ctime,
|
8840 |
"uuid": instance.uuid,
|
8841 |
} |
8842 |
|
8843 |
return result
|
8844 |
|
8845 |
|
8846 |
def PrepareContainerMods(mods, private_fn): |
8847 |
"""Prepares a list of container modifications by adding a private data field.
|
8848 |
|
8849 |
@type mods: list of tuples; (operation, index, parameters)
|
8850 |
@param mods: List of modifications
|
8851 |
@type private_fn: callable or None
|
8852 |
@param private_fn: Callable for constructing a private data field for a
|
8853 |
modification
|
8854 |
@rtype: list
|
8855 |
|
8856 |
"""
|
8857 |
if private_fn is None: |
8858 |
fn = lambda: None |
8859 |
else:
|
8860 |
fn = private_fn |
8861 |
|
8862 |
return [(op, idx, params, fn()) for (op, idx, params) in mods] |
8863 |
|
8864 |
|
8865 |
def GetItemFromContainer(identifier, kind, container): |
8866 |
"""Return the item refered by the identifier.
|
8867 |
|
8868 |
@type identifier: string
|
8869 |
@param identifier: Item index or name or UUID
|
8870 |
@type kind: string
|
8871 |
@param kind: One-word item description
|
8872 |
@type container: list
|
8873 |
@param container: Container to get the item from
|
8874 |
|
8875 |
"""
|
8876 |
# Index
|
8877 |
try:
|
8878 |
idx = int(identifier)
|
8879 |
if idx == -1: |
8880 |
# Append
|
8881 |
absidx = len(container) - 1 |
8882 |
elif idx < 0: |
8883 |
raise IndexError("Not accepting negative indices other than -1") |
8884 |
elif idx > len(container): |
8885 |
raise IndexError("Got %s index %s, but there are only %s" % |
8886 |
(kind, idx, len(container)))
|
8887 |
else:
|
8888 |
absidx = idx |
8889 |
return (absidx, container[idx])
|
8890 |
except ValueError: |
8891 |
pass
|
8892 |
|
8893 |
for idx, item in enumerate(container): |
8894 |
if item.uuid == identifier or item.name == identifier: |
8895 |
return (idx, item)
|
8896 |
|
8897 |
raise errors.OpPrereqError("Cannot find %s with identifier %s" % |
8898 |
(kind, identifier), errors.ECODE_NOENT) |
8899 |
|
8900 |
|
8901 |
#: Type description for changes as returned by L{ApplyContainerMods}'s
|
8902 |
#: callbacks
|
8903 |
_TApplyContModsCbChanges = \ |
8904 |
ht.TMaybeListOf(ht.TAnd(ht.TIsLength(2), ht.TItems([
|
8905 |
ht.TNonEmptyString, |
8906 |
ht.TAny, |
8907 |
]))) |
8908 |
|
8909 |
|
8910 |
def ApplyContainerMods(kind, container, chgdesc, mods, |
8911 |
create_fn, modify_fn, remove_fn): |
8912 |
"""Applies descriptions in C{mods} to C{container}.
|
8913 |
|
8914 |
@type kind: string
|
8915 |
@param kind: One-word item description
|
8916 |
@type container: list
|
8917 |
@param container: Container to modify
|
8918 |
@type chgdesc: None or list
|
8919 |
@param chgdesc: List of applied changes
|
8920 |
@type mods: list
|
8921 |
@param mods: Modifications as returned by L{PrepareContainerMods}
|
8922 |
@type create_fn: callable
|
8923 |
@param create_fn: Callback for creating a new item (L{constants.DDM_ADD});
|
8924 |
receives absolute item index, parameters and private data object as added
|
8925 |
by L{PrepareContainerMods}, returns tuple containing new item and changes
|
8926 |
as list
|
8927 |
@type modify_fn: callable
|
8928 |
@param modify_fn: Callback for modifying an existing item
|
8929 |
(L{constants.DDM_MODIFY}); receives absolute item index, item, parameters
|
8930 |
and private data object as added by L{PrepareContainerMods}, returns
|
8931 |
changes as list
|
8932 |
@type remove_fn: callable
|
8933 |
@param remove_fn: Callback on removing item; receives absolute item index,
|
8934 |
item and private data object as added by L{PrepareContainerMods}
|
8935 |
|
8936 |
"""
|
8937 |
for (op, identifier, params, private) in mods: |
8938 |
changes = None
|
8939 |
|
8940 |
if op == constants.DDM_ADD:
|
8941 |
# Calculate where item will be added
|
8942 |
# When adding an item, identifier can only be an index
|
8943 |
try:
|
8944 |
idx = int(identifier)
|
8945 |
except ValueError: |
8946 |
raise errors.OpPrereqError("Only possitive integer or -1 is accepted as" |
8947 |
" identifier for %s" % constants.DDM_ADD,
|
8948 |
errors.ECODE_INVAL) |
8949 |
if idx == -1: |
8950 |
addidx = len(container)
|
8951 |
else:
|
8952 |
if idx < 0: |
8953 |
raise IndexError("Not accepting negative indices other than -1") |
8954 |
elif idx > len(container): |
8955 |
raise IndexError("Got %s index %s, but there are only %s" % |
8956 |
(kind, idx, len(container)))
|
8957 |
addidx = idx |
8958 |
|
8959 |
if create_fn is None: |
8960 |
item = params |
8961 |
else:
|
8962 |
(item, changes) = create_fn(addidx, params, private) |
8963 |
|
8964 |
if idx == -1: |
8965 |
container.append(item) |
8966 |
else:
|
8967 |
assert idx >= 0 |
8968 |
assert idx <= len(container) |
8969 |
# list.insert does so before the specified index
|
8970 |
container.insert(idx, item) |
8971 |
else:
|
8972 |
# Retrieve existing item
|
8973 |
(absidx, item) = GetItemFromContainer(identifier, kind, container) |
8974 |
|
8975 |
if op == constants.DDM_REMOVE:
|
8976 |
assert not params |
8977 |
|
8978 |
if remove_fn is not None: |
8979 |
remove_fn(absidx, item, private) |
8980 |
|
8981 |
changes = [("%s/%s" % (kind, absidx), "remove")] |
8982 |
|
8983 |
assert container[absidx] == item
|
8984 |
del container[absidx]
|
8985 |
elif op == constants.DDM_MODIFY:
|
8986 |
if modify_fn is not None: |
8987 |
changes = modify_fn(absidx, item, params, private) |
8988 |
else:
|
8989 |
raise errors.ProgrammerError("Unhandled operation '%s'" % op) |
8990 |
|
8991 |
assert _TApplyContModsCbChanges(changes)
|
8992 |
|
8993 |
if not (chgdesc is None or changes is None): |
8994 |
chgdesc.extend(changes) |
8995 |
|
8996 |
|
8997 |
def _UpdateIvNames(base_index, disks): |
8998 |
"""Updates the C{iv_name} attribute of disks.
|
8999 |
|
9000 |
@type disks: list of L{objects.Disk}
|
9001 |
|
9002 |
"""
|
9003 |
for (idx, disk) in enumerate(disks): |
9004 |
disk.iv_name = "disk/%s" % (base_index + idx, )
|
9005 |
|
9006 |
|
9007 |
class _InstNicModPrivate: |
9008 |
"""Data structure for network interface modifications.
|
9009 |
|
9010 |
Used by L{LUInstanceSetParams}.
|
9011 |
|
9012 |
"""
|
9013 |
def __init__(self): |
9014 |
self.params = None |
9015 |
self.filled = None |
9016 |
|
9017 |
|
9018 |
class LUInstanceSetParams(LogicalUnit): |
9019 |
"""Modifies an instances's parameters.
|
9020 |
|
9021 |
"""
|
9022 |
HPATH = "instance-modify"
|
9023 |
HTYPE = constants.HTYPE_INSTANCE |
9024 |
REQ_BGL = False
|
9025 |
|
9026 |
@staticmethod
|
9027 |
def _UpgradeDiskNicMods(kind, mods, verify_fn): |
9028 |
assert ht.TList(mods)
|
9029 |
assert not mods or len(mods[0]) in (2, 3) |
9030 |
|
9031 |
if mods and len(mods[0]) == 2: |
9032 |
result = [] |
9033 |
|
9034 |
addremove = 0
|
9035 |
for op, params in mods: |
9036 |
if op in (constants.DDM_ADD, constants.DDM_REMOVE): |
9037 |
result.append((op, -1, params))
|
9038 |
addremove += 1
|
9039 |
|
9040 |
if addremove > 1: |
9041 |
raise errors.OpPrereqError("Only one %s add or remove operation is" |
9042 |
" supported at a time" % kind,
|
9043 |
errors.ECODE_INVAL) |
9044 |
else:
|
9045 |
result.append((constants.DDM_MODIFY, op, params)) |
9046 |
|
9047 |
assert verify_fn(result)
|
9048 |
else:
|
9049 |
result = mods |
9050 |
|
9051 |
return result
|
9052 |
|
9053 |
@staticmethod
|
9054 |
def _CheckMods(kind, mods, key_types, item_fn): |
9055 |
"""Ensures requested disk/NIC modifications are valid.
|
9056 |
|
9057 |
"""
|
9058 |
for (op, _, params) in mods: |
9059 |
assert ht.TDict(params)
|
9060 |
|
9061 |
# If 'key_types' is an empty dict, we assume we have an
|
9062 |
# 'ext' template and thus do not ForceDictType
|
9063 |
if key_types:
|
9064 |
utils.ForceDictType(params, key_types) |
9065 |
|
9066 |
if op == constants.DDM_REMOVE:
|
9067 |
if params:
|
9068 |
raise errors.OpPrereqError("No settings should be passed when" |
9069 |
" removing a %s" % kind,
|
9070 |
errors.ECODE_INVAL) |
9071 |
elif op in (constants.DDM_ADD, constants.DDM_MODIFY): |
9072 |
item_fn(op, params) |
9073 |
else:
|
9074 |
raise errors.ProgrammerError("Unhandled operation '%s'" % op) |
9075 |
|
9076 |
@staticmethod
|
9077 |
def _VerifyDiskModification(op, params): |
9078 |
"""Verifies a disk modification.
|
9079 |
|
9080 |
"""
|
9081 |
if op == constants.DDM_ADD:
|
9082 |
mode = params.setdefault(constants.IDISK_MODE, constants.DISK_RDWR) |
9083 |
if mode not in constants.DISK_ACCESS_SET: |
9084 |
raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode, |
9085 |
errors.ECODE_INVAL) |
9086 |
|
9087 |
size = params.get(constants.IDISK_SIZE, None)
|
9088 |
if size is None: |
9089 |
raise errors.OpPrereqError("Required disk parameter '%s' missing" % |
9090 |
constants.IDISK_SIZE, errors.ECODE_INVAL) |
9091 |
|
9092 |
try:
|
9093 |
size = int(size)
|
9094 |
except (TypeError, ValueError), err: |
9095 |
raise errors.OpPrereqError("Invalid disk size parameter: %s" % err, |
9096 |
errors.ECODE_INVAL) |
9097 |
|
9098 |
params[constants.IDISK_SIZE] = size |
9099 |
name = params.get(constants.IDISK_NAME, None)
|
9100 |
if name is not None and name.lower() == constants.VALUE_NONE: |
9101 |
params[constants.IDISK_NAME] = None
|
9102 |
|
9103 |
elif op == constants.DDM_MODIFY:
|
9104 |
if constants.IDISK_SIZE in params: |
9105 |
raise errors.OpPrereqError("Disk size change not possible, use" |
9106 |
" grow-disk", errors.ECODE_INVAL)
|
9107 |
if len(params) > 2: |
9108 |
raise errors.OpPrereqError("Disk modification doesn't support" |
9109 |
" additional arbitrary parameters",
|
9110 |
errors.ECODE_INVAL) |
9111 |
name = params.get(constants.IDISK_NAME, None)
|
9112 |
if name is not None and name.lower() == constants.VALUE_NONE: |
9113 |
params[constants.IDISK_NAME] = None
|
9114 |
|
9115 |
@staticmethod
|
9116 |
def _VerifyNicModification(op, params): |
9117 |
"""Verifies a network interface modification.
|
9118 |
|
9119 |
"""
|
9120 |
if op in (constants.DDM_ADD, constants.DDM_MODIFY): |
9121 |
ip = params.get(constants.INIC_IP, None)
|
9122 |
name = params.get(constants.INIC_NAME, None)
|
9123 |
req_net = params.get(constants.INIC_NETWORK, None)
|
9124 |
link = params.get(constants.NIC_LINK, None)
|
9125 |
mode = params.get(constants.NIC_MODE, None)
|
9126 |
if name is not None and name.lower() == constants.VALUE_NONE: |
9127 |
params[constants.INIC_NAME] = None
|
9128 |
if req_net is not None: |
9129 |
if req_net.lower() == constants.VALUE_NONE:
|
9130 |
params[constants.INIC_NETWORK] = None
|
9131 |
req_net = None
|
9132 |
elif link is not None or mode is not None: |
9133 |
raise errors.OpPrereqError("If network is given" |
9134 |
" mode or link should not",
|
9135 |
errors.ECODE_INVAL) |
9136 |
|
9137 |
if op == constants.DDM_ADD:
|
9138 |
macaddr = params.get(constants.INIC_MAC, None)
|
9139 |
if macaddr is None: |
9140 |
params[constants.INIC_MAC] = constants.VALUE_AUTO |
9141 |
|
9142 |
if ip is not None: |
9143 |
if ip.lower() == constants.VALUE_NONE:
|
9144 |
params[constants.INIC_IP] = None
|
9145 |
else:
|
9146 |
if ip.lower() == constants.NIC_IP_POOL:
|
9147 |
if op == constants.DDM_ADD and req_net is None: |
9148 |
raise errors.OpPrereqError("If ip=pool, parameter network" |
9149 |
" cannot be none",
|
9150 |
errors.ECODE_INVAL) |
9151 |
else:
|
9152 |
if not netutils.IPAddress.IsValid(ip): |
9153 |
raise errors.OpPrereqError("Invalid IP address '%s'" % ip, |
9154 |
errors.ECODE_INVAL) |
9155 |
|
9156 |
if constants.INIC_MAC in params: |
9157 |
macaddr = params[constants.INIC_MAC] |
9158 |
if macaddr not in (constants.VALUE_AUTO, constants.VALUE_GENERATE): |
9159 |
macaddr = utils.NormalizeAndValidateMac(macaddr) |
9160 |
|
9161 |
if op == constants.DDM_MODIFY and macaddr == constants.VALUE_AUTO: |
9162 |
raise errors.OpPrereqError("'auto' is not a valid MAC address when" |
9163 |
" modifying an existing NIC",
|
9164 |
errors.ECODE_INVAL) |
9165 |
|
9166 |
def CheckArguments(self): |
9167 |
if not (self.op.nics or self.op.disks or self.op.disk_template or |
9168 |
self.op.hvparams or self.op.beparams or self.op.os_name or |
9169 |
self.op.offline is not None or self.op.runtime_mem or |
9170 |
self.op.pnode):
|
9171 |
raise errors.OpPrereqError("No changes submitted", errors.ECODE_INVAL) |
9172 |
|
9173 |
if self.op.hvparams: |
9174 |
_CheckParamsNotGlobal(self.op.hvparams, constants.HVC_GLOBALS,
|
9175 |
"hypervisor", "instance", "cluster") |
9176 |
|
9177 |
self.op.disks = self._UpgradeDiskNicMods( |
9178 |
"disk", self.op.disks, opcodes.OpInstanceSetParams.TestDiskModifications) |
9179 |
self.op.nics = self._UpgradeDiskNicMods( |
9180 |
"NIC", self.op.nics, opcodes.OpInstanceSetParams.TestNicModifications) |
9181 |
|
9182 |
if self.op.disks and self.op.disk_template is not None: |
9183 |
raise errors.OpPrereqError("Disk template conversion and other disk" |
9184 |
" changes not supported at the same time",
|
9185 |
errors.ECODE_INVAL) |
9186 |
|
9187 |
if (self.op.disk_template and |
9188 |
self.op.disk_template in constants.DTS_INT_MIRROR and |
9189 |
self.op.remote_node is None): |
9190 |
raise errors.OpPrereqError("Changing the disk template to a mirrored" |
9191 |
" one requires specifying a secondary node",
|
9192 |
errors.ECODE_INVAL) |
9193 |
|
9194 |
# Check NIC modifications
|
9195 |
self._CheckMods("NIC", self.op.nics, constants.INIC_PARAMS_TYPES, |
9196 |
self._VerifyNicModification)
|
9197 |
|
9198 |
if self.op.pnode: |
9199 |
self.op.pnode = _ExpandNodeName(self.cfg, self.op.pnode) |
9200 |
|
9201 |
def ExpandNames(self): |
9202 |
self._ExpandAndLockInstance()
|
9203 |
self.needed_locks[locking.LEVEL_NODEGROUP] = []
|
9204 |
# Can't even acquire node locks in shared mode as upcoming changes in
|
9205 |
# Ganeti 2.6 will start to modify the node object on disk conversion
|
9206 |
self.needed_locks[locking.LEVEL_NODE] = []
|
9207 |
self.needed_locks[locking.LEVEL_NODE_RES] = []
|
9208 |
self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
|
9209 |
# Look node group to look up the ipolicy
|
9210 |
self.share_locks[locking.LEVEL_NODEGROUP] = 1 |
9211 |
|
9212 |
def DeclareLocks(self, level): |
9213 |
if level == locking.LEVEL_NODEGROUP:
|
9214 |
assert not self.needed_locks[locking.LEVEL_NODEGROUP] |
9215 |
# Acquire locks for the instance's nodegroups optimistically. Needs
|
9216 |
# to be verified in CheckPrereq
|
9217 |
self.needed_locks[locking.LEVEL_NODEGROUP] = \
|
9218 |
self.cfg.GetInstanceNodeGroups(self.op.instance_name) |
9219 |
elif level == locking.LEVEL_NODE:
|
9220 |
self._LockInstancesNodes()
|
9221 |
if self.op.disk_template and self.op.remote_node: |
9222 |
self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node) |
9223 |
self.needed_locks[locking.LEVEL_NODE].append(self.op.remote_node) |
9224 |
elif level == locking.LEVEL_NODE_RES and self.op.disk_template: |
9225 |
# Copy node locks
|
9226 |
self.needed_locks[locking.LEVEL_NODE_RES] = \
|
9227 |
_CopyLockList(self.needed_locks[locking.LEVEL_NODE])
|
9228 |
|
9229 |
def BuildHooksEnv(self): |
9230 |
"""Build hooks env.
|
9231 |
|
9232 |
This runs on the master, primary and secondaries.
|
9233 |
|
9234 |
"""
|
9235 |
args = {} |
9236 |
if constants.BE_MINMEM in self.be_new: |
9237 |
args["minmem"] = self.be_new[constants.BE_MINMEM] |
9238 |
if constants.BE_MAXMEM in self.be_new: |
9239 |
args["maxmem"] = self.be_new[constants.BE_MAXMEM] |
9240 |
if constants.BE_VCPUS in self.be_new: |
9241 |
args["vcpus"] = self.be_new[constants.BE_VCPUS] |
9242 |
# TODO: export disk changes. Note: _BuildInstanceHookEnv* don't export disk
|
9243 |
# information at all.
|
9244 |
|
9245 |
if self._new_nics is not None: |
9246 |
nics = [] |
9247 |
|
9248 |
for nic in self._new_nics: |
9249 |
n = copy.deepcopy(nic) |
9250 |
nicparams = self.cluster.SimpleFillNIC(n.nicparams)
|
9251 |
n.nicparams = nicparams |
9252 |
nics.append(_NICToTuple(self, n))
|
9253 |
|
9254 |
args["nics"] = nics
|
9255 |
|
9256 |
env = _BuildInstanceHookEnvByObject(self, self.instance, override=args) |
9257 |
if self.op.disk_template: |
9258 |
env["NEW_DISK_TEMPLATE"] = self.op.disk_template |
9259 |
if self.op.runtime_mem: |
9260 |
env["RUNTIME_MEMORY"] = self.op.runtime_mem |
9261 |
|
9262 |
return env
|
9263 |
|
9264 |
def BuildHooksNodes(self): |
9265 |
"""Build hooks nodes.
|
9266 |
|
9267 |
"""
|
9268 |
nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes) |
9269 |
return (nl, nl)
|
9270 |
|
9271 |
def _PrepareNicModification(self, params, private, old_ip, old_net_uuid, |
9272 |
old_params, cluster, pnode): |
9273 |
|
9274 |
update_params_dict = dict([(key, params[key])
|
9275 |
for key in constants.NICS_PARAMETERS |
9276 |
if key in params]) |
9277 |
|
9278 |
req_link = update_params_dict.get(constants.NIC_LINK, None)
|
9279 |
req_mode = update_params_dict.get(constants.NIC_MODE, None)
|
9280 |
|
9281 |
new_net_uuid = None
|
9282 |
new_net_uuid_or_name = params.get(constants.INIC_NETWORK, old_net_uuid) |
9283 |
if new_net_uuid_or_name:
|
9284 |
new_net_uuid = self.cfg.LookupNetwork(new_net_uuid_or_name)
|
9285 |
new_net_obj = self.cfg.GetNetwork(new_net_uuid)
|
9286 |
|
9287 |
if old_net_uuid:
|
9288 |
old_net_obj = self.cfg.GetNetwork(old_net_uuid)
|
9289 |
|
9290 |
if new_net_uuid:
|
9291 |
netparams = self.cfg.GetGroupNetParams(new_net_uuid, pnode)
|
9292 |
if not netparams: |
9293 |
raise errors.OpPrereqError("No netparams found for the network" |
9294 |
" %s, probably not connected" %
|
9295 |
new_net_obj.name, errors.ECODE_INVAL) |
9296 |
new_params = dict(netparams)
|
9297 |
else:
|
9298 |
new_params = _GetUpdatedParams(old_params, update_params_dict) |
9299 |
|
9300 |
utils.ForceDictType(new_params, constants.NICS_PARAMETER_TYPES) |
9301 |
|
9302 |
new_filled_params = cluster.SimpleFillNIC(new_params) |
9303 |
objects.NIC.CheckParameterSyntax(new_filled_params) |
9304 |
|
9305 |
new_mode = new_filled_params[constants.NIC_MODE] |
9306 |
if new_mode == constants.NIC_MODE_BRIDGED:
|
9307 |
bridge = new_filled_params[constants.NIC_LINK] |
9308 |
msg = self.rpc.call_bridges_exist(pnode, [bridge]).fail_msg
|
9309 |
if msg:
|
9310 |
msg = "Error checking bridges on node '%s': %s" % (pnode, msg)
|
9311 |
if self.op.force: |
9312 |
self.warn.append(msg)
|
9313 |
else:
|
9314 |
raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
|
9315 |
|
9316 |
elif new_mode == constants.NIC_MODE_ROUTED:
|
9317 |
ip = params.get(constants.INIC_IP, old_ip) |
9318 |
if ip is None: |
9319 |
raise errors.OpPrereqError("Cannot set the NIC IP address to None" |
9320 |
" on a routed NIC", errors.ECODE_INVAL)
|
9321 |
|
9322 |
elif new_mode == constants.NIC_MODE_OVS:
|
9323 |
# TODO: check OVS link
|
9324 |
self.LogInfo("OVS links are currently not checked for correctness") |
9325 |
|
9326 |
if constants.INIC_MAC in params: |
9327 |
mac = params[constants.INIC_MAC] |
9328 |
if mac is None: |
9329 |
raise errors.OpPrereqError("Cannot unset the NIC MAC address", |
9330 |
errors.ECODE_INVAL) |
9331 |
elif mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE): |
9332 |
# otherwise generate the MAC address
|
9333 |
params[constants.INIC_MAC] = \ |
9334 |
self.cfg.GenerateMAC(new_net_uuid, self.proc.GetECId()) |
9335 |
else:
|
9336 |
# or validate/reserve the current one
|
9337 |
try:
|
9338 |
self.cfg.ReserveMAC(mac, self.proc.GetECId()) |
9339 |
except errors.ReservationError:
|
9340 |
raise errors.OpPrereqError("MAC address '%s' already in use" |
9341 |
" in cluster" % mac,
|
9342 |
errors.ECODE_NOTUNIQUE) |
9343 |
elif new_net_uuid != old_net_uuid:
|
9344 |
|
9345 |
def get_net_prefix(net_uuid): |
9346 |
mac_prefix = None
|
9347 |
if net_uuid:
|
9348 |
nobj = self.cfg.GetNetwork(net_uuid)
|
9349 |
mac_prefix = nobj.mac_prefix |
9350 |
|
9351 |
return mac_prefix
|
9352 |
|
9353 |
new_prefix = get_net_prefix(new_net_uuid) |
9354 |
old_prefix = get_net_prefix(old_net_uuid) |
9355 |
if old_prefix != new_prefix:
|
9356 |
params[constants.INIC_MAC] = \ |
9357 |
self.cfg.GenerateMAC(new_net_uuid, self.proc.GetECId()) |
9358 |
|
9359 |
# if there is a change in (ip, network) tuple
|
9360 |
new_ip = params.get(constants.INIC_IP, old_ip) |
9361 |
if (new_ip, new_net_uuid) != (old_ip, old_net_uuid):
|
9362 |
if new_ip:
|
9363 |
# if IP is pool then require a network and generate one IP
|
9364 |
if new_ip.lower() == constants.NIC_IP_POOL:
|
9365 |
if new_net_uuid:
|
9366 |
try:
|
9367 |
new_ip = self.cfg.GenerateIp(new_net_uuid, self.proc.GetECId()) |
9368 |
except errors.ReservationError:
|
9369 |
raise errors.OpPrereqError("Unable to get a free IP" |
9370 |
" from the address pool",
|
9371 |
errors.ECODE_STATE) |
9372 |
self.LogInfo("Chose IP %s from network %s", |
9373 |
new_ip, |
9374 |
new_net_obj.name) |
9375 |
params[constants.INIC_IP] = new_ip |
9376 |
else:
|
9377 |
raise errors.OpPrereqError("ip=pool, but no network found", |
9378 |
errors.ECODE_INVAL) |
9379 |
# Reserve new IP if in the new network if any
|
9380 |
elif new_net_uuid:
|
9381 |
try:
|
9382 |
self.cfg.ReserveIp(new_net_uuid, new_ip, self.proc.GetECId()) |
9383 |
self.LogInfo("Reserving IP %s in network %s", |
9384 |
new_ip, new_net_obj.name) |
9385 |
except errors.ReservationError:
|
9386 |
raise errors.OpPrereqError("IP %s not available in network %s" % |
9387 |
(new_ip, new_net_obj.name), |
9388 |
errors.ECODE_NOTUNIQUE) |
9389 |
# new network is None so check if new IP is a conflicting IP
|
9390 |
elif self.op.conflicts_check: |
9391 |
_CheckForConflictingIp(self, new_ip, pnode)
|
9392 |
|
9393 |
# release old IP if old network is not None
|
9394 |
if old_ip and old_net_uuid: |
9395 |
try:
|
9396 |
self.cfg.ReleaseIp(old_net_uuid, old_ip, self.proc.GetECId()) |
9397 |
except errors.AddressPoolError:
|
9398 |
logging.warning("Release IP %s not contained in network %s",
|
9399 |
old_ip, old_net_obj.name) |
9400 |
|
9401 |
# there are no changes in (ip, network) tuple and old network is not None
|
9402 |
elif (old_net_uuid is not None and |
9403 |
(req_link is not None or req_mode is not None)): |
9404 |
raise errors.OpPrereqError("Not allowed to change link or mode of" |
9405 |
" a NIC that is connected to a network",
|
9406 |
errors.ECODE_INVAL) |
9407 |
|
9408 |
private.params = new_params |
9409 |
private.filled = new_filled_params |
9410 |
|
9411 |
def _PreCheckDiskTemplate(self, pnode_info): |
9412 |
"""CheckPrereq checks related to a new disk template."""
|
9413 |
# Arguments are passed to avoid configuration lookups
|
9414 |
instance = self.instance
|
9415 |
pnode = instance.primary_node |
9416 |
cluster = self.cluster
|
9417 |
if instance.disk_template == self.op.disk_template: |
9418 |
raise errors.OpPrereqError("Instance already has disk template %s" % |
9419 |
instance.disk_template, errors.ECODE_INVAL) |
9420 |
|
9421 |
if (instance.disk_template,
|
9422 |
self.op.disk_template) not in self._DISK_CONVERSIONS: |
9423 |
raise errors.OpPrereqError("Unsupported disk template conversion from" |
9424 |
" %s to %s" % (instance.disk_template,
|
9425 |
self.op.disk_template),
|
9426 |
errors.ECODE_INVAL) |
9427 |
_CheckInstanceState(self, instance, INSTANCE_DOWN,
|
9428 |
msg="cannot change disk template")
|
9429 |
if self.op.disk_template in constants.DTS_INT_MIRROR: |
9430 |
if self.op.remote_node == pnode: |
9431 |
raise errors.OpPrereqError("Given new secondary node %s is the same" |
9432 |
" as the primary node of the instance" %
|
9433 |
self.op.remote_node, errors.ECODE_STATE)
|
9434 |
_CheckNodeOnline(self, self.op.remote_node) |
9435 |
_CheckNodeNotDrained(self, self.op.remote_node) |
9436 |
# FIXME: here we assume that the old instance type is DT_PLAIN
|
9437 |
assert instance.disk_template == constants.DT_PLAIN
|
9438 |
disks = [{constants.IDISK_SIZE: d.size, |
9439 |
constants.IDISK_VG: d.logical_id[0]}
|
9440 |
for d in instance.disks] |
9441 |
required = _ComputeDiskSizePerVG(self.op.disk_template, disks)
|
9442 |
_CheckNodesFreeDiskPerVG(self, [self.op.remote_node], required) |
9443 |
|
9444 |
snode_info = self.cfg.GetNodeInfo(self.op.remote_node) |
9445 |
snode_group = self.cfg.GetNodeGroup(snode_info.group)
|
9446 |
ipolicy = ganeti.masterd.instance.CalculateGroupIPolicy(cluster, |
9447 |
snode_group) |
9448 |
_CheckTargetNodeIPolicy(self, ipolicy, instance, snode_info, self.cfg, |
9449 |
ignore=self.op.ignore_ipolicy)
|
9450 |
if pnode_info.group != snode_info.group:
|
9451 |
self.LogWarning("The primary and secondary nodes are in two" |
9452 |
" different node groups; the disk parameters"
|
9453 |
" from the first disk's node group will be"
|
9454 |
" used")
|
9455 |
|
9456 |
if not self.op.disk_template in constants.DTS_EXCL_STORAGE: |
9457 |
# Make sure none of the nodes require exclusive storage
|
9458 |
nodes = [pnode_info] |
9459 |
if self.op.disk_template in constants.DTS_INT_MIRROR: |
9460 |
assert snode_info
|
9461 |
nodes.append(snode_info) |
9462 |
has_es = lambda n: _IsExclusiveStorageEnabledNode(self.cfg, n) |
9463 |
if compat.any(map(has_es, nodes)): |
9464 |
errmsg = ("Cannot convert disk template from %s to %s when exclusive"
|
9465 |
" storage is enabled" % (instance.disk_template,
|
9466 |
self.op.disk_template))
|
9467 |
raise errors.OpPrereqError(errmsg, errors.ECODE_STATE)
|
9468 |
|
9469 |
def CheckPrereq(self): |
9470 |
"""Check prerequisites.
|
9471 |
|
9472 |
This only checks the instance list against the existing names.
|
9473 |
|
9474 |
"""
|
9475 |
assert self.op.instance_name in self.owned_locks(locking.LEVEL_INSTANCE) |
9476 |
instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name) |
9477 |
|
9478 |
cluster = self.cluster = self.cfg.GetClusterInfo() |
9479 |
assert self.instance is not None, \ |
9480 |
"Cannot retrieve locked instance %s" % self.op.instance_name |
9481 |
|
9482 |
pnode = instance.primary_node |
9483 |
|
9484 |
self.warn = []
|
9485 |
|
9486 |
if (self.op.pnode is not None and self.op.pnode != pnode and |
9487 |
not self.op.force): |
9488 |
# verify that the instance is not up
|
9489 |
instance_info = self.rpc.call_instance_info(pnode, instance.name,
|
9490 |
instance.hypervisor) |
9491 |
if instance_info.fail_msg:
|
9492 |
self.warn.append("Can't get instance runtime information: %s" % |
9493 |
instance_info.fail_msg) |
9494 |
elif instance_info.payload:
|
9495 |
raise errors.OpPrereqError("Instance is still running on %s" % pnode, |
9496 |
errors.ECODE_STATE) |
9497 |
|
9498 |
assert pnode in self.owned_locks(locking.LEVEL_NODE) |
9499 |
nodelist = list(instance.all_nodes)
|
9500 |
pnode_info = self.cfg.GetNodeInfo(pnode)
|
9501 |
self.diskparams = self.cfg.GetInstanceDiskParams(instance) |
9502 |
|
9503 |
#_CheckInstanceNodeGroups(self.cfg, self.op.instance_name, owned_groups)
|
9504 |
assert pnode_info.group in self.owned_locks(locking.LEVEL_NODEGROUP) |
9505 |
group_info = self.cfg.GetNodeGroup(pnode_info.group)
|
9506 |
|
9507 |
# dictionary with instance information after the modification
|
9508 |
ispec = {} |
9509 |
|
9510 |
# Check disk modifications. This is done here and not in CheckArguments
|
9511 |
# (as with NICs), because we need to know the instance's disk template
|
9512 |
if instance.disk_template == constants.DT_EXT:
|
9513 |
self._CheckMods("disk", self.op.disks, {}, |
9514 |
self._VerifyDiskModification)
|
9515 |
else:
|
9516 |
self._CheckMods("disk", self.op.disks, constants.IDISK_PARAMS_TYPES, |
9517 |
self._VerifyDiskModification)
|
9518 |
|
9519 |
# Prepare disk/NIC modifications
|
9520 |
self.diskmod = PrepareContainerMods(self.op.disks, None) |
9521 |
self.nicmod = PrepareContainerMods(self.op.nics, _InstNicModPrivate) |
9522 |
|
9523 |
# Check the validity of the `provider' parameter
|
9524 |
if instance.disk_template in constants.DT_EXT: |
9525 |
for mod in self.diskmod: |
9526 |
ext_provider = mod[2].get(constants.IDISK_PROVIDER, None) |
9527 |
if mod[0] == constants.DDM_ADD: |
9528 |
if ext_provider is None: |
9529 |
raise errors.OpPrereqError("Instance template is '%s' and parameter" |
9530 |
" '%s' missing, during disk add" %
|
9531 |
(constants.DT_EXT, |
9532 |
constants.IDISK_PROVIDER), |
9533 |
errors.ECODE_NOENT) |
9534 |
elif mod[0] == constants.DDM_MODIFY: |
9535 |
if ext_provider:
|
9536 |
raise errors.OpPrereqError("Parameter '%s' is invalid during disk" |
9537 |
" modification" %
|
9538 |
constants.IDISK_PROVIDER, |
9539 |
errors.ECODE_INVAL) |
9540 |
else:
|
9541 |
for mod in self.diskmod: |
9542 |
ext_provider = mod[2].get(constants.IDISK_PROVIDER, None) |
9543 |
if ext_provider is not None: |
9544 |
raise errors.OpPrereqError("Parameter '%s' is only valid for" |
9545 |
" instances of type '%s'" %
|
9546 |
(constants.IDISK_PROVIDER, |
9547 |
constants.DT_EXT), |
9548 |
errors.ECODE_INVAL) |
9549 |
|
9550 |
# OS change
|
9551 |
if self.op.os_name and not self.op.force: |
9552 |
_CheckNodeHasOS(self, instance.primary_node, self.op.os_name, |
9553 |
self.op.force_variant)
|
9554 |
instance_os = self.op.os_name
|
9555 |
else:
|
9556 |
instance_os = instance.os |
9557 |
|
9558 |
assert not (self.op.disk_template and self.op.disks), \ |
9559 |
"Can't modify disk template and apply disk changes at the same time"
|
9560 |
|
9561 |
if self.op.disk_template: |
9562 |
self._PreCheckDiskTemplate(pnode_info)
|
9563 |
|
9564 |
# hvparams processing
|
9565 |
if self.op.hvparams: |
9566 |
hv_type = instance.hypervisor |
9567 |
i_hvdict = _GetUpdatedParams(instance.hvparams, self.op.hvparams)
|
9568 |
utils.ForceDictType(i_hvdict, constants.HVS_PARAMETER_TYPES) |
9569 |
hv_new = cluster.SimpleFillHV(hv_type, instance.os, i_hvdict) |
9570 |
|
9571 |
# local check
|
9572 |
hypervisor.GetHypervisorClass(hv_type).CheckParameterSyntax(hv_new) |
9573 |
_CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
|
9574 |
self.hv_proposed = self.hv_new = hv_new # the new actual values |
9575 |
self.hv_inst = i_hvdict # the new dict (without defaults) |
9576 |
else:
|
9577 |
self.hv_proposed = cluster.SimpleFillHV(instance.hypervisor, instance.os,
|
9578 |
instance.hvparams) |
9579 |
self.hv_new = self.hv_inst = {} |
9580 |
|
9581 |
# beparams processing
|
9582 |
if self.op.beparams: |
9583 |
i_bedict = _GetUpdatedParams(instance.beparams, self.op.beparams,
|
9584 |
use_none=True)
|
9585 |
objects.UpgradeBeParams(i_bedict) |
9586 |
utils.ForceDictType(i_bedict, constants.BES_PARAMETER_TYPES) |
9587 |
be_new = cluster.SimpleFillBE(i_bedict) |
9588 |
self.be_proposed = self.be_new = be_new # the new actual values |
9589 |
self.be_inst = i_bedict # the new dict (without defaults) |
9590 |
else:
|
9591 |
self.be_new = self.be_inst = {} |
9592 |
self.be_proposed = cluster.SimpleFillBE(instance.beparams)
|
9593 |
be_old = cluster.FillBE(instance) |
9594 |
|
9595 |
# CPU param validation -- checking every time a parameter is
|
9596 |
# changed to cover all cases where either CPU mask or vcpus have
|
9597 |
# changed
|
9598 |
if (constants.BE_VCPUS in self.be_proposed and |
9599 |
constants.HV_CPU_MASK in self.hv_proposed): |
9600 |
cpu_list = \ |
9601 |
utils.ParseMultiCpuMask(self.hv_proposed[constants.HV_CPU_MASK])
|
9602 |
# Verify mask is consistent with number of vCPUs. Can skip this
|
9603 |
# test if only 1 entry in the CPU mask, which means same mask
|
9604 |
# is applied to all vCPUs.
|
9605 |
if (len(cpu_list) > 1 and |
9606 |
len(cpu_list) != self.be_proposed[constants.BE_VCPUS]): |
9607 |
raise errors.OpPrereqError("Number of vCPUs [%d] does not match the" |
9608 |
" CPU mask [%s]" %
|
9609 |
(self.be_proposed[constants.BE_VCPUS],
|
9610 |
self.hv_proposed[constants.HV_CPU_MASK]),
|
9611 |
errors.ECODE_INVAL) |
9612 |
|
9613 |
# Only perform this test if a new CPU mask is given
|
9614 |
if constants.HV_CPU_MASK in self.hv_new: |
9615 |
# Calculate the largest CPU number requested
|
9616 |
max_requested_cpu = max(map(max, cpu_list)) |
9617 |
# Check that all of the instance's nodes have enough physical CPUs to
|
9618 |
# satisfy the requested CPU mask
|
9619 |
_CheckNodesPhysicalCPUs(self, instance.all_nodes,
|
9620 |
max_requested_cpu + 1, instance.hypervisor)
|
9621 |
|
9622 |
# osparams processing
|
9623 |
if self.op.osparams: |
9624 |
i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
|
9625 |
_CheckOSParams(self, True, nodelist, instance_os, i_osdict) |
9626 |
self.os_inst = i_osdict # the new dict (without defaults) |
9627 |
else:
|
9628 |
self.os_inst = {}
|
9629 |
|
9630 |
#TODO(dynmem): do the appropriate check involving MINMEM
|
9631 |
if (constants.BE_MAXMEM in self.op.beparams and not self.op.force and |
9632 |
be_new[constants.BE_MAXMEM] > be_old[constants.BE_MAXMEM]): |
9633 |
mem_check_list = [pnode] |
9634 |
if be_new[constants.BE_AUTO_BALANCE]:
|
9635 |
# either we changed auto_balance to yes or it was from before
|
9636 |
mem_check_list.extend(instance.secondary_nodes) |
9637 |
instance_info = self.rpc.call_instance_info(pnode, instance.name,
|
9638 |
instance.hypervisor) |
9639 |
nodeinfo = self.rpc.call_node_info(mem_check_list, None, |
9640 |
[instance.hypervisor], False)
|
9641 |
pninfo = nodeinfo[pnode] |
9642 |
msg = pninfo.fail_msg |
9643 |
if msg:
|
9644 |
# Assume the primary node is unreachable and go ahead
|
9645 |
self.warn.append("Can't get info from primary node %s: %s" % |
9646 |
(pnode, msg)) |
9647 |
else:
|
9648 |
(_, _, (pnhvinfo, )) = pninfo.payload |
9649 |
if not isinstance(pnhvinfo.get("memory_free", None), int): |
9650 |
self.warn.append("Node data from primary node %s doesn't contain" |
9651 |
" free memory information" % pnode)
|
9652 |
elif instance_info.fail_msg:
|
9653 |
self.warn.append("Can't get instance runtime information: %s" % |
9654 |
instance_info.fail_msg) |
9655 |
else:
|
9656 |
if instance_info.payload:
|
9657 |
current_mem = int(instance_info.payload["memory"]) |
9658 |
else:
|
9659 |
# Assume instance not running
|
9660 |
# (there is a slight race condition here, but it's not very
|
9661 |
# probable, and we have no other way to check)
|
9662 |
# TODO: Describe race condition
|
9663 |
current_mem = 0
|
9664 |
#TODO(dynmem): do the appropriate check involving MINMEM
|
9665 |
miss_mem = (be_new[constants.BE_MAXMEM] - current_mem - |
9666 |
pnhvinfo["memory_free"])
|
9667 |
if miss_mem > 0: |
9668 |
raise errors.OpPrereqError("This change will prevent the instance" |
9669 |
" from starting, due to %d MB of memory"
|
9670 |
" missing on its primary node" %
|
9671 |
miss_mem, errors.ECODE_NORES) |
9672 |
|
9673 |
if be_new[constants.BE_AUTO_BALANCE]:
|
9674 |
for node, nres in nodeinfo.items(): |
9675 |
if node not in instance.secondary_nodes: |
9676 |
continue
|
9677 |
nres.Raise("Can't get info from secondary node %s" % node,
|
9678 |
prereq=True, ecode=errors.ECODE_STATE)
|
9679 |
(_, _, (nhvinfo, )) = nres.payload |
9680 |
if not isinstance(nhvinfo.get("memory_free", None), int): |
9681 |
raise errors.OpPrereqError("Secondary node %s didn't return free" |
9682 |
" memory information" % node,
|
9683 |
errors.ECODE_STATE) |
9684 |
#TODO(dynmem): do the appropriate check involving MINMEM
|
9685 |
elif be_new[constants.BE_MAXMEM] > nhvinfo["memory_free"]: |
9686 |
raise errors.OpPrereqError("This change will prevent the instance" |
9687 |
" from failover to its secondary node"
|
9688 |
" %s, due to not enough memory" % node,
|
9689 |
errors.ECODE_STATE) |
9690 |
|
9691 |
if self.op.runtime_mem: |
9692 |
remote_info = self.rpc.call_instance_info(instance.primary_node,
|
9693 |
instance.name, |
9694 |
instance.hypervisor) |
9695 |
remote_info.Raise("Error checking node %s" % instance.primary_node)
|
9696 |
if not remote_info.payload: # not running already |
9697 |
raise errors.OpPrereqError("Instance %s is not running" % |
9698 |
instance.name, errors.ECODE_STATE) |
9699 |
|
9700 |
current_memory = remote_info.payload["memory"]
|
9701 |
if (not self.op.force and |
9702 |
(self.op.runtime_mem > self.be_proposed[constants.BE_MAXMEM] or |
9703 |
self.op.runtime_mem < self.be_proposed[constants.BE_MINMEM])): |
9704 |
raise errors.OpPrereqError("Instance %s must have memory between %d" |
9705 |
" and %d MB of memory unless --force is"
|
9706 |
" given" %
|
9707 |
(instance.name, |
9708 |
self.be_proposed[constants.BE_MINMEM],
|
9709 |
self.be_proposed[constants.BE_MAXMEM]),
|
9710 |
errors.ECODE_INVAL) |
9711 |
|
9712 |
delta = self.op.runtime_mem - current_memory
|
9713 |
if delta > 0: |
9714 |
_CheckNodeFreeMemory(self, instance.primary_node,
|
9715 |
"ballooning memory for instance %s" %
|
9716 |
instance.name, delta, instance.hypervisor) |
9717 |
|
9718 |
if self.op.disks and instance.disk_template == constants.DT_DISKLESS: |
9719 |
raise errors.OpPrereqError("Disk operations not supported for" |
9720 |
" diskless instances", errors.ECODE_INVAL)
|
9721 |
|
9722 |
def _PrepareNicCreate(_, params, private): |
9723 |
self._PrepareNicModification(params, private, None, None, |
9724 |
{}, cluster, pnode) |
9725 |
return (None, None) |
9726 |
|
9727 |
def _PrepareNicMod(_, nic, params, private): |
9728 |
self._PrepareNicModification(params, private, nic.ip, nic.network,
|
9729 |
nic.nicparams, cluster, pnode) |
9730 |
return None |
9731 |
|
9732 |
def _PrepareNicRemove(_, params, __): |
9733 |
ip = params.ip |
9734 |
net = params.network |
9735 |
if net is not None and ip is not None: |
9736 |
self.cfg.ReleaseIp(net, ip, self.proc.GetECId()) |
9737 |
|
9738 |
# Verify NIC changes (operating on copy)
|
9739 |
nics = instance.nics[:] |
9740 |
ApplyContainerMods("NIC", nics, None, self.nicmod, |
9741 |
_PrepareNicCreate, _PrepareNicMod, _PrepareNicRemove) |
9742 |
if len(nics) > constants.MAX_NICS: |
9743 |
raise errors.OpPrereqError("Instance has too many network interfaces" |
9744 |
" (%d), cannot add more" % constants.MAX_NICS,
|
9745 |
errors.ECODE_STATE) |
9746 |
|
9747 |
def _PrepareDiskMod(_, disk, params, __): |
9748 |
disk.name = params.get(constants.IDISK_NAME, None)
|
9749 |
|
9750 |
# Verify disk changes (operating on a copy)
|
9751 |
disks = copy.deepcopy(instance.disks) |
9752 |
ApplyContainerMods("disk", disks, None, self.diskmod, None, _PrepareDiskMod, |
9753 |
None)
|
9754 |
utils.ValidateDeviceNames("disk", disks)
|
9755 |
if len(disks) > constants.MAX_DISKS: |
9756 |
raise errors.OpPrereqError("Instance has too many disks (%d), cannot add" |
9757 |
" more" % constants.MAX_DISKS,
|
9758 |
errors.ECODE_STATE) |
9759 |
disk_sizes = [disk.size for disk in instance.disks] |
9760 |
disk_sizes.extend(params["size"] for (op, idx, params, private) in |
9761 |
self.diskmod if op == constants.DDM_ADD) |
9762 |
ispec[constants.ISPEC_DISK_COUNT] = len(disk_sizes)
|
9763 |
ispec[constants.ISPEC_DISK_SIZE] = disk_sizes |
9764 |
|
9765 |
if self.op.offline is not None and self.op.offline: |
9766 |
_CheckInstanceState(self, instance, CAN_CHANGE_INSTANCE_OFFLINE,
|
9767 |
msg="can't change to offline")
|
9768 |
|
9769 |
# Pre-compute NIC changes (necessary to use result in hooks)
|
9770 |
self._nic_chgdesc = []
|
9771 |
if self.nicmod: |
9772 |
# Operate on copies as this is still in prereq
|
9773 |
nics = [nic.Copy() for nic in instance.nics] |
9774 |
ApplyContainerMods("NIC", nics, self._nic_chgdesc, self.nicmod, |
9775 |
self._CreateNewNic, self._ApplyNicMods, None) |
9776 |
# Verify that NIC names are unique and valid
|
9777 |
utils.ValidateDeviceNames("NIC", nics)
|
9778 |
self._new_nics = nics
|
9779 |
ispec[constants.ISPEC_NIC_COUNT] = len(self._new_nics) |
9780 |
else:
|
9781 |
self._new_nics = None |
9782 |
ispec[constants.ISPEC_NIC_COUNT] = len(instance.nics)
|
9783 |
|
9784 |
if not self.op.ignore_ipolicy: |
9785 |
ipolicy = ganeti.masterd.instance.CalculateGroupIPolicy(cluster, |
9786 |
group_info) |
9787 |
|
9788 |
# Fill ispec with backend parameters
|
9789 |
ispec[constants.ISPEC_SPINDLE_USE] = \ |
9790 |
self.be_new.get(constants.BE_SPINDLE_USE, None) |
9791 |
ispec[constants.ISPEC_CPU_COUNT] = self.be_new.get(constants.BE_VCPUS,
|
9792 |
None)
|
9793 |
|
9794 |
# Copy ispec to verify parameters with min/max values separately
|
9795 |
if self.op.disk_template: |
9796 |
new_disk_template = self.op.disk_template
|
9797 |
else:
|
9798 |
new_disk_template = instance.disk_template |
9799 |
ispec_max = ispec.copy() |
9800 |
ispec_max[constants.ISPEC_MEM_SIZE] = \ |
9801 |
self.be_new.get(constants.BE_MAXMEM, None) |
9802 |
res_max = _ComputeIPolicyInstanceSpecViolation(ipolicy, ispec_max, |
9803 |
new_disk_template) |
9804 |
ispec_min = ispec.copy() |
9805 |
ispec_min[constants.ISPEC_MEM_SIZE] = \ |
9806 |
self.be_new.get(constants.BE_MINMEM, None) |
9807 |
res_min = _ComputeIPolicyInstanceSpecViolation(ipolicy, ispec_min, |
9808 |
new_disk_template) |
9809 |
|
9810 |
if (res_max or res_min): |
9811 |
# FIXME: Improve error message by including information about whether
|
9812 |
# the upper or lower limit of the parameter fails the ipolicy.
|
9813 |
msg = ("Instance allocation to group %s (%s) violates policy: %s" %
|
9814 |
(group_info, group_info.name, |
9815 |
utils.CommaJoin(set(res_max + res_min))))
|
9816 |
raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
|
9817 |
|
9818 |
def _ConvertPlainToDrbd(self, feedback_fn): |
9819 |
"""Converts an instance from plain to drbd.
|
9820 |
|
9821 |
"""
|
9822 |
feedback_fn("Converting template to drbd")
|
9823 |
instance = self.instance
|
9824 |
pnode = instance.primary_node |
9825 |
snode = self.op.remote_node
|
9826 |
|
9827 |
assert instance.disk_template == constants.DT_PLAIN
|
9828 |
|
9829 |
# create a fake disk info for _GenerateDiskTemplate
|
9830 |
disk_info = [{constants.IDISK_SIZE: d.size, constants.IDISK_MODE: d.mode, |
9831 |
constants.IDISK_VG: d.logical_id[0],
|
9832 |
constants.IDISK_NAME: d.name} |
9833 |
for d in instance.disks] |
9834 |
new_disks = _GenerateDiskTemplate(self, self.op.disk_template, |
9835 |
instance.name, pnode, [snode], |
9836 |
disk_info, None, None, 0, feedback_fn, |
9837 |
self.diskparams)
|
9838 |
anno_disks = rpc.AnnotateDiskParams(constants.DT_DRBD8, new_disks, |
9839 |
self.diskparams)
|
9840 |
p_excl_stor = _IsExclusiveStorageEnabledNodeName(self.cfg, pnode)
|
9841 |
s_excl_stor = _IsExclusiveStorageEnabledNodeName(self.cfg, snode)
|
9842 |
info = _GetInstanceInfoText(instance) |
9843 |
feedback_fn("Creating additional volumes...")
|
9844 |
# first, create the missing data and meta devices
|
9845 |
for disk in anno_disks: |
9846 |
# unfortunately this is... not too nice
|
9847 |
_CreateSingleBlockDev(self, pnode, instance, disk.children[1], |
9848 |
info, True, p_excl_stor)
|
9849 |
for child in disk.children: |
9850 |
_CreateSingleBlockDev(self, snode, instance, child, info, True, |
9851 |
s_excl_stor) |
9852 |
# at this stage, all new LVs have been created, we can rename the
|
9853 |
# old ones
|
9854 |
feedback_fn("Renaming original volumes...")
|
9855 |
rename_list = [(o, n.children[0].logical_id)
|
9856 |
for (o, n) in zip(instance.disks, new_disks)] |
9857 |
result = self.rpc.call_blockdev_rename(pnode, rename_list)
|
9858 |
result.Raise("Failed to rename original LVs")
|
9859 |
|
9860 |
feedback_fn("Initializing DRBD devices...")
|
9861 |
# all child devices are in place, we can now create the DRBD devices
|
9862 |
try:
|
9863 |
for disk in anno_disks: |
9864 |
for (node, excl_stor) in [(pnode, p_excl_stor), (snode, s_excl_stor)]: |
9865 |
f_create = node == pnode |
9866 |
_CreateSingleBlockDev(self, node, instance, disk, info, f_create,
|
9867 |
excl_stor) |
9868 |
except errors.GenericError, e:
|
9869 |
feedback_fn("Initializing of DRBD devices failed;"
|
9870 |
" renaming back original volumes...")
|
9871 |
for disk in new_disks: |
9872 |
self.cfg.SetDiskID(disk, pnode)
|
9873 |
rename_back_list = [(n.children[0], o.logical_id)
|
9874 |
for (n, o) in zip(new_disks, instance.disks)] |
9875 |
result = self.rpc.call_blockdev_rename(pnode, rename_back_list)
|
9876 |
result.Raise("Failed to rename LVs back after error %s" % str(e)) |
9877 |
raise
|
9878 |
|
9879 |
# at this point, the instance has been modified
|
9880 |
instance.disk_template = constants.DT_DRBD8 |
9881 |
instance.disks = new_disks |
9882 |
self.cfg.Update(instance, feedback_fn)
|
9883 |
|
9884 |
# Release node locks while waiting for sync
|
9885 |
_ReleaseLocks(self, locking.LEVEL_NODE)
|
9886 |
|
9887 |
# disks are created, waiting for sync
|
9888 |
disk_abort = not _WaitForSync(self, instance, |
9889 |
oneshot=not self.op.wait_for_sync) |
9890 |
if disk_abort:
|
9891 |
raise errors.OpExecError("There are some degraded disks for" |
9892 |
" this instance, please cleanup manually")
|
9893 |
|
9894 |
# Node resource locks will be released by caller
|
9895 |
|
9896 |
def _ConvertDrbdToPlain(self, feedback_fn): |
9897 |
"""Converts an instance from drbd to plain.
|
9898 |
|
9899 |
"""
|
9900 |
instance = self.instance
|
9901 |
|
9902 |
assert len(instance.secondary_nodes) == 1 |
9903 |
assert instance.disk_template == constants.DT_DRBD8
|
9904 |
|
9905 |
pnode = instance.primary_node |
9906 |
snode = instance.secondary_nodes[0]
|
9907 |
feedback_fn("Converting template to plain")
|
9908 |
|
9909 |
old_disks = _AnnotateDiskParams(instance, instance.disks, self.cfg)
|
9910 |
new_disks = [d.children[0] for d in instance.disks] |
9911 |
|
9912 |
# copy over size, mode and name
|
9913 |
for parent, child in zip(old_disks, new_disks): |
9914 |
child.size = parent.size |
9915 |
child.mode = parent.mode |
9916 |
child.name = parent.name |
9917 |
|
9918 |
# this is a DRBD disk, return its port to the pool
|
9919 |
# NOTE: this must be done right before the call to cfg.Update!
|
9920 |
for disk in old_disks: |
9921 |
tcp_port = disk.logical_id[2]
|
9922 |
self.cfg.AddTcpUdpPort(tcp_port)
|
9923 |
|
9924 |
# update instance structure
|
9925 |
instance.disks = new_disks |
9926 |
instance.disk_template = constants.DT_PLAIN |
9927 |
_UpdateIvNames(0, instance.disks)
|
9928 |
self.cfg.Update(instance, feedback_fn)
|
9929 |
|
9930 |
# Release locks in case removing disks takes a while
|
9931 |
_ReleaseLocks(self, locking.LEVEL_NODE)
|
9932 |
|
9933 |
feedback_fn("Removing volumes on the secondary node...")
|
9934 |
for disk in old_disks: |
9935 |
self.cfg.SetDiskID(disk, snode)
|
9936 |
msg = self.rpc.call_blockdev_remove(snode, disk).fail_msg
|
9937 |
if msg:
|
9938 |
self.LogWarning("Could not remove block device %s on node %s," |
9939 |
" continuing anyway: %s", disk.iv_name, snode, msg)
|
9940 |
|
9941 |
feedback_fn("Removing unneeded volumes on the primary node...")
|
9942 |
for idx, disk in enumerate(old_disks): |
9943 |
meta = disk.children[1]
|
9944 |
self.cfg.SetDiskID(meta, pnode)
|
9945 |
msg = self.rpc.call_blockdev_remove(pnode, meta).fail_msg
|
9946 |
if msg:
|
9947 |
self.LogWarning("Could not remove metadata for disk %d on node %s," |
9948 |
" continuing anyway: %s", idx, pnode, msg)
|
9949 |
|
9950 |
def _CreateNewDisk(self, idx, params, _): |
9951 |
"""Creates a new disk.
|
9952 |
|
9953 |
"""
|
9954 |
instance = self.instance
|
9955 |
|
9956 |
# add a new disk
|
9957 |
if instance.disk_template in constants.DTS_FILEBASED: |
9958 |
(file_driver, file_path) = instance.disks[0].logical_id
|
9959 |
file_path = os.path.dirname(file_path) |
9960 |
else:
|
9961 |
file_driver = file_path = None
|
9962 |
|
9963 |
disk = \ |
9964 |
_GenerateDiskTemplate(self, instance.disk_template, instance.name,
|
9965 |
instance.primary_node, instance.secondary_nodes, |
9966 |
[params], file_path, file_driver, idx, |
9967 |
self.Log, self.diskparams)[0] |
9968 |
|
9969 |
info = _GetInstanceInfoText(instance) |
9970 |
|
9971 |
logging.info("Creating volume %s for instance %s",
|
9972 |
disk.iv_name, instance.name) |
9973 |
# Note: this needs to be kept in sync with _CreateDisks
|
9974 |
#HARDCODE
|
9975 |
for node in instance.all_nodes: |
9976 |
f_create = (node == instance.primary_node) |
9977 |
try:
|
9978 |
_CreateBlockDev(self, node, instance, disk, f_create, info, f_create)
|
9979 |
except errors.OpExecError, err:
|
9980 |
self.LogWarning("Failed to create volume %s (%s) on node '%s': %s", |
9981 |
disk.iv_name, disk, node, err) |
9982 |
|
9983 |
if self.cluster.prealloc_wipe_disks: |
9984 |
# Wipe new disk
|
9985 |
_WipeDisks(self, instance,
|
9986 |
disks=[(idx, disk, 0)])
|
9987 |
|
9988 |
return (disk, [
|
9989 |
("disk/%d" % idx, "add:size=%s,mode=%s" % (disk.size, disk.mode)), |
9990 |
]) |
9991 |
|
9992 |
@staticmethod
|
9993 |
def _ModifyDisk(idx, disk, params, _): |
9994 |
"""Modifies a disk.
|
9995 |
|
9996 |
"""
|
9997 |
changes = [] |
9998 |
mode = params.get(constants.IDISK_MODE, None)
|
9999 |
if mode:
|
10000 |
disk.mode = mode |
10001 |
changes.append(("disk.mode/%d" % idx, disk.mode))
|
10002 |
|
10003 |
name = params.get(constants.IDISK_NAME, None)
|
10004 |
disk.name = name |
10005 |
changes.append(("disk.name/%d" % idx, disk.name))
|
10006 |
|
10007 |
return changes
|
10008 |
|
10009 |
def _RemoveDisk(self, idx, root, _): |
10010 |
"""Removes a disk.
|
10011 |
|
10012 |
"""
|
10013 |
(anno_disk,) = _AnnotateDiskParams(self.instance, [root], self.cfg) |
10014 |
for node, disk in anno_disk.ComputeNodeTree(self.instance.primary_node): |
10015 |
self.cfg.SetDiskID(disk, node)
|
10016 |
msg = self.rpc.call_blockdev_remove(node, disk).fail_msg
|
10017 |
if msg:
|
10018 |
self.LogWarning("Could not remove disk/%d on node '%s': %s," |
10019 |
" continuing anyway", idx, node, msg)
|
10020 |
|
10021 |
# if this is a DRBD disk, return its port to the pool
|
10022 |
if root.dev_type in constants.LDS_DRBD: |
10023 |
self.cfg.AddTcpUdpPort(root.logical_id[2]) |
10024 |
|
10025 |
def _CreateNewNic(self, idx, params, private): |
10026 |
"""Creates data structure for a new network interface.
|
10027 |
|
10028 |
"""
|
10029 |
mac = params[constants.INIC_MAC] |
10030 |
ip = params.get(constants.INIC_IP, None)
|
10031 |
net = params.get(constants.INIC_NETWORK, None)
|
10032 |
name = params.get(constants.INIC_NAME, None)
|
10033 |
net_uuid = self.cfg.LookupNetwork(net)
|
10034 |
#TODO: not private.filled?? can a nic have no nicparams??
|
10035 |
nicparams = private.filled |
10036 |
nobj = objects.NIC(mac=mac, ip=ip, network=net_uuid, name=name, |
10037 |
nicparams=nicparams) |
10038 |
nobj.uuid = self.cfg.GenerateUniqueID(self.proc.GetECId()) |
10039 |
|
10040 |
return (nobj, [
|
10041 |
("nic.%d" % idx,
|
10042 |
"add:mac=%s,ip=%s,mode=%s,link=%s,network=%s" %
|
10043 |
(mac, ip, private.filled[constants.NIC_MODE], |
10044 |
private.filled[constants.NIC_LINK], |
10045 |
net)), |
10046 |
]) |
10047 |
|
10048 |
def _ApplyNicMods(self, idx, nic, params, private): |
10049 |
"""Modifies a network interface.
|
10050 |
|
10051 |
"""
|
10052 |
changes = [] |
10053 |
|
10054 |
for key in [constants.INIC_MAC, constants.INIC_IP, constants.INIC_NAME]: |
10055 |
if key in params: |
10056 |
changes.append(("nic.%s/%d" % (key, idx), params[key]))
|
10057 |
setattr(nic, key, params[key])
|
10058 |
|
10059 |
new_net = params.get(constants.INIC_NETWORK, nic.network) |
10060 |
new_net_uuid = self.cfg.LookupNetwork(new_net)
|
10061 |
if new_net_uuid != nic.network:
|
10062 |
changes.append(("nic.network/%d" % idx, new_net))
|
10063 |
nic.network = new_net_uuid |
10064 |
|
10065 |
if private.filled:
|
10066 |
nic.nicparams = private.filled |
10067 |
|
10068 |
for (key, val) in nic.nicparams.items(): |
10069 |
changes.append(("nic.%s/%d" % (key, idx), val))
|
10070 |
|
10071 |
return changes
|
10072 |
|
10073 |
def Exec(self, feedback_fn): |
10074 |
"""Modifies an instance.
|
10075 |
|
10076 |
All parameters take effect only at the next restart of the instance.
|
10077 |
|
10078 |
"""
|
10079 |
# Process here the warnings from CheckPrereq, as we don't have a
|
10080 |
# feedback_fn there.
|
10081 |
# TODO: Replace with self.LogWarning
|
10082 |
for warn in self.warn: |
10083 |
feedback_fn("WARNING: %s" % warn)
|
10084 |
|
10085 |
assert ((self.op.disk_template is None) ^ |
10086 |
bool(self.owned_locks(locking.LEVEL_NODE_RES))), \ |
10087 |
"Not owning any node resource locks"
|
10088 |
|
10089 |
result = [] |
10090 |
instance = self.instance
|
10091 |
|
10092 |
# New primary node
|
10093 |
if self.op.pnode: |
10094 |
instance.primary_node = self.op.pnode
|
10095 |
|
10096 |
# runtime memory
|
10097 |
if self.op.runtime_mem: |
10098 |
rpcres = self.rpc.call_instance_balloon_memory(instance.primary_node,
|
10099 |
instance, |
10100 |
self.op.runtime_mem)
|
10101 |
rpcres.Raise("Cannot modify instance runtime memory")
|
10102 |
result.append(("runtime_memory", self.op.runtime_mem)) |
10103 |
|
10104 |
# Apply disk changes
|
10105 |
ApplyContainerMods("disk", instance.disks, result, self.diskmod, |
10106 |
self._CreateNewDisk, self._ModifyDisk, self._RemoveDisk) |
10107 |
_UpdateIvNames(0, instance.disks)
|
10108 |
|
10109 |
if self.op.disk_template: |
10110 |
if __debug__:
|
10111 |
check_nodes = set(instance.all_nodes)
|
10112 |
if self.op.remote_node: |
10113 |
check_nodes.add(self.op.remote_node)
|
10114 |
for level in [locking.LEVEL_NODE, locking.LEVEL_NODE_RES]: |
10115 |
owned = self.owned_locks(level)
|
10116 |
assert not (check_nodes - owned), \ |
10117 |
("Not owning the correct locks, owning %r, expected at least %r" %
|
10118 |
(owned, check_nodes)) |
10119 |
|
10120 |
r_shut = _ShutdownInstanceDisks(self, instance)
|
10121 |
if not r_shut: |
10122 |
raise errors.OpExecError("Cannot shutdown instance disks, unable to" |
10123 |
" proceed with disk template conversion")
|
10124 |
mode = (instance.disk_template, self.op.disk_template)
|
10125 |
try:
|
10126 |
self._DISK_CONVERSIONS[mode](self, feedback_fn) |
10127 |
except:
|
10128 |
self.cfg.ReleaseDRBDMinors(instance.name)
|
10129 |
raise
|
10130 |
result.append(("disk_template", self.op.disk_template)) |
10131 |
|
10132 |
assert instance.disk_template == self.op.disk_template, \ |
10133 |
("Expected disk template '%s', found '%s'" %
|
10134 |
(self.op.disk_template, instance.disk_template))
|
10135 |
|
10136 |
# Release node and resource locks if there are any (they might already have
|
10137 |
# been released during disk conversion)
|
10138 |
_ReleaseLocks(self, locking.LEVEL_NODE)
|
10139 |
_ReleaseLocks(self, locking.LEVEL_NODE_RES)
|
10140 |
|
10141 |
# Apply NIC changes
|
10142 |
if self._new_nics is not None: |
10143 |
instance.nics = self._new_nics
|
10144 |
result.extend(self._nic_chgdesc)
|
10145 |
|
10146 |
# hvparams changes
|
10147 |
if self.op.hvparams: |
10148 |
instance.hvparams = self.hv_inst
|
10149 |
for key, val in self.op.hvparams.iteritems(): |
10150 |
result.append(("hv/%s" % key, val))
|
10151 |
|
10152 |
# beparams changes
|
10153 |
if self.op.beparams: |
10154 |
instance.beparams = self.be_inst
|
10155 |
for key, val in self.op.beparams.iteritems(): |
10156 |
result.append(("be/%s" % key, val))
|
10157 |
|
10158 |
# OS change
|
10159 |
if self.op.os_name: |
10160 |
instance.os = self.op.os_name
|
10161 |
|
10162 |
# osparams changes
|
10163 |
if self.op.osparams: |
10164 |
instance.osparams = self.os_inst
|
10165 |
for key, val in self.op.osparams.iteritems(): |
10166 |
result.append(("os/%s" % key, val))
|
10167 |
|
10168 |
if self.op.offline is None: |
10169 |
# Ignore
|
10170 |
pass
|
10171 |
elif self.op.offline: |
10172 |
# Mark instance as offline
|
10173 |
self.cfg.MarkInstanceOffline(instance.name)
|
10174 |
result.append(("admin_state", constants.ADMINST_OFFLINE))
|
10175 |
else:
|
10176 |
# Mark instance as online, but stopped
|
10177 |
self.cfg.MarkInstanceDown(instance.name)
|
10178 |
result.append(("admin_state", constants.ADMINST_DOWN))
|
10179 |
|
10180 |
self.cfg.Update(instance, feedback_fn, self.proc.GetECId()) |
10181 |
|
10182 |
assert not (self.owned_locks(locking.LEVEL_NODE_RES) or |
10183 |
self.owned_locks(locking.LEVEL_NODE)), \
|
10184 |
"All node locks should have been released by now"
|
10185 |
|
10186 |
return result
|
10187 |
|
10188 |
_DISK_CONVERSIONS = { |
10189 |
(constants.DT_PLAIN, constants.DT_DRBD8): _ConvertPlainToDrbd, |
10190 |
(constants.DT_DRBD8, constants.DT_PLAIN): _ConvertDrbdToPlain, |
10191 |
} |
10192 |
|
10193 |
|
10194 |
class LUInstanceChangeGroup(LogicalUnit): |
10195 |
HPATH = "instance-change-group"
|
10196 |
HTYPE = constants.HTYPE_INSTANCE |
10197 |
REQ_BGL = False
|
10198 |
|
10199 |
def ExpandNames(self): |
10200 |
self.share_locks = _ShareAll()
|
10201 |
|
10202 |
self.needed_locks = {
|
10203 |
locking.LEVEL_NODEGROUP: [], |
10204 |
locking.LEVEL_NODE: [], |
10205 |
locking.LEVEL_NODE_ALLOC: locking.ALL_SET, |
10206 |
} |
10207 |
|
10208 |
self._ExpandAndLockInstance()
|
10209 |
|
10210 |
if self.op.target_groups: |
10211 |
self.req_target_uuids = map(self.cfg.LookupNodeGroup, |
10212 |
self.op.target_groups)
|
10213 |
else:
|
10214 |
self.req_target_uuids = None |
10215 |
|
10216 |
self.op.iallocator = _GetDefaultIAllocator(self.cfg, self.op.iallocator) |
10217 |
|
10218 |
def DeclareLocks(self, level): |
10219 |
if level == locking.LEVEL_NODEGROUP:
|
10220 |
assert not self.needed_locks[locking.LEVEL_NODEGROUP] |
10221 |
|
10222 |
if self.req_target_uuids: |
10223 |
lock_groups = set(self.req_target_uuids) |
10224 |
|
10225 |
# Lock all groups used by instance optimistically; this requires going
|
10226 |
# via the node before it's locked, requiring verification later on
|
10227 |
instance_groups = self.cfg.GetInstanceNodeGroups(self.op.instance_name) |
10228 |
lock_groups.update(instance_groups) |
10229 |
else:
|
10230 |
# No target groups, need to lock all of them
|
10231 |
lock_groups = locking.ALL_SET |
10232 |
|
10233 |
self.needed_locks[locking.LEVEL_NODEGROUP] = lock_groups
|
10234 |
|
10235 |
elif level == locking.LEVEL_NODE:
|
10236 |
if self.req_target_uuids: |
10237 |
# Lock all nodes used by instances
|
10238 |
self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
|
10239 |
self._LockInstancesNodes()
|
10240 |
|
10241 |
# Lock all nodes in all potential target groups
|
10242 |
lock_groups = (frozenset(self.owned_locks(locking.LEVEL_NODEGROUP)) - |
10243 |
self.cfg.GetInstanceNodeGroups(self.op.instance_name)) |
10244 |
member_nodes = [node_name |
10245 |
for group in lock_groups |
10246 |
for node_name in self.cfg.GetNodeGroup(group).members] |
10247 |
self.needed_locks[locking.LEVEL_NODE].extend(member_nodes)
|
10248 |
else:
|
10249 |
# Lock all nodes as all groups are potential targets
|
10250 |
self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
|
10251 |
|
10252 |
def CheckPrereq(self): |
10253 |
owned_instances = frozenset(self.owned_locks(locking.LEVEL_INSTANCE)) |
10254 |
owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP)) |
10255 |
owned_nodes = frozenset(self.owned_locks(locking.LEVEL_NODE)) |
10256 |
|
10257 |
assert (self.req_target_uuids is None or |
10258 |
owned_groups.issuperset(self.req_target_uuids))
|
10259 |
assert owned_instances == set([self.op.instance_name]) |
10260 |
|
10261 |
# Get instance information
|
10262 |
self.instance = self.cfg.GetInstanceInfo(self.op.instance_name) |
10263 |
|
10264 |
# Check if node groups for locked instance are still correct
|
10265 |
assert owned_nodes.issuperset(self.instance.all_nodes), \ |
10266 |
("Instance %s's nodes changed while we kept the lock" %
|
10267 |
self.op.instance_name)
|
10268 |
|
10269 |
inst_groups = _CheckInstanceNodeGroups(self.cfg, self.op.instance_name, |
10270 |
owned_groups) |
10271 |
|
10272 |
if self.req_target_uuids: |
10273 |
# User requested specific target groups
|
10274 |
self.target_uuids = frozenset(self.req_target_uuids) |
10275 |
else:
|
10276 |
# All groups except those used by the instance are potential targets
|
10277 |
self.target_uuids = owned_groups - inst_groups
|
10278 |
|
10279 |
conflicting_groups = self.target_uuids & inst_groups
|
10280 |
if conflicting_groups:
|
10281 |
raise errors.OpPrereqError("Can't use group(s) '%s' as targets, they are" |
10282 |
" used by the instance '%s'" %
|
10283 |
(utils.CommaJoin(conflicting_groups), |
10284 |
self.op.instance_name),
|
10285 |
errors.ECODE_INVAL) |
10286 |
|
10287 |
if not self.target_uuids: |
10288 |
raise errors.OpPrereqError("There are no possible target groups", |
10289 |
errors.ECODE_INVAL) |
10290 |
|
10291 |
def BuildHooksEnv(self): |
10292 |
"""Build hooks env.
|
10293 |
|
10294 |
"""
|
10295 |
assert self.target_uuids |
10296 |
|
10297 |
env = { |
10298 |
"TARGET_GROUPS": " ".join(self.target_uuids), |
10299 |
} |
10300 |
|
10301 |
env.update(_BuildInstanceHookEnvByObject(self, self.instance)) |
10302 |
|
10303 |
return env
|
10304 |
|
10305 |
def BuildHooksNodes(self): |
10306 |
"""Build hooks nodes.
|
10307 |
|
10308 |
"""
|
10309 |
mn = self.cfg.GetMasterNode()
|
10310 |
return ([mn], [mn])
|
10311 |
|
10312 |
def Exec(self, feedback_fn): |
10313 |
instances = list(self.owned_locks(locking.LEVEL_INSTANCE)) |
10314 |
|
10315 |
assert instances == [self.op.instance_name], "Instance not locked" |
10316 |
|
10317 |
req = iallocator.IAReqGroupChange(instances=instances, |
10318 |
target_groups=list(self.target_uuids)) |
10319 |
ial = iallocator.IAllocator(self.cfg, self.rpc, req) |
10320 |
|
10321 |
ial.Run(self.op.iallocator)
|
10322 |
|
10323 |
if not ial.success: |
10324 |
raise errors.OpPrereqError("Can't compute solution for changing group of" |
10325 |
" instance '%s' using iallocator '%s': %s" %
|
10326 |
(self.op.instance_name, self.op.iallocator, |
10327 |
ial.info), errors.ECODE_NORES) |
10328 |
|
10329 |
jobs = _LoadNodeEvacResult(self, ial.result, self.op.early_release, False) |
10330 |
|
10331 |
self.LogInfo("Iallocator returned %s job(s) for changing group of" |
10332 |
" instance '%s'", len(jobs), self.op.instance_name) |
10333 |
|
10334 |
return ResultWithJobs(jobs)
|
10335 |
|
10336 |
|
10337 |
class LUBackupQuery(NoHooksLU): |
10338 |
"""Query the exports list
|
10339 |
|
10340 |
"""
|
10341 |
REQ_BGL = False
|
10342 |
|
10343 |
def CheckArguments(self): |
10344 |
self.expq = _ExportQuery(qlang.MakeSimpleFilter("node", self.op.nodes), |
10345 |
["node", "export"], self.op.use_locking) |
10346 |
|
10347 |
def ExpandNames(self): |
10348 |
self.expq.ExpandNames(self) |
10349 |
|
10350 |
def DeclareLocks(self, level): |
10351 |
self.expq.DeclareLocks(self, level) |
10352 |
|
10353 |
def Exec(self, feedback_fn): |
10354 |
result = {} |
10355 |
|
10356 |
for (node, expname) in self.expq.OldStyleQuery(self): |
10357 |
if expname is None: |
10358 |
result[node] = False
|
10359 |
else:
|
10360 |
result.setdefault(node, []).append(expname) |
10361 |
|
10362 |
return result
|
10363 |
|
10364 |
|
10365 |
class _ExportQuery(_QueryBase): |
10366 |
FIELDS = query.EXPORT_FIELDS |
10367 |
|
10368 |
#: The node name is not a unique key for this query
|
10369 |
SORT_FIELD = "node"
|
10370 |
|
10371 |
def ExpandNames(self, lu): |
10372 |
lu.needed_locks = {} |
10373 |
|
10374 |
# The following variables interact with _QueryBase._GetNames
|
10375 |
if self.names: |
10376 |
self.wanted = _GetWantedNodes(lu, self.names) |
10377 |
else:
|
10378 |
self.wanted = locking.ALL_SET
|
10379 |
|
10380 |
self.do_locking = self.use_locking |
10381 |
|
10382 |
if self.do_locking: |
10383 |
lu.share_locks = _ShareAll() |
10384 |
lu.needed_locks = { |
10385 |
locking.LEVEL_NODE: self.wanted,
|
10386 |
} |
10387 |
|
10388 |
if not self.names: |
10389 |
lu.needed_locks[locking.LEVEL_NODE_ALLOC] = locking.ALL_SET |
10390 |
|
10391 |
def DeclareLocks(self, lu, level): |
10392 |
pass
|
10393 |
|
10394 |
def _GetQueryData(self, lu): |
10395 |
"""Computes the list of nodes and their attributes.
|
10396 |
|
10397 |
"""
|
10398 |
# Locking is not used
|
10399 |
# TODO
|
10400 |
assert not (compat.any(lu.glm.is_owned(level) |
10401 |
for level in locking.LEVELS |
10402 |
if level != locking.LEVEL_CLUSTER) or |
10403 |
self.do_locking or self.use_locking) |
10404 |
|
10405 |
nodes = self._GetNames(lu, lu.cfg.GetNodeList(), locking.LEVEL_NODE)
|
10406 |
|
10407 |
result = [] |
10408 |
|
10409 |
for (node, nres) in lu.rpc.call_export_list(nodes).items(): |
10410 |
if nres.fail_msg:
|
10411 |
result.append((node, None))
|
10412 |
else:
|
10413 |
result.extend((node, expname) for expname in nres.payload) |
10414 |
|
10415 |
return result
|
10416 |
|
10417 |
|
10418 |
class LUBackupPrepare(NoHooksLU): |
10419 |
"""Prepares an instance for an export and returns useful information.
|
10420 |
|
10421 |
"""
|
10422 |
REQ_BGL = False
|
10423 |
|
10424 |
def ExpandNames(self): |
10425 |
self._ExpandAndLockInstance()
|
10426 |
|
10427 |
def CheckPrereq(self): |
10428 |
"""Check prerequisites.
|
10429 |
|
10430 |
"""
|
10431 |
instance_name = self.op.instance_name
|
10432 |
|
10433 |
self.instance = self.cfg.GetInstanceInfo(instance_name) |
10434 |
assert self.instance is not None, \ |
10435 |
"Cannot retrieve locked instance %s" % self.op.instance_name |
10436 |
_CheckNodeOnline(self, self.instance.primary_node) |
10437 |
|
10438 |
self._cds = _GetClusterDomainSecret()
|
10439 |
|
10440 |
def Exec(self, feedback_fn): |
10441 |
"""Prepares an instance for an export.
|
10442 |
|
10443 |
"""
|
10444 |
instance = self.instance
|
10445 |
|
10446 |
if self.op.mode == constants.EXPORT_MODE_REMOTE: |
10447 |
salt = utils.GenerateSecret(8)
|
10448 |
|
10449 |
feedback_fn("Generating X509 certificate on %s" % instance.primary_node)
|
10450 |
result = self.rpc.call_x509_cert_create(instance.primary_node,
|
10451 |
constants.RIE_CERT_VALIDITY) |
10452 |
result.Raise("Can't create X509 key and certificate on %s" % result.node)
|
10453 |
|
10454 |
(name, cert_pem) = result.payload |
10455 |
|
10456 |
cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM, |
10457 |
cert_pem) |
10458 |
|
10459 |
return {
|
10460 |
"handshake": masterd.instance.ComputeRemoteExportHandshake(self._cds), |
10461 |
"x509_key_name": (name, utils.Sha1Hmac(self._cds, name, salt=salt), |
10462 |
salt), |
10463 |
"x509_ca": utils.SignX509Certificate(cert, self._cds, salt), |
10464 |
} |
10465 |
|
10466 |
return None |
10467 |
|
10468 |
|
10469 |
class LUBackupExport(LogicalUnit): |
10470 |
"""Export an instance to an image in the cluster.
|
10471 |
|
10472 |
"""
|
10473 |
HPATH = "instance-export"
|
10474 |
HTYPE = constants.HTYPE_INSTANCE |
10475 |
REQ_BGL = False
|
10476 |
|
10477 |
def CheckArguments(self): |
10478 |
"""Check the arguments.
|
10479 |
|
10480 |
"""
|
10481 |
self.x509_key_name = self.op.x509_key_name |
10482 |
self.dest_x509_ca_pem = self.op.destination_x509_ca |
10483 |
|
10484 |
if self.op.mode == constants.EXPORT_MODE_REMOTE: |
10485 |
if not self.x509_key_name: |
10486 |
raise errors.OpPrereqError("Missing X509 key name for encryption", |
10487 |
errors.ECODE_INVAL) |
10488 |
|
10489 |
if not self.dest_x509_ca_pem: |
10490 |
raise errors.OpPrereqError("Missing destination X509 CA", |
10491 |
errors.ECODE_INVAL) |
10492 |
|
10493 |
def ExpandNames(self): |
10494 |
self._ExpandAndLockInstance()
|
10495 |
|
10496 |
# Lock all nodes for local exports
|
10497 |
if self.op.mode == constants.EXPORT_MODE_LOCAL: |
10498 |
# FIXME: lock only instance primary and destination node
|
10499 |
#
|
10500 |
# Sad but true, for now we have do lock all nodes, as we don't know where
|
10501 |
# the previous export might be, and in this LU we search for it and
|
10502 |
# remove it from its current node. In the future we could fix this by:
|
10503 |
# - making a tasklet to search (share-lock all), then create the
|
10504 |
# new one, then one to remove, after
|
10505 |
# - removing the removal operation altogether
|
10506 |
self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
|
10507 |
|
10508 |
# Allocations should be stopped while this LU runs with node locks, but
|
10509 |
# it doesn't have to be exclusive
|
10510 |
self.share_locks[locking.LEVEL_NODE_ALLOC] = 1 |
10511 |
self.needed_locks[locking.LEVEL_NODE_ALLOC] = locking.ALL_SET
|
10512 |
|
10513 |
def DeclareLocks(self, level): |
10514 |
"""Last minute lock declaration."""
|
10515 |
# All nodes are locked anyway, so nothing to do here.
|
10516 |
|
10517 |
def BuildHooksEnv(self): |
10518 |
"""Build hooks env.
|
10519 |
|
10520 |
This will run on the master, primary node and target node.
|
10521 |
|
10522 |
"""
|
10523 |
env = { |
10524 |
"EXPORT_MODE": self.op.mode, |
10525 |
"EXPORT_NODE": self.op.target_node, |
10526 |
"EXPORT_DO_SHUTDOWN": self.op.shutdown, |
10527 |
"SHUTDOWN_TIMEOUT": self.op.shutdown_timeout, |
10528 |
# TODO: Generic function for boolean env variables
|
10529 |
"REMOVE_INSTANCE": str(bool(self.op.remove_instance)), |
10530 |
} |
10531 |
|
10532 |
env.update(_BuildInstanceHookEnvByObject(self, self.instance)) |
10533 |
|
10534 |
return env
|
10535 |
|
10536 |
def BuildHooksNodes(self): |
10537 |
"""Build hooks nodes.
|
10538 |
|
10539 |
"""
|
10540 |
nl = [self.cfg.GetMasterNode(), self.instance.primary_node] |
10541 |
|
10542 |
if self.op.mode == constants.EXPORT_MODE_LOCAL: |
10543 |
nl.append(self.op.target_node)
|
10544 |
|
10545 |
return (nl, nl)
|
10546 |
|
10547 |
def CheckPrereq(self): |
10548 |
"""Check prerequisites.
|
10549 |
|
10550 |
This checks that the instance and node names are valid.
|
10551 |
|
10552 |
"""
|
10553 |
instance_name = self.op.instance_name
|
10554 |
|
10555 |
self.instance = self.cfg.GetInstanceInfo(instance_name) |
10556 |
assert self.instance is not None, \ |
10557 |
"Cannot retrieve locked instance %s" % self.op.instance_name |
10558 |
_CheckNodeOnline(self, self.instance.primary_node) |
10559 |
|
10560 |
if (self.op.remove_instance and |
10561 |
self.instance.admin_state == constants.ADMINST_UP and |
10562 |
not self.op.shutdown): |
10563 |
raise errors.OpPrereqError("Can not remove instance without shutting it" |
10564 |
" down before", errors.ECODE_STATE)
|
10565 |
|
10566 |
if self.op.mode == constants.EXPORT_MODE_LOCAL: |
10567 |
self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node) |
10568 |
self.dst_node = self.cfg.GetNodeInfo(self.op.target_node) |
10569 |
assert self.dst_node is not None |
10570 |
|
10571 |
_CheckNodeOnline(self, self.dst_node.name) |
10572 |
_CheckNodeNotDrained(self, self.dst_node.name) |
10573 |
|
10574 |
self._cds = None |
10575 |
self.dest_disk_info = None |
10576 |
self.dest_x509_ca = None |
10577 |
|
10578 |
elif self.op.mode == constants.EXPORT_MODE_REMOTE: |
10579 |
self.dst_node = None |
10580 |
|
10581 |
if len(self.op.target_node) != len(self.instance.disks): |
10582 |
raise errors.OpPrereqError(("Received destination information for %s" |
10583 |
" disks, but instance %s has %s disks") %
|
10584 |
(len(self.op.target_node), instance_name, |
10585 |
len(self.instance.disks)), |
10586 |
errors.ECODE_INVAL) |
10587 |
|
10588 |
cds = _GetClusterDomainSecret() |
10589 |
|
10590 |
# Check X509 key name
|
10591 |
try:
|
10592 |
(key_name, hmac_digest, hmac_salt) = self.x509_key_name
|
10593 |
except (TypeError, ValueError), err: |
10594 |
raise errors.OpPrereqError("Invalid data for X509 key name: %s" % err, |
10595 |
errors.ECODE_INVAL) |
10596 |
|
10597 |
if not utils.VerifySha1Hmac(cds, key_name, hmac_digest, salt=hmac_salt): |
10598 |
raise errors.OpPrereqError("HMAC for X509 key name is wrong", |
10599 |
errors.ECODE_INVAL) |
10600 |
|
10601 |
# Load and verify CA
|
10602 |
try:
|
10603 |
(cert, _) = utils.LoadSignedX509Certificate(self.dest_x509_ca_pem, cds)
|
10604 |
except OpenSSL.crypto.Error, err:
|
10605 |
raise errors.OpPrereqError("Unable to load destination X509 CA (%s)" % |
10606 |
(err, ), errors.ECODE_INVAL) |
10607 |
|
10608 |
(errcode, msg) = utils.VerifyX509Certificate(cert, None, None) |
10609 |
if errcode is not None: |
10610 |
raise errors.OpPrereqError("Invalid destination X509 CA (%s)" % |
10611 |
(msg, ), errors.ECODE_INVAL) |
10612 |
|
10613 |
self.dest_x509_ca = cert
|
10614 |
|
10615 |
# Verify target information
|
10616 |
disk_info = [] |
10617 |
for idx, disk_data in enumerate(self.op.target_node): |
10618 |
try:
|
10619 |
(host, port, magic) = \ |
10620 |
masterd.instance.CheckRemoteExportDiskInfo(cds, idx, disk_data) |
10621 |
except errors.GenericError, err:
|
10622 |
raise errors.OpPrereqError("Target info for disk %s: %s" % |
10623 |
(idx, err), errors.ECODE_INVAL) |
10624 |
|
10625 |
disk_info.append((host, port, magic)) |
10626 |
|
10627 |
assert len(disk_info) == len(self.op.target_node) |
10628 |
self.dest_disk_info = disk_info
|
10629 |
|
10630 |
else:
|
10631 |
raise errors.ProgrammerError("Unhandled export mode %r" % |
10632 |
self.op.mode)
|
10633 |
|
10634 |
# instance disk type verification
|
10635 |
# TODO: Implement export support for file-based disks
|
10636 |
for disk in self.instance.disks: |
10637 |
if disk.dev_type == constants.LD_FILE:
|
10638 |
raise errors.OpPrereqError("Export not supported for instances with" |
10639 |
" file-based disks", errors.ECODE_INVAL)
|
10640 |
|
10641 |
def _CleanupExports(self, feedback_fn): |
10642 |
"""Removes exports of current instance from all other nodes.
|
10643 |
|
10644 |
If an instance in a cluster with nodes A..D was exported to node C, its
|
10645 |
exports will be removed from the nodes A, B and D.
|
10646 |
|
10647 |
"""
|
10648 |
assert self.op.mode != constants.EXPORT_MODE_REMOTE |
10649 |
|
10650 |
nodelist = self.cfg.GetNodeList()
|
10651 |
nodelist.remove(self.dst_node.name)
|
10652 |
|
10653 |
# on one-node clusters nodelist will be empty after the removal
|
10654 |
# if we proceed the backup would be removed because OpBackupQuery
|
10655 |
# substitutes an empty list with the full cluster node list.
|
10656 |
iname = self.instance.name
|
10657 |
if nodelist:
|
10658 |
feedback_fn("Removing old exports for instance %s" % iname)
|
10659 |
exportlist = self.rpc.call_export_list(nodelist)
|
10660 |
for node in exportlist: |
10661 |
if exportlist[node].fail_msg:
|
10662 |
continue
|
10663 |
if iname in exportlist[node].payload: |
10664 |
msg = self.rpc.call_export_remove(node, iname).fail_msg
|
10665 |
if msg:
|
10666 |
self.LogWarning("Could not remove older export for instance %s" |
10667 |
" on node %s: %s", iname, node, msg)
|
10668 |
|
10669 |
def Exec(self, feedback_fn): |
10670 |
"""Export an instance to an image in the cluster.
|
10671 |
|
10672 |
"""
|
10673 |
assert self.op.mode in constants.EXPORT_MODES |
10674 |
|
10675 |
instance = self.instance
|
10676 |
src_node = instance.primary_node |
10677 |
|
10678 |
if self.op.shutdown: |
10679 |
# shutdown the instance, but not the disks
|
10680 |
feedback_fn("Shutting down instance %s" % instance.name)
|
10681 |
result = self.rpc.call_instance_shutdown(src_node, instance,
|
10682 |
self.op.shutdown_timeout,
|
10683 |
self.op.reason)
|
10684 |
# TODO: Maybe ignore failures if ignore_remove_failures is set
|
10685 |
result.Raise("Could not shutdown instance %s on"
|
10686 |
" node %s" % (instance.name, src_node))
|
10687 |
|
10688 |
# set the disks ID correctly since call_instance_start needs the
|
10689 |
# correct drbd minor to create the symlinks
|
10690 |
for disk in instance.disks: |
10691 |
self.cfg.SetDiskID(disk, src_node)
|
10692 |
|
10693 |
activate_disks = (instance.admin_state != constants.ADMINST_UP) |
10694 |
|
10695 |
if activate_disks:
|
10696 |
# Activate the instance disks if we'exporting a stopped instance
|
10697 |
feedback_fn("Activating disks for %s" % instance.name)
|
10698 |
_StartInstanceDisks(self, instance, None) |
10699 |
|
10700 |
try:
|
10701 |
helper = masterd.instance.ExportInstanceHelper(self, feedback_fn,
|
10702 |
instance) |
10703 |
|
10704 |
helper.CreateSnapshots() |
10705 |
try:
|
10706 |
if (self.op.shutdown and |
10707 |
instance.admin_state == constants.ADMINST_UP and
|
10708 |
not self.op.remove_instance): |
10709 |
assert not activate_disks |
10710 |
feedback_fn("Starting instance %s" % instance.name)
|
10711 |
result = self.rpc.call_instance_start(src_node,
|
10712 |
(instance, None, None), False, |
10713 |
self.op.reason)
|
10714 |
msg = result.fail_msg |
10715 |
if msg:
|
10716 |
feedback_fn("Failed to start instance: %s" % msg)
|
10717 |
_ShutdownInstanceDisks(self, instance)
|
10718 |
raise errors.OpExecError("Could not start instance: %s" % msg) |
10719 |
|
10720 |
if self.op.mode == constants.EXPORT_MODE_LOCAL: |
10721 |
(fin_resu, dresults) = helper.LocalExport(self.dst_node)
|
10722 |
elif self.op.mode == constants.EXPORT_MODE_REMOTE: |
10723 |
connect_timeout = constants.RIE_CONNECT_TIMEOUT |
10724 |
timeouts = masterd.instance.ImportExportTimeouts(connect_timeout) |
10725 |
|
10726 |
(key_name, _, _) = self.x509_key_name
|
10727 |
|
10728 |
dest_ca_pem = \ |
10729 |
OpenSSL.crypto.dump_certificate(OpenSSL.crypto.FILETYPE_PEM, |
10730 |
self.dest_x509_ca)
|
10731 |
|
10732 |
(fin_resu, dresults) = helper.RemoteExport(self.dest_disk_info,
|
10733 |
key_name, dest_ca_pem, |
10734 |
timeouts) |
10735 |
finally:
|
10736 |
helper.Cleanup() |
10737 |
|
10738 |
# Check for backwards compatibility
|
10739 |
assert len(dresults) == len(instance.disks) |
10740 |
assert compat.all(isinstance(i, bool) for i in dresults), \ |
10741 |
"Not all results are boolean: %r" % dresults
|
10742 |
|
10743 |
finally:
|
10744 |
if activate_disks:
|
10745 |
feedback_fn("Deactivating disks for %s" % instance.name)
|
10746 |
_ShutdownInstanceDisks(self, instance)
|
10747 |
|
10748 |
if not (compat.all(dresults) and fin_resu): |
10749 |
failures = [] |
10750 |
if not fin_resu: |
10751 |
failures.append("export finalization")
|
10752 |
if not compat.all(dresults): |
10753 |
fdsk = utils.CommaJoin(idx for (idx, dsk) in enumerate(dresults) |
10754 |
if not dsk) |
10755 |
failures.append("disk export: disk(s) %s" % fdsk)
|
10756 |
|
10757 |
raise errors.OpExecError("Export failed, errors in %s" % |
10758 |
utils.CommaJoin(failures)) |
10759 |
|
10760 |
# At this point, the export was successful, we can cleanup/finish
|
10761 |
|
10762 |
# Remove instance if requested
|
10763 |
if self.op.remove_instance: |
10764 |
feedback_fn("Removing instance %s" % instance.name)
|
10765 |
_RemoveInstance(self, feedback_fn, instance,
|
10766 |
self.op.ignore_remove_failures)
|
10767 |
|
10768 |
if self.op.mode == constants.EXPORT_MODE_LOCAL: |
10769 |
self._CleanupExports(feedback_fn)
|
10770 |
|
10771 |
return fin_resu, dresults
|
10772 |
|
10773 |
|
10774 |
class LUBackupRemove(NoHooksLU): |
10775 |
"""Remove exports related to the named instance.
|
10776 |
|
10777 |
"""
|
10778 |
REQ_BGL = False
|
10779 |
|
10780 |
def ExpandNames(self): |
10781 |
self.needed_locks = {
|
10782 |
# We need all nodes to be locked in order for RemoveExport to work, but
|
10783 |
# we don't need to lock the instance itself, as nothing will happen to it
|
10784 |
# (and we can remove exports also for a removed instance)
|
10785 |
locking.LEVEL_NODE: locking.ALL_SET, |
10786 |
|
10787 |
# Removing backups is quick, so blocking allocations is justified
|
10788 |
locking.LEVEL_NODE_ALLOC: locking.ALL_SET, |
10789 |
} |
10790 |
|
10791 |
# Allocations should be stopped while this LU runs with node locks, but it
|
10792 |
# doesn't have to be exclusive
|
10793 |
self.share_locks[locking.LEVEL_NODE_ALLOC] = 1 |
10794 |
|
10795 |
def Exec(self, feedback_fn): |
10796 |
"""Remove any export.
|
10797 |
|
10798 |
"""
|
10799 |
instance_name = self.cfg.ExpandInstanceName(self.op.instance_name) |
10800 |
# If the instance was not found we'll try with the name that was passed in.
|
10801 |
# This will only work if it was an FQDN, though.
|
10802 |
fqdn_warn = False
|
10803 |
if not instance_name: |
10804 |
fqdn_warn = True
|
10805 |
instance_name = self.op.instance_name
|
10806 |
|
10807 |
locked_nodes = self.owned_locks(locking.LEVEL_NODE)
|
10808 |
exportlist = self.rpc.call_export_list(locked_nodes)
|
10809 |
found = False
|
10810 |
for node in exportlist: |
10811 |
msg = exportlist[node].fail_msg |
10812 |
if msg:
|
10813 |
self.LogWarning("Failed to query node %s (continuing): %s", node, msg) |
10814 |
continue
|
10815 |
if instance_name in exportlist[node].payload: |
10816 |
found = True
|
10817 |
result = self.rpc.call_export_remove(node, instance_name)
|
10818 |
msg = result.fail_msg |
10819 |
if msg:
|
10820 |
logging.error("Could not remove export for instance %s"
|
10821 |
" on node %s: %s", instance_name, node, msg)
|
10822 |
|
10823 |
if fqdn_warn and not found: |
10824 |
feedback_fn("Export not found. If trying to remove an export belonging"
|
10825 |
" to a deleted instance please use its Fully Qualified"
|
10826 |
" Domain Name.")
|
10827 |
|
10828 |
|
10829 |
class LURestrictedCommand(NoHooksLU): |
10830 |
"""Logical unit for executing restricted commands.
|
10831 |
|
10832 |
"""
|
10833 |
REQ_BGL = False
|
10834 |
|
10835 |
def ExpandNames(self): |
10836 |
if self.op.nodes: |
10837 |
self.op.nodes = _GetWantedNodes(self, self.op.nodes) |
10838 |
|
10839 |
self.needed_locks = {
|
10840 |
locking.LEVEL_NODE: self.op.nodes,
|
10841 |
} |
10842 |
self.share_locks = {
|
10843 |
locking.LEVEL_NODE: not self.op.use_locking, |
10844 |
} |
10845 |
|
10846 |
def CheckPrereq(self): |
10847 |
"""Check prerequisites.
|
10848 |
|
10849 |
"""
|
10850 |
|
10851 |
def Exec(self, feedback_fn): |
10852 |
"""Execute restricted command and return output.
|
10853 |
|
10854 |
"""
|
10855 |
owned_nodes = frozenset(self.owned_locks(locking.LEVEL_NODE)) |
10856 |
|
10857 |
# Check if correct locks are held
|
10858 |
assert set(self.op.nodes).issubset(owned_nodes) |
10859 |
|
10860 |
rpcres = self.rpc.call_restricted_command(self.op.nodes, self.op.command) |
10861 |
|
10862 |
result = [] |
10863 |
|
10864 |
for node_name in self.op.nodes: |
10865 |
nres = rpcres[node_name] |
10866 |
if nres.fail_msg:
|
10867 |
msg = ("Command '%s' on node '%s' failed: %s" %
|
10868 |
(self.op.command, node_name, nres.fail_msg))
|
10869 |
result.append((False, msg))
|
10870 |
else:
|
10871 |
result.append((True, nres.payload))
|
10872 |
|
10873 |
return result
|
10874 |
|
10875 |
|
10876 |
#: Query type implementations
|
10877 |
_QUERY_IMPL = { |
10878 |
constants.QR_CLUSTER: _ClusterQuery, |
10879 |
constants.QR_INSTANCE: _InstanceQuery, |
10880 |
constants.QR_NODE: _NodeQuery, |
10881 |
constants.QR_GROUP: _GroupQuery, |
10882 |
constants.QR_NETWORK: _NetworkQuery, |
10883 |
constants.QR_OS: _OsQuery, |
10884 |
constants.QR_EXTSTORAGE: _ExtStorageQuery, |
10885 |
constants.QR_EXPORT: _ExportQuery, |
10886 |
} |
10887 |
|
10888 |
assert set(_QUERY_IMPL.keys()) == constants.QR_VIA_OP |
10889 |
|
10890 |
|
10891 |
def _GetQueryImplementation(name): |
10892 |
"""Returns the implemtnation for a query type.
|
10893 |
|
10894 |
@param name: Query type, must be one of L{constants.QR_VIA_OP}
|
10895 |
|
10896 |
"""
|
10897 |
try:
|
10898 |
return _QUERY_IMPL[name]
|
10899 |
except KeyError: |
10900 |
raise errors.OpPrereqError("Unknown query resource '%s'" % name, |
10901 |
errors.ECODE_INVAL) |
10902 |
|
10903 |
|
10904 |
def _CheckForConflictingIp(lu, ip, node): |
10905 |
"""In case of conflicting IP address raise error.
|
10906 |
|
10907 |
@type ip: string
|
10908 |
@param ip: IP address
|
10909 |
@type node: string
|
10910 |
@param node: node name
|
10911 |
|
10912 |
"""
|
10913 |
(conf_net, _) = lu.cfg.CheckIPInNodeGroup(ip, node) |
10914 |
if conf_net is not None: |
10915 |
raise errors.OpPrereqError(("The requested IP address (%s) belongs to" |
10916 |
" network %s, but the target NIC does not." %
|
10917 |
(ip, conf_net)), |
10918 |
errors.ECODE_STATE) |
10919 |
|
10920 |
return (None, None) |