root / lib / cmdlib / instance_migration.py @ 0973f9ed
History | View | Annotate | Download (36 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 |
"""Logical units dealing with instance migration an failover."""
|
23 |
|
24 |
import logging |
25 |
import time |
26 |
|
27 |
from ganeti import constants |
28 |
from ganeti import errors |
29 |
from ganeti import locking |
30 |
from ganeti.masterd import iallocator |
31 |
from ganeti import utils |
32 |
from ganeti.cmdlib.base import LogicalUnit, Tasklet |
33 |
from ganeti.cmdlib.common import ExpandInstanceName, \ |
34 |
CheckIAllocatorOrNode, ExpandNodeName |
35 |
from ganeti.cmdlib.instance_storage import CheckDiskConsistency, \ |
36 |
ExpandCheckDisks, ShutdownInstanceDisks, AssembleInstanceDisks |
37 |
from ganeti.cmdlib.instance_utils import BuildInstanceHookEnvByObject, \ |
38 |
CheckTargetNodeIPolicy, ReleaseLocks, CheckNodeNotDrained, \ |
39 |
CopyLockList, CheckNodeFreeMemory, CheckInstanceBridgesExist |
40 |
|
41 |
import ganeti.masterd.instance |
42 |
|
43 |
|
44 |
def _ExpandNamesForMigration(lu): |
45 |
"""Expands names for use with L{TLMigrateInstance}.
|
46 |
|
47 |
@type lu: L{LogicalUnit}
|
48 |
|
49 |
"""
|
50 |
if lu.op.target_node is not None: |
51 |
lu.op.target_node = ExpandNodeName(lu.cfg, lu.op.target_node) |
52 |
|
53 |
lu.needed_locks[locking.LEVEL_NODE] = [] |
54 |
lu.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE |
55 |
|
56 |
lu.needed_locks[locking.LEVEL_NODE_RES] = [] |
57 |
lu.recalculate_locks[locking.LEVEL_NODE_RES] = constants.LOCKS_REPLACE |
58 |
|
59 |
# The node allocation lock is actually only needed for externally replicated
|
60 |
# instances (e.g. sharedfile or RBD) and if an iallocator is used.
|
61 |
lu.needed_locks[locking.LEVEL_NODE_ALLOC] = [] |
62 |
|
63 |
|
64 |
def _DeclareLocksForMigration(lu, level): |
65 |
"""Declares locks for L{TLMigrateInstance}.
|
66 |
|
67 |
@type lu: L{LogicalUnit}
|
68 |
@param level: Lock level
|
69 |
|
70 |
"""
|
71 |
if level == locking.LEVEL_NODE_ALLOC:
|
72 |
assert lu.op.instance_name in lu.owned_locks(locking.LEVEL_INSTANCE) |
73 |
|
74 |
instance = lu.cfg.GetInstanceInfo(lu.op.instance_name) |
75 |
|
76 |
# Node locks are already declared here rather than at LEVEL_NODE as we need
|
77 |
# the instance object anyway to declare the node allocation lock.
|
78 |
if instance.disk_template in constants.DTS_EXT_MIRROR: |
79 |
if lu.op.target_node is None: |
80 |
lu.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET |
81 |
lu.needed_locks[locking.LEVEL_NODE_ALLOC] = locking.ALL_SET |
82 |
else:
|
83 |
lu.needed_locks[locking.LEVEL_NODE] = [instance.primary_node, |
84 |
lu.op.target_node] |
85 |
del lu.recalculate_locks[locking.LEVEL_NODE]
|
86 |
else:
|
87 |
lu._LockInstancesNodes() # pylint: disable=W0212
|
88 |
|
89 |
elif level == locking.LEVEL_NODE:
|
90 |
# Node locks are declared together with the node allocation lock
|
91 |
assert (lu.needed_locks[locking.LEVEL_NODE] or |
92 |
lu.needed_locks[locking.LEVEL_NODE] is locking.ALL_SET)
|
93 |
|
94 |
elif level == locking.LEVEL_NODE_RES:
|
95 |
# Copy node locks
|
96 |
lu.needed_locks[locking.LEVEL_NODE_RES] = \ |
97 |
CopyLockList(lu.needed_locks[locking.LEVEL_NODE]) |
98 |
|
99 |
|
100 |
class LUInstanceFailover(LogicalUnit): |
101 |
"""Failover an instance.
|
102 |
|
103 |
"""
|
104 |
HPATH = "instance-failover"
|
105 |
HTYPE = constants.HTYPE_INSTANCE |
106 |
REQ_BGL = False
|
107 |
|
108 |
def CheckArguments(self): |
109 |
"""Check the arguments.
|
110 |
|
111 |
"""
|
112 |
self.iallocator = getattr(self.op, "iallocator", None) |
113 |
self.target_node = getattr(self.op, "target_node", None) |
114 |
|
115 |
def ExpandNames(self): |
116 |
self._ExpandAndLockInstance()
|
117 |
_ExpandNamesForMigration(self)
|
118 |
|
119 |
self._migrater = \
|
120 |
TLMigrateInstance(self, self.op.instance_name, self.op.cleanup, True, |
121 |
False, self.op.ignore_consistency, True, |
122 |
self.op.shutdown_timeout, self.op.ignore_ipolicy) |
123 |
|
124 |
self.tasklets = [self._migrater] |
125 |
|
126 |
def DeclareLocks(self, level): |
127 |
_DeclareLocksForMigration(self, level)
|
128 |
|
129 |
def BuildHooksEnv(self): |
130 |
"""Build hooks env.
|
131 |
|
132 |
This runs on master, primary and secondary nodes of the instance.
|
133 |
|
134 |
"""
|
135 |
instance = self._migrater.instance
|
136 |
source_node = instance.primary_node |
137 |
target_node = self.op.target_node
|
138 |
env = { |
139 |
"IGNORE_CONSISTENCY": self.op.ignore_consistency, |
140 |
"SHUTDOWN_TIMEOUT": self.op.shutdown_timeout, |
141 |
"OLD_PRIMARY": source_node,
|
142 |
"NEW_PRIMARY": target_node,
|
143 |
"FAILOVER_CLEANUP": self.op.cleanup, |
144 |
} |
145 |
|
146 |
if instance.disk_template in constants.DTS_INT_MIRROR: |
147 |
env["OLD_SECONDARY"] = instance.secondary_nodes[0] |
148 |
env["NEW_SECONDARY"] = source_node
|
149 |
else:
|
150 |
env["OLD_SECONDARY"] = env["NEW_SECONDARY"] = "" |
151 |
|
152 |
env.update(BuildInstanceHookEnvByObject(self, instance))
|
153 |
|
154 |
return env
|
155 |
|
156 |
def BuildHooksNodes(self): |
157 |
"""Build hooks nodes.
|
158 |
|
159 |
"""
|
160 |
instance = self._migrater.instance
|
161 |
nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes) |
162 |
return (nl, nl + [instance.primary_node])
|
163 |
|
164 |
|
165 |
class LUInstanceMigrate(LogicalUnit): |
166 |
"""Migrate an instance.
|
167 |
|
168 |
This is migration without shutting down, compared to the failover,
|
169 |
which is done with shutdown.
|
170 |
|
171 |
"""
|
172 |
HPATH = "instance-migrate"
|
173 |
HTYPE = constants.HTYPE_INSTANCE |
174 |
REQ_BGL = False
|
175 |
|
176 |
def ExpandNames(self): |
177 |
self._ExpandAndLockInstance()
|
178 |
_ExpandNamesForMigration(self)
|
179 |
|
180 |
self._migrater = \
|
181 |
TLMigrateInstance(self, self.op.instance_name, self.op.cleanup, |
182 |
False, self.op.allow_failover, False, |
183 |
self.op.allow_runtime_changes,
|
184 |
constants.DEFAULT_SHUTDOWN_TIMEOUT, |
185 |
self.op.ignore_ipolicy)
|
186 |
|
187 |
self.tasklets = [self._migrater] |
188 |
|
189 |
def DeclareLocks(self, level): |
190 |
_DeclareLocksForMigration(self, level)
|
191 |
|
192 |
def BuildHooksEnv(self): |
193 |
"""Build hooks env.
|
194 |
|
195 |
This runs on master, primary and secondary nodes of the instance.
|
196 |
|
197 |
"""
|
198 |
instance = self._migrater.instance
|
199 |
source_node = instance.primary_node |
200 |
target_node = self.op.target_node
|
201 |
env = BuildInstanceHookEnvByObject(self, instance)
|
202 |
env.update({ |
203 |
"MIGRATE_LIVE": self._migrater.live, |
204 |
"MIGRATE_CLEANUP": self.op.cleanup, |
205 |
"OLD_PRIMARY": source_node,
|
206 |
"NEW_PRIMARY": target_node,
|
207 |
"ALLOW_RUNTIME_CHANGES": self.op.allow_runtime_changes, |
208 |
}) |
209 |
|
210 |
if instance.disk_template in constants.DTS_INT_MIRROR: |
211 |
env["OLD_SECONDARY"] = target_node
|
212 |
env["NEW_SECONDARY"] = source_node
|
213 |
else:
|
214 |
env["OLD_SECONDARY"] = env["NEW_SECONDARY"] = None |
215 |
|
216 |
return env
|
217 |
|
218 |
def BuildHooksNodes(self): |
219 |
"""Build hooks nodes.
|
220 |
|
221 |
"""
|
222 |
instance = self._migrater.instance
|
223 |
snodes = list(instance.secondary_nodes)
|
224 |
nl = [self.cfg.GetMasterNode(), instance.primary_node] + snodes
|
225 |
return (nl, nl)
|
226 |
|
227 |
|
228 |
class TLMigrateInstance(Tasklet): |
229 |
"""Tasklet class for instance migration.
|
230 |
|
231 |
@type live: boolean
|
232 |
@ivar live: whether the migration will be done live or non-live;
|
233 |
this variable is initalized only after CheckPrereq has run
|
234 |
@type cleanup: boolean
|
235 |
@ivar cleanup: Wheater we cleanup from a failed migration
|
236 |
@type iallocator: string
|
237 |
@ivar iallocator: The iallocator used to determine target_node
|
238 |
@type target_node: string
|
239 |
@ivar target_node: If given, the target_node to reallocate the instance to
|
240 |
@type failover: boolean
|
241 |
@ivar failover: Whether operation results in failover or migration
|
242 |
@type fallback: boolean
|
243 |
@ivar fallback: Whether fallback to failover is allowed if migration not
|
244 |
possible
|
245 |
@type ignore_consistency: boolean
|
246 |
@ivar ignore_consistency: Wheter we should ignore consistency between source
|
247 |
and target node
|
248 |
@type shutdown_timeout: int
|
249 |
@ivar shutdown_timeout: In case of failover timeout of the shutdown
|
250 |
@type ignore_ipolicy: bool
|
251 |
@ivar ignore_ipolicy: If true, we can ignore instance policy when migrating
|
252 |
|
253 |
"""
|
254 |
|
255 |
# Constants
|
256 |
_MIGRATION_POLL_INTERVAL = 1 # seconds |
257 |
_MIGRATION_FEEDBACK_INTERVAL = 10 # seconds |
258 |
|
259 |
def __init__(self, lu, instance_name, cleanup, failover, fallback, |
260 |
ignore_consistency, allow_runtime_changes, shutdown_timeout, |
261 |
ignore_ipolicy): |
262 |
"""Initializes this class.
|
263 |
|
264 |
"""
|
265 |
Tasklet.__init__(self, lu)
|
266 |
|
267 |
# Parameters
|
268 |
self.instance_name = instance_name
|
269 |
self.cleanup = cleanup
|
270 |
self.live = False # will be overridden later |
271 |
self.failover = failover
|
272 |
self.fallback = fallback
|
273 |
self.ignore_consistency = ignore_consistency
|
274 |
self.shutdown_timeout = shutdown_timeout
|
275 |
self.ignore_ipolicy = ignore_ipolicy
|
276 |
self.allow_runtime_changes = allow_runtime_changes
|
277 |
|
278 |
def CheckPrereq(self): |
279 |
"""Check prerequisites.
|
280 |
|
281 |
This checks that the instance is in the cluster.
|
282 |
|
283 |
"""
|
284 |
instance_name = ExpandInstanceName(self.lu.cfg, self.instance_name) |
285 |
instance = self.cfg.GetInstanceInfo(instance_name)
|
286 |
assert instance is not None |
287 |
self.instance = instance
|
288 |
cluster = self.cfg.GetClusterInfo()
|
289 |
|
290 |
if (not self.cleanup and |
291 |
not instance.admin_state == constants.ADMINST_UP and |
292 |
not self.failover and self.fallback): |
293 |
self.lu.LogInfo("Instance is marked down or offline, fallback allowed," |
294 |
" switching to failover")
|
295 |
self.failover = True |
296 |
|
297 |
if instance.disk_template not in constants.DTS_MIRRORED: |
298 |
if self.failover: |
299 |
text = "failovers"
|
300 |
else:
|
301 |
text = "migrations"
|
302 |
raise errors.OpPrereqError("Instance's disk layout '%s' does not allow" |
303 |
" %s" % (instance.disk_template, text),
|
304 |
errors.ECODE_STATE) |
305 |
|
306 |
if instance.disk_template in constants.DTS_EXT_MIRROR: |
307 |
CheckIAllocatorOrNode(self.lu, "iallocator", "target_node") |
308 |
|
309 |
if self.lu.op.iallocator: |
310 |
assert locking.NAL in self.lu.owned_locks(locking.LEVEL_NODE_ALLOC) |
311 |
self._RunAllocator()
|
312 |
else:
|
313 |
# We set set self.target_node as it is required by
|
314 |
# BuildHooksEnv
|
315 |
self.target_node = self.lu.op.target_node |
316 |
|
317 |
# Check that the target node is correct in terms of instance policy
|
318 |
nodeinfo = self.cfg.GetNodeInfo(self.target_node) |
319 |
group_info = self.cfg.GetNodeGroup(nodeinfo.group)
|
320 |
ipolicy = ganeti.masterd.instance.CalculateGroupIPolicy(cluster, |
321 |
group_info) |
322 |
CheckTargetNodeIPolicy(self.lu, ipolicy, instance, nodeinfo, self.cfg, |
323 |
ignore=self.ignore_ipolicy)
|
324 |
|
325 |
# self.target_node is already populated, either directly or by the
|
326 |
# iallocator run
|
327 |
target_node = self.target_node
|
328 |
if self.target_node == instance.primary_node: |
329 |
raise errors.OpPrereqError("Cannot migrate instance %s" |
330 |
" to its primary (%s)" %
|
331 |
(instance.name, instance.primary_node), |
332 |
errors.ECODE_STATE) |
333 |
|
334 |
if len(self.lu.tasklets) == 1: |
335 |
# It is safe to release locks only when we're the only tasklet
|
336 |
# in the LU
|
337 |
ReleaseLocks(self.lu, locking.LEVEL_NODE,
|
338 |
keep=[instance.primary_node, self.target_node])
|
339 |
ReleaseLocks(self.lu, locking.LEVEL_NODE_ALLOC)
|
340 |
|
341 |
else:
|
342 |
assert not self.lu.glm.is_owned(locking.LEVEL_NODE_ALLOC) |
343 |
|
344 |
secondary_nodes = instance.secondary_nodes |
345 |
if not secondary_nodes: |
346 |
raise errors.ConfigurationError("No secondary node but using" |
347 |
" %s disk template" %
|
348 |
instance.disk_template) |
349 |
target_node = secondary_nodes[0]
|
350 |
if self.lu.op.iallocator or (self.lu.op.target_node and |
351 |
self.lu.op.target_node != target_node):
|
352 |
if self.failover: |
353 |
text = "failed over"
|
354 |
else:
|
355 |
text = "migrated"
|
356 |
raise errors.OpPrereqError("Instances with disk template %s cannot" |
357 |
" be %s to arbitrary nodes"
|
358 |
" (neither an iallocator nor a target"
|
359 |
" node can be passed)" %
|
360 |
(instance.disk_template, text), |
361 |
errors.ECODE_INVAL) |
362 |
nodeinfo = self.cfg.GetNodeInfo(target_node)
|
363 |
group_info = self.cfg.GetNodeGroup(nodeinfo.group)
|
364 |
ipolicy = ganeti.masterd.instance.CalculateGroupIPolicy(cluster, |
365 |
group_info) |
366 |
CheckTargetNodeIPolicy(self.lu, ipolicy, instance, nodeinfo, self.cfg, |
367 |
ignore=self.ignore_ipolicy)
|
368 |
|
369 |
i_be = cluster.FillBE(instance) |
370 |
|
371 |
# check memory requirements on the secondary node
|
372 |
if (not self.cleanup and |
373 |
(not self.failover or instance.admin_state == constants.ADMINST_UP)): |
374 |
self.tgt_free_mem = CheckNodeFreeMemory(self.lu, target_node, |
375 |
"migrating instance %s" %
|
376 |
instance.name, |
377 |
i_be[constants.BE_MINMEM], |
378 |
instance.hypervisor) |
379 |
else:
|
380 |
self.lu.LogInfo("Not checking memory on the secondary node as" |
381 |
" instance will not be started")
|
382 |
|
383 |
# check if failover must be forced instead of migration
|
384 |
if (not self.cleanup and not self.failover and |
385 |
i_be[constants.BE_ALWAYS_FAILOVER]): |
386 |
self.lu.LogInfo("Instance configured to always failover; fallback" |
387 |
" to failover")
|
388 |
self.failover = True |
389 |
|
390 |
# check bridge existance
|
391 |
CheckInstanceBridgesExist(self.lu, instance, node=target_node)
|
392 |
|
393 |
if not self.cleanup: |
394 |
CheckNodeNotDrained(self.lu, target_node)
|
395 |
if not self.failover: |
396 |
result = self.rpc.call_instance_migratable(instance.primary_node,
|
397 |
instance) |
398 |
if result.fail_msg and self.fallback: |
399 |
self.lu.LogInfo("Can't migrate, instance offline, fallback to" |
400 |
" failover")
|
401 |
self.failover = True |
402 |
else:
|
403 |
result.Raise("Can't migrate, please use failover",
|
404 |
prereq=True, ecode=errors.ECODE_STATE)
|
405 |
|
406 |
assert not (self.failover and self.cleanup) |
407 |
|
408 |
if not self.failover: |
409 |
if self.lu.op.live is not None and self.lu.op.mode is not None: |
410 |
raise errors.OpPrereqError("Only one of the 'live' and 'mode'" |
411 |
" parameters are accepted",
|
412 |
errors.ECODE_INVAL) |
413 |
if self.lu.op.live is not None: |
414 |
if self.lu.op.live: |
415 |
self.lu.op.mode = constants.HT_MIGRATION_LIVE
|
416 |
else:
|
417 |
self.lu.op.mode = constants.HT_MIGRATION_NONLIVE
|
418 |
# reset the 'live' parameter to None so that repeated
|
419 |
# invocations of CheckPrereq do not raise an exception
|
420 |
self.lu.op.live = None |
421 |
elif self.lu.op.mode is None: |
422 |
# read the default value from the hypervisor
|
423 |
i_hv = cluster.FillHV(self.instance, skip_globals=False) |
424 |
self.lu.op.mode = i_hv[constants.HV_MIGRATION_MODE]
|
425 |
|
426 |
self.live = self.lu.op.mode == constants.HT_MIGRATION_LIVE |
427 |
else:
|
428 |
# Failover is never live
|
429 |
self.live = False |
430 |
|
431 |
if not (self.failover or self.cleanup): |
432 |
remote_info = self.rpc.call_instance_info(instance.primary_node,
|
433 |
instance.name, |
434 |
instance.hypervisor) |
435 |
remote_info.Raise("Error checking instance on node %s" %
|
436 |
instance.primary_node) |
437 |
instance_running = bool(remote_info.payload)
|
438 |
if instance_running:
|
439 |
self.current_mem = int(remote_info.payload["memory"]) |
440 |
|
441 |
def _RunAllocator(self): |
442 |
"""Run the allocator based on input opcode.
|
443 |
|
444 |
"""
|
445 |
assert locking.NAL in self.lu.owned_locks(locking.LEVEL_NODE_ALLOC) |
446 |
|
447 |
# FIXME: add a self.ignore_ipolicy option
|
448 |
req = iallocator.IAReqRelocate(name=self.instance_name,
|
449 |
relocate_from=[self.instance.primary_node])
|
450 |
ial = iallocator.IAllocator(self.cfg, self.rpc, req) |
451 |
|
452 |
ial.Run(self.lu.op.iallocator)
|
453 |
|
454 |
if not ial.success: |
455 |
raise errors.OpPrereqError("Can't compute nodes using" |
456 |
" iallocator '%s': %s" %
|
457 |
(self.lu.op.iallocator, ial.info),
|
458 |
errors.ECODE_NORES) |
459 |
self.target_node = ial.result[0] |
460 |
self.lu.LogInfo("Selected nodes for instance %s via iallocator %s: %s", |
461 |
self.instance_name, self.lu.op.iallocator, |
462 |
utils.CommaJoin(ial.result)) |
463 |
|
464 |
def _WaitUntilSync(self): |
465 |
"""Poll with custom rpc for disk sync.
|
466 |
|
467 |
This uses our own step-based rpc call.
|
468 |
|
469 |
"""
|
470 |
self.feedback_fn("* wait until resync is done") |
471 |
all_done = False
|
472 |
while not all_done: |
473 |
all_done = True
|
474 |
result = self.rpc.call_drbd_wait_sync(self.all_nodes, |
475 |
self.nodes_ip,
|
476 |
(self.instance.disks,
|
477 |
self.instance))
|
478 |
min_percent = 100
|
479 |
for node, nres in result.items(): |
480 |
nres.Raise("Cannot resync disks on node %s" % node)
|
481 |
node_done, node_percent = nres.payload |
482 |
all_done = all_done and node_done
|
483 |
if node_percent is not None: |
484 |
min_percent = min(min_percent, node_percent)
|
485 |
if not all_done: |
486 |
if min_percent < 100: |
487 |
self.feedback_fn(" - progress: %.1f%%" % min_percent) |
488 |
time.sleep(2)
|
489 |
|
490 |
def _EnsureSecondary(self, node): |
491 |
"""Demote a node to secondary.
|
492 |
|
493 |
"""
|
494 |
self.feedback_fn("* switching node %s to secondary mode" % node) |
495 |
|
496 |
for dev in self.instance.disks: |
497 |
self.cfg.SetDiskID(dev, node)
|
498 |
|
499 |
result = self.rpc.call_blockdev_close(node, self.instance.name, |
500 |
self.instance.disks)
|
501 |
result.Raise("Cannot change disk to secondary on node %s" % node)
|
502 |
|
503 |
def _GoStandalone(self): |
504 |
"""Disconnect from the network.
|
505 |
|
506 |
"""
|
507 |
self.feedback_fn("* changing into standalone mode") |
508 |
result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip, |
509 |
self.instance.disks)
|
510 |
for node, nres in result.items(): |
511 |
nres.Raise("Cannot disconnect disks node %s" % node)
|
512 |
|
513 |
def _GoReconnect(self, multimaster): |
514 |
"""Reconnect to the network.
|
515 |
|
516 |
"""
|
517 |
if multimaster:
|
518 |
msg = "dual-master"
|
519 |
else:
|
520 |
msg = "single-master"
|
521 |
self.feedback_fn("* changing disks into %s mode" % msg) |
522 |
result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip, |
523 |
(self.instance.disks, self.instance), |
524 |
self.instance.name, multimaster)
|
525 |
for node, nres in result.items(): |
526 |
nres.Raise("Cannot change disks config on node %s" % node)
|
527 |
|
528 |
def _ExecCleanup(self): |
529 |
"""Try to cleanup after a failed migration.
|
530 |
|
531 |
The cleanup is done by:
|
532 |
- check that the instance is running only on one node
|
533 |
(and update the config if needed)
|
534 |
- change disks on its secondary node to secondary
|
535 |
- wait until disks are fully synchronized
|
536 |
- disconnect from the network
|
537 |
- change disks into single-master mode
|
538 |
- wait again until disks are fully synchronized
|
539 |
|
540 |
"""
|
541 |
instance = self.instance
|
542 |
target_node = self.target_node
|
543 |
source_node = self.source_node
|
544 |
|
545 |
# check running on only one node
|
546 |
self.feedback_fn("* checking where the instance actually runs" |
547 |
" (if this hangs, the hypervisor might be in"
|
548 |
" a bad state)")
|
549 |
ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor]) |
550 |
for node, result in ins_l.items(): |
551 |
result.Raise("Can't contact node %s" % node)
|
552 |
|
553 |
runningon_source = instance.name in ins_l[source_node].payload
|
554 |
runningon_target = instance.name in ins_l[target_node].payload
|
555 |
|
556 |
if runningon_source and runningon_target: |
557 |
raise errors.OpExecError("Instance seems to be running on two nodes," |
558 |
" or the hypervisor is confused; you will have"
|
559 |
" to ensure manually that it runs only on one"
|
560 |
" and restart this operation")
|
561 |
|
562 |
if not (runningon_source or runningon_target): |
563 |
raise errors.OpExecError("Instance does not seem to be running at all;" |
564 |
" in this case it's safer to repair by"
|
565 |
" running 'gnt-instance stop' to ensure disk"
|
566 |
" shutdown, and then restarting it")
|
567 |
|
568 |
if runningon_target:
|
569 |
# the migration has actually succeeded, we need to update the config
|
570 |
self.feedback_fn("* instance running on secondary node (%s)," |
571 |
" updating config" % target_node)
|
572 |
instance.primary_node = target_node |
573 |
self.cfg.Update(instance, self.feedback_fn) |
574 |
demoted_node = source_node |
575 |
else:
|
576 |
self.feedback_fn("* instance confirmed to be running on its" |
577 |
" primary node (%s)" % source_node)
|
578 |
demoted_node = target_node |
579 |
|
580 |
if instance.disk_template in constants.DTS_INT_MIRROR: |
581 |
self._EnsureSecondary(demoted_node)
|
582 |
try:
|
583 |
self._WaitUntilSync()
|
584 |
except errors.OpExecError:
|
585 |
# we ignore here errors, since if the device is standalone, it
|
586 |
# won't be able to sync
|
587 |
pass
|
588 |
self._GoStandalone()
|
589 |
self._GoReconnect(False) |
590 |
self._WaitUntilSync()
|
591 |
|
592 |
self.feedback_fn("* done") |
593 |
|
594 |
def _RevertDiskStatus(self): |
595 |
"""Try to revert the disk status after a failed migration.
|
596 |
|
597 |
"""
|
598 |
target_node = self.target_node
|
599 |
if self.instance.disk_template in constants.DTS_EXT_MIRROR: |
600 |
return
|
601 |
|
602 |
try:
|
603 |
self._EnsureSecondary(target_node)
|
604 |
self._GoStandalone()
|
605 |
self._GoReconnect(False) |
606 |
self._WaitUntilSync()
|
607 |
except errors.OpExecError, err:
|
608 |
self.lu.LogWarning("Migration failed and I can't reconnect the drives," |
609 |
" please try to recover the instance manually;"
|
610 |
" error '%s'" % str(err)) |
611 |
|
612 |
def _AbortMigration(self): |
613 |
"""Call the hypervisor code to abort a started migration.
|
614 |
|
615 |
"""
|
616 |
instance = self.instance
|
617 |
target_node = self.target_node
|
618 |
source_node = self.source_node
|
619 |
migration_info = self.migration_info
|
620 |
|
621 |
abort_result = self.rpc.call_instance_finalize_migration_dst(target_node,
|
622 |
instance, |
623 |
migration_info, |
624 |
False)
|
625 |
abort_msg = abort_result.fail_msg |
626 |
if abort_msg:
|
627 |
logging.error("Aborting migration failed on target node %s: %s",
|
628 |
target_node, abort_msg) |
629 |
# Don't raise an exception here, as we stil have to try to revert the
|
630 |
# disk status, even if this step failed.
|
631 |
|
632 |
abort_result = self.rpc.call_instance_finalize_migration_src(
|
633 |
source_node, instance, False, self.live) |
634 |
abort_msg = abort_result.fail_msg |
635 |
if abort_msg:
|
636 |
logging.error("Aborting migration failed on source node %s: %s",
|
637 |
source_node, abort_msg) |
638 |
|
639 |
def _ExecMigration(self): |
640 |
"""Migrate an instance.
|
641 |
|
642 |
The migrate is done by:
|
643 |
- change the disks into dual-master mode
|
644 |
- wait until disks are fully synchronized again
|
645 |
- migrate the instance
|
646 |
- change disks on the new secondary node (the old primary) to secondary
|
647 |
- wait until disks are fully synchronized
|
648 |
- change disks into single-master mode
|
649 |
|
650 |
"""
|
651 |
instance = self.instance
|
652 |
target_node = self.target_node
|
653 |
source_node = self.source_node
|
654 |
|
655 |
# Check for hypervisor version mismatch and warn the user.
|
656 |
nodeinfo = self.rpc.call_node_info([source_node, target_node],
|
657 |
None, [self.instance.hypervisor], False) |
658 |
for ninfo in nodeinfo.values(): |
659 |
ninfo.Raise("Unable to retrieve node information from node '%s'" %
|
660 |
ninfo.node) |
661 |
(_, _, (src_info, )) = nodeinfo[source_node].payload |
662 |
(_, _, (dst_info, )) = nodeinfo[target_node].payload |
663 |
|
664 |
if ((constants.HV_NODEINFO_KEY_VERSION in src_info) and |
665 |
(constants.HV_NODEINFO_KEY_VERSION in dst_info)):
|
666 |
src_version = src_info[constants.HV_NODEINFO_KEY_VERSION] |
667 |
dst_version = dst_info[constants.HV_NODEINFO_KEY_VERSION] |
668 |
if src_version != dst_version:
|
669 |
self.feedback_fn("* warning: hypervisor version mismatch between" |
670 |
" source (%s) and target (%s) node" %
|
671 |
(src_version, dst_version)) |
672 |
|
673 |
self.feedback_fn("* checking disk consistency between source and target") |
674 |
for (idx, dev) in enumerate(instance.disks): |
675 |
if not CheckDiskConsistency(self.lu, instance, dev, target_node, False): |
676 |
raise errors.OpExecError("Disk %s is degraded or not fully" |
677 |
" synchronized on target node,"
|
678 |
" aborting migration" % idx)
|
679 |
|
680 |
if self.current_mem > self.tgt_free_mem: |
681 |
if not self.allow_runtime_changes: |
682 |
raise errors.OpExecError("Memory ballooning not allowed and not enough" |
683 |
" free memory to fit instance %s on target"
|
684 |
" node %s (have %dMB, need %dMB)" %
|
685 |
(instance.name, target_node, |
686 |
self.tgt_free_mem, self.current_mem)) |
687 |
self.feedback_fn("* setting instance memory to %s" % self.tgt_free_mem) |
688 |
rpcres = self.rpc.call_instance_balloon_memory(instance.primary_node,
|
689 |
instance, |
690 |
self.tgt_free_mem)
|
691 |
rpcres.Raise("Cannot modify instance runtime memory")
|
692 |
|
693 |
# First get the migration information from the remote node
|
694 |
result = self.rpc.call_migration_info(source_node, instance)
|
695 |
msg = result.fail_msg |
696 |
if msg:
|
697 |
log_err = ("Failed fetching source migration information from %s: %s" %
|
698 |
(source_node, msg)) |
699 |
logging.error(log_err) |
700 |
raise errors.OpExecError(log_err)
|
701 |
|
702 |
self.migration_info = migration_info = result.payload
|
703 |
|
704 |
if self.instance.disk_template not in constants.DTS_EXT_MIRROR: |
705 |
# Then switch the disks to master/master mode
|
706 |
self._EnsureSecondary(target_node)
|
707 |
self._GoStandalone()
|
708 |
self._GoReconnect(True) |
709 |
self._WaitUntilSync()
|
710 |
|
711 |
self.feedback_fn("* preparing %s to accept the instance" % target_node) |
712 |
# This fills physical_id slot that may be missing on newly created disks
|
713 |
for disk in instance.disks: |
714 |
self.cfg.SetDiskID(disk, target_node)
|
715 |
result = self.rpc.call_accept_instance(target_node,
|
716 |
instance, |
717 |
migration_info, |
718 |
self.nodes_ip[target_node])
|
719 |
|
720 |
msg = result.fail_msg |
721 |
if msg:
|
722 |
logging.error("Instance pre-migration failed, trying to revert"
|
723 |
" disk status: %s", msg)
|
724 |
self.feedback_fn("Pre-migration failed, aborting") |
725 |
self._AbortMigration()
|
726 |
self._RevertDiskStatus()
|
727 |
raise errors.OpExecError("Could not pre-migrate instance %s: %s" % |
728 |
(instance.name, msg)) |
729 |
|
730 |
self.feedback_fn("* migrating instance to %s" % target_node) |
731 |
result = self.rpc.call_instance_migrate(source_node, instance,
|
732 |
self.nodes_ip[target_node],
|
733 |
self.live)
|
734 |
msg = result.fail_msg |
735 |
if msg:
|
736 |
logging.error("Instance migration failed, trying to revert"
|
737 |
" disk status: %s", msg)
|
738 |
self.feedback_fn("Migration failed, aborting") |
739 |
self._AbortMigration()
|
740 |
self._RevertDiskStatus()
|
741 |
raise errors.OpExecError("Could not migrate instance %s: %s" % |
742 |
(instance.name, msg)) |
743 |
|
744 |
self.feedback_fn("* starting memory transfer") |
745 |
last_feedback = time.time() |
746 |
while True: |
747 |
result = self.rpc.call_instance_get_migration_status(source_node,
|
748 |
instance) |
749 |
msg = result.fail_msg |
750 |
ms = result.payload # MigrationStatus instance
|
751 |
if msg or (ms.status in constants.HV_MIGRATION_FAILED_STATUSES): |
752 |
logging.error("Instance migration failed, trying to revert"
|
753 |
" disk status: %s", msg)
|
754 |
self.feedback_fn("Migration failed, aborting") |
755 |
self._AbortMigration()
|
756 |
self._RevertDiskStatus()
|
757 |
if not msg: |
758 |
msg = "hypervisor returned failure"
|
759 |
raise errors.OpExecError("Could not migrate instance %s: %s" % |
760 |
(instance.name, msg)) |
761 |
|
762 |
if result.payload.status != constants.HV_MIGRATION_ACTIVE:
|
763 |
self.feedback_fn("* memory transfer complete") |
764 |
break
|
765 |
|
766 |
if (utils.TimeoutExpired(last_feedback,
|
767 |
self._MIGRATION_FEEDBACK_INTERVAL) and |
768 |
ms.transferred_ram is not None): |
769 |
mem_progress = 100 * float(ms.transferred_ram) / float(ms.total_ram) |
770 |
self.feedback_fn("* memory transfer progress: %.2f %%" % mem_progress) |
771 |
last_feedback = time.time() |
772 |
|
773 |
time.sleep(self._MIGRATION_POLL_INTERVAL)
|
774 |
|
775 |
result = self.rpc.call_instance_finalize_migration_src(source_node,
|
776 |
instance, |
777 |
True,
|
778 |
self.live)
|
779 |
msg = result.fail_msg |
780 |
if msg:
|
781 |
logging.error("Instance migration succeeded, but finalization failed"
|
782 |
" on the source node: %s", msg)
|
783 |
raise errors.OpExecError("Could not finalize instance migration: %s" % |
784 |
msg) |
785 |
|
786 |
instance.primary_node = target_node |
787 |
|
788 |
# distribute new instance config to the other nodes
|
789 |
self.cfg.Update(instance, self.feedback_fn) |
790 |
|
791 |
result = self.rpc.call_instance_finalize_migration_dst(target_node,
|
792 |
instance, |
793 |
migration_info, |
794 |
True)
|
795 |
msg = result.fail_msg |
796 |
if msg:
|
797 |
logging.error("Instance migration succeeded, but finalization failed"
|
798 |
" on the target node: %s", msg)
|
799 |
raise errors.OpExecError("Could not finalize instance migration: %s" % |
800 |
msg) |
801 |
|
802 |
if self.instance.disk_template not in constants.DTS_EXT_MIRROR: |
803 |
self._EnsureSecondary(source_node)
|
804 |
self._WaitUntilSync()
|
805 |
self._GoStandalone()
|
806 |
self._GoReconnect(False) |
807 |
self._WaitUntilSync()
|
808 |
|
809 |
# If the instance's disk template is `rbd' or `ext' and there was a
|
810 |
# successful migration, unmap the device from the source node.
|
811 |
if self.instance.disk_template in (constants.DT_RBD, constants.DT_EXT): |
812 |
disks = ExpandCheckDisks(instance, instance.disks) |
813 |
self.feedback_fn("* unmapping instance's disks from %s" % source_node) |
814 |
for disk in disks: |
815 |
result = self.rpc.call_blockdev_shutdown(source_node, (disk, instance))
|
816 |
msg = result.fail_msg |
817 |
if msg:
|
818 |
logging.error("Migration was successful, but couldn't unmap the"
|
819 |
" block device %s on source node %s: %s",
|
820 |
disk.iv_name, source_node, msg) |
821 |
logging.error("You need to unmap the device %s manually on %s",
|
822 |
disk.iv_name, source_node) |
823 |
|
824 |
self.feedback_fn("* done") |
825 |
|
826 |
def _ExecFailover(self): |
827 |
"""Failover an instance.
|
828 |
|
829 |
The failover is done by shutting it down on its present node and
|
830 |
starting it on the secondary.
|
831 |
|
832 |
"""
|
833 |
instance = self.instance
|
834 |
primary_node = self.cfg.GetNodeInfo(instance.primary_node)
|
835 |
|
836 |
source_node = instance.primary_node |
837 |
target_node = self.target_node
|
838 |
|
839 |
if instance.disks_active:
|
840 |
self.feedback_fn("* checking disk consistency between source and target") |
841 |
for (idx, dev) in enumerate(instance.disks): |
842 |
# for drbd, these are drbd over lvm
|
843 |
if not CheckDiskConsistency(self.lu, instance, dev, target_node, |
844 |
False):
|
845 |
if primary_node.offline:
|
846 |
self.feedback_fn("Node %s is offline, ignoring degraded disk %s on" |
847 |
" target node %s" %
|
848 |
(primary_node.name, idx, target_node)) |
849 |
elif not self.ignore_consistency: |
850 |
raise errors.OpExecError("Disk %s is degraded on target node," |
851 |
" aborting failover" % idx)
|
852 |
else:
|
853 |
self.feedback_fn("* not checking disk consistency as instance is not" |
854 |
" running")
|
855 |
|
856 |
self.feedback_fn("* shutting down instance on source node") |
857 |
logging.info("Shutting down instance %s on node %s",
|
858 |
instance.name, source_node) |
859 |
|
860 |
result = self.rpc.call_instance_shutdown(source_node, instance,
|
861 |
self.shutdown_timeout,
|
862 |
self.lu.op.reason)
|
863 |
msg = result.fail_msg |
864 |
if msg:
|
865 |
if self.ignore_consistency or primary_node.offline: |
866 |
self.lu.LogWarning("Could not shutdown instance %s on node %s," |
867 |
" proceeding anyway; please make sure node"
|
868 |
" %s is down; error details: %s",
|
869 |
instance.name, source_node, source_node, msg) |
870 |
else:
|
871 |
raise errors.OpExecError("Could not shutdown instance %s on" |
872 |
" node %s: %s" %
|
873 |
(instance.name, source_node, msg)) |
874 |
|
875 |
self.feedback_fn("* deactivating the instance's disks on source node") |
876 |
if not ShutdownInstanceDisks(self.lu, instance, ignore_primary=True): |
877 |
raise errors.OpExecError("Can't shut down the instance's disks") |
878 |
|
879 |
instance.primary_node = target_node |
880 |
# distribute new instance config to the other nodes
|
881 |
self.cfg.Update(instance, self.feedback_fn) |
882 |
|
883 |
# Only start the instance if it's marked as up
|
884 |
if instance.admin_state == constants.ADMINST_UP:
|
885 |
self.feedback_fn("* activating the instance's disks on target node %s" % |
886 |
target_node) |
887 |
logging.info("Starting instance %s on node %s",
|
888 |
instance.name, target_node) |
889 |
|
890 |
disks_ok, _ = AssembleInstanceDisks(self.lu, instance,
|
891 |
ignore_secondaries=True)
|
892 |
if not disks_ok: |
893 |
ShutdownInstanceDisks(self.lu, instance)
|
894 |
raise errors.OpExecError("Can't activate the instance's disks") |
895 |
|
896 |
self.feedback_fn("* starting the instance on the target node %s" % |
897 |
target_node) |
898 |
result = self.rpc.call_instance_start(target_node, (instance, None, None), |
899 |
False, self.lu.op.reason) |
900 |
msg = result.fail_msg |
901 |
if msg:
|
902 |
ShutdownInstanceDisks(self.lu, instance)
|
903 |
raise errors.OpExecError("Could not start instance %s on node %s: %s" % |
904 |
(instance.name, target_node, msg)) |
905 |
|
906 |
def Exec(self, feedback_fn): |
907 |
"""Perform the migration.
|
908 |
|
909 |
"""
|
910 |
self.feedback_fn = feedback_fn
|
911 |
self.source_node = self.instance.primary_node |
912 |
|
913 |
# FIXME: if we implement migrate-to-any in DRBD, this needs fixing
|
914 |
if self.instance.disk_template in constants.DTS_INT_MIRROR: |
915 |
self.target_node = self.instance.secondary_nodes[0] |
916 |
# Otherwise self.target_node has been populated either
|
917 |
# directly, or through an iallocator.
|
918 |
|
919 |
self.all_nodes = [self.source_node, self.target_node] |
920 |
self.nodes_ip = dict((name, node.secondary_ip) for (name, node) |
921 |
in self.cfg.GetMultiNodeInfo(self.all_nodes)) |
922 |
|
923 |
if self.failover: |
924 |
feedback_fn("Failover instance %s" % self.instance.name) |
925 |
self._ExecFailover()
|
926 |
else:
|
927 |
feedback_fn("Migrating instance %s" % self.instance.name) |
928 |
|
929 |
if self.cleanup: |
930 |
return self._ExecCleanup() |
931 |
else:
|
932 |
return self._ExecMigration() |