root / lib / cmdlib.py @ 7b2cd2b4
History | View | Annotate | Download (356.3 kB)
1 |
#
|
---|---|
2 |
#
|
3 |
|
4 |
# Copyright (C) 2006, 2007, 2008 Google Inc.
|
5 |
#
|
6 |
# This program is free software; you can redistribute it and/or modify
|
7 |
# it under the terms of the GNU General Public License as published by
|
8 |
# the Free Software Foundation; either version 2 of the License, or
|
9 |
# (at your option) any later version.
|
10 |
#
|
11 |
# This program is distributed in the hope that it will be useful, but
|
12 |
# WITHOUT ANY WARRANTY; without even the implied warranty of
|
13 |
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
14 |
# General Public License for more details.
|
15 |
#
|
16 |
# You should have received a copy of the GNU General Public License
|
17 |
# along with this program; if not, write to the Free Software
|
18 |
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
|
19 |
# 02110-1301, USA.
|
20 |
|
21 |
|
22 |
"""Module implementing the master-side code."""
|
23 |
|
24 |
# pylint: disable-msg=W0201,C0302
|
25 |
|
26 |
# W0201 since most LU attributes are defined in CheckPrereq or similar
|
27 |
# functions
|
28 |
|
29 |
# C0302: since we have waaaay to many lines in this module
|
30 |
|
31 |
import os |
32 |
import os.path |
33 |
import time |
34 |
import re |
35 |
import platform |
36 |
import logging |
37 |
import copy |
38 |
import OpenSSL |
39 |
|
40 |
from ganeti import ssh |
41 |
from ganeti import utils |
42 |
from ganeti import errors |
43 |
from ganeti import hypervisor |
44 |
from ganeti import locking |
45 |
from ganeti import constants |
46 |
from ganeti import objects |
47 |
from ganeti import serializer |
48 |
from ganeti import ssconf |
49 |
from ganeti import uidpool |
50 |
from ganeti import compat |
51 |
from ganeti import masterd |
52 |
|
53 |
import ganeti.masterd.instance # pylint: disable-msg=W0611 |
54 |
|
55 |
|
56 |
# Modifiable default values; need to define these here before the
|
57 |
# actual LUs
|
58 |
|
59 |
def _EmptyList(): |
60 |
"""Returns an empty list.
|
61 |
|
62 |
"""
|
63 |
return []
|
64 |
|
65 |
|
66 |
def _EmptyDict(): |
67 |
"""Returns an empty dict.
|
68 |
|
69 |
"""
|
70 |
return {}
|
71 |
|
72 |
|
73 |
#: The without-default default value
|
74 |
_NoDefault = object()
|
75 |
|
76 |
|
77 |
#: The no-type (value to complex to check it in the type system)
|
78 |
_NoType = object()
|
79 |
|
80 |
|
81 |
# Some basic types
|
82 |
def _TNotNone(val): |
83 |
"""Checks if the given value is not None.
|
84 |
|
85 |
"""
|
86 |
return val is not None |
87 |
|
88 |
|
89 |
def _TNone(val): |
90 |
"""Checks if the given value is None.
|
91 |
|
92 |
"""
|
93 |
return val is None |
94 |
|
95 |
|
96 |
def _TBool(val): |
97 |
"""Checks if the given value is a boolean.
|
98 |
|
99 |
"""
|
100 |
return isinstance(val, bool) |
101 |
|
102 |
|
103 |
def _TInt(val): |
104 |
"""Checks if the given value is an integer.
|
105 |
|
106 |
"""
|
107 |
return isinstance(val, int) |
108 |
|
109 |
|
110 |
def _TFloat(val): |
111 |
"""Checks if the given value is a float.
|
112 |
|
113 |
"""
|
114 |
return isinstance(val, float) |
115 |
|
116 |
|
117 |
def _TString(val): |
118 |
"""Checks if the given value is a string.
|
119 |
|
120 |
"""
|
121 |
return isinstance(val, basestring) |
122 |
|
123 |
|
124 |
def _TTrue(val): |
125 |
"""Checks if a given value evaluates to a boolean True value.
|
126 |
|
127 |
"""
|
128 |
return bool(val) |
129 |
|
130 |
|
131 |
def _TElemOf(target_list): |
132 |
"""Builds a function that checks if a given value is a member of a list.
|
133 |
|
134 |
"""
|
135 |
return lambda val: val in target_list |
136 |
|
137 |
|
138 |
# Container types
|
139 |
def _TList(val): |
140 |
"""Checks if the given value is a list.
|
141 |
|
142 |
"""
|
143 |
return isinstance(val, list) |
144 |
|
145 |
|
146 |
def _TDict(val): |
147 |
"""Checks if the given value is a dictionary.
|
148 |
|
149 |
"""
|
150 |
return isinstance(val, dict) |
151 |
|
152 |
|
153 |
# Combinator types
|
154 |
def _TAnd(*args): |
155 |
"""Combine multiple functions using an AND operation.
|
156 |
|
157 |
"""
|
158 |
def fn(val): |
159 |
return compat.all(t(val) for t in args) |
160 |
return fn
|
161 |
|
162 |
|
163 |
def _TOr(*args): |
164 |
"""Combine multiple functions using an AND operation.
|
165 |
|
166 |
"""
|
167 |
def fn(val): |
168 |
return compat.any(t(val) for t in args) |
169 |
return fn
|
170 |
|
171 |
|
172 |
# Type aliases
|
173 |
|
174 |
#: a non-empty string
|
175 |
_TNonEmptyString = _TAnd(_TString, _TTrue) |
176 |
|
177 |
|
178 |
#: a maybe non-empty string
|
179 |
_TMaybeString = _TOr(_TNonEmptyString, _TNone) |
180 |
|
181 |
|
182 |
#: a maybe boolean (bool or none)
|
183 |
_TMaybeBool = _TOr(_TBool, _TNone) |
184 |
|
185 |
|
186 |
#: a positive integer
|
187 |
_TPositiveInt = _TAnd(_TInt, lambda v: v >= 0) |
188 |
|
189 |
#: a strictly positive integer
|
190 |
_TStrictPositiveInt = _TAnd(_TInt, lambda v: v > 0) |
191 |
|
192 |
|
193 |
def _TListOf(my_type): |
194 |
"""Checks if a given value is a list with all elements of the same type.
|
195 |
|
196 |
"""
|
197 |
return _TAnd(_TList,
|
198 |
lambda lst: compat.all(my_type(v) for v in lst)) |
199 |
|
200 |
|
201 |
def _TDictOf(key_type, val_type): |
202 |
"""Checks a dict type for the type of its key/values.
|
203 |
|
204 |
"""
|
205 |
return _TAnd(_TDict,
|
206 |
lambda my_dict: (compat.all(key_type(v) for v in my_dict.keys()) |
207 |
and compat.all(val_type(v)
|
208 |
for v in my_dict.values()))) |
209 |
|
210 |
|
211 |
# Common opcode attributes
|
212 |
|
213 |
#: output fields for a query operation
|
214 |
_POutputFields = ("output_fields", _NoDefault, _TListOf(_TNonEmptyString))
|
215 |
|
216 |
|
217 |
#: the shutdown timeout
|
218 |
_PShutdownTimeout = ("shutdown_timeout", constants.DEFAULT_SHUTDOWN_TIMEOUT,
|
219 |
_TPositiveInt) |
220 |
|
221 |
#: the force parameter
|
222 |
_PForce = ("force", False, _TBool) |
223 |
|
224 |
#: a required instance name (for single-instance LUs)
|
225 |
_PInstanceName = ("instance_name", _NoDefault, _TNonEmptyString)
|
226 |
|
227 |
|
228 |
#: a required node name (for single-node LUs)
|
229 |
_PNodeName = ("node_name", _NoDefault, _TNonEmptyString)
|
230 |
|
231 |
|
232 |
# End types
|
233 |
class LogicalUnit(object): |
234 |
"""Logical Unit base class.
|
235 |
|
236 |
Subclasses must follow these rules:
|
237 |
- implement ExpandNames
|
238 |
- implement CheckPrereq (except when tasklets are used)
|
239 |
- implement Exec (except when tasklets are used)
|
240 |
- implement BuildHooksEnv
|
241 |
- redefine HPATH and HTYPE
|
242 |
- optionally redefine their run requirements:
|
243 |
REQ_BGL: the LU needs to hold the Big Ganeti Lock exclusively
|
244 |
|
245 |
Note that all commands require root permissions.
|
246 |
|
247 |
@ivar dry_run_result: the value (if any) that will be returned to the caller
|
248 |
in dry-run mode (signalled by opcode dry_run parameter)
|
249 |
@cvar _OP_PARAMS: a list of opcode attributes, their defaults values
|
250 |
they should get if not already defined, and types they must match
|
251 |
|
252 |
"""
|
253 |
HPATH = None
|
254 |
HTYPE = None
|
255 |
_OP_PARAMS = [] |
256 |
REQ_BGL = True
|
257 |
|
258 |
def __init__(self, processor, op, context, rpc): |
259 |
"""Constructor for LogicalUnit.
|
260 |
|
261 |
This needs to be overridden in derived classes in order to check op
|
262 |
validity.
|
263 |
|
264 |
"""
|
265 |
self.proc = processor
|
266 |
self.op = op
|
267 |
self.cfg = context.cfg
|
268 |
self.context = context
|
269 |
self.rpc = rpc
|
270 |
# Dicts used to declare locking needs to mcpu
|
271 |
self.needed_locks = None |
272 |
self.acquired_locks = {}
|
273 |
self.share_locks = dict.fromkeys(locking.LEVELS, 0) |
274 |
self.add_locks = {}
|
275 |
self.remove_locks = {}
|
276 |
# Used to force good behavior when calling helper functions
|
277 |
self.recalculate_locks = {}
|
278 |
self.__ssh = None |
279 |
# logging
|
280 |
self.LogWarning = processor.LogWarning # pylint: disable-msg=C0103 |
281 |
self.LogInfo = processor.LogInfo # pylint: disable-msg=C0103 |
282 |
self.LogStep = processor.LogStep # pylint: disable-msg=C0103 |
283 |
# support for dry-run
|
284 |
self.dry_run_result = None |
285 |
# support for generic debug attribute
|
286 |
if (not hasattr(self.op, "debug_level") or |
287 |
not isinstance(self.op.debug_level, int)): |
288 |
self.op.debug_level = 0 |
289 |
|
290 |
# Tasklets
|
291 |
self.tasklets = None |
292 |
|
293 |
# The new kind-of-type-system
|
294 |
op_id = self.op.OP_ID
|
295 |
for attr_name, aval, test in self._OP_PARAMS: |
296 |
if not hasattr(op, attr_name): |
297 |
if aval == _NoDefault:
|
298 |
raise errors.OpPrereqError("Required parameter '%s.%s' missing" % |
299 |
(op_id, attr_name), errors.ECODE_INVAL) |
300 |
else:
|
301 |
if callable(aval): |
302 |
dval = aval() |
303 |
else:
|
304 |
dval = aval |
305 |
setattr(self.op, attr_name, dval) |
306 |
attr_val = getattr(op, attr_name)
|
307 |
if test == _NoType:
|
308 |
# no tests here
|
309 |
continue
|
310 |
if not callable(test): |
311 |
raise errors.ProgrammerError("Validation for parameter '%s.%s' failed," |
312 |
" given type is not a proper type (%s)" %
|
313 |
(op_id, attr_name, test)) |
314 |
if not test(attr_val): |
315 |
logging.error("OpCode %s, parameter %s, has invalid type %s/value %s",
|
316 |
self.op.OP_ID, attr_name, type(attr_val), attr_val) |
317 |
raise errors.OpPrereqError("Parameter '%s.%s' fails validation" % |
318 |
(op_id, attr_name), errors.ECODE_INVAL) |
319 |
|
320 |
self.CheckArguments()
|
321 |
|
322 |
def __GetSSH(self): |
323 |
"""Returns the SshRunner object
|
324 |
|
325 |
"""
|
326 |
if not self.__ssh: |
327 |
self.__ssh = ssh.SshRunner(self.cfg.GetClusterName()) |
328 |
return self.__ssh |
329 |
|
330 |
ssh = property(fget=__GetSSH)
|
331 |
|
332 |
def CheckArguments(self): |
333 |
"""Check syntactic validity for the opcode arguments.
|
334 |
|
335 |
This method is for doing a simple syntactic check and ensure
|
336 |
validity of opcode parameters, without any cluster-related
|
337 |
checks. While the same can be accomplished in ExpandNames and/or
|
338 |
CheckPrereq, doing these separate is better because:
|
339 |
|
340 |
- ExpandNames is left as as purely a lock-related function
|
341 |
- CheckPrereq is run after we have acquired locks (and possible
|
342 |
waited for them)
|
343 |
|
344 |
The function is allowed to change the self.op attribute so that
|
345 |
later methods can no longer worry about missing parameters.
|
346 |
|
347 |
"""
|
348 |
pass
|
349 |
|
350 |
def ExpandNames(self): |
351 |
"""Expand names for this LU.
|
352 |
|
353 |
This method is called before starting to execute the opcode, and it should
|
354 |
update all the parameters of the opcode to their canonical form (e.g. a
|
355 |
short node name must be fully expanded after this method has successfully
|
356 |
completed). This way locking, hooks, logging, ecc. can work correctly.
|
357 |
|
358 |
LUs which implement this method must also populate the self.needed_locks
|
359 |
member, as a dict with lock levels as keys, and a list of needed lock names
|
360 |
as values. Rules:
|
361 |
|
362 |
- use an empty dict if you don't need any lock
|
363 |
- if you don't need any lock at a particular level omit that level
|
364 |
- don't put anything for the BGL level
|
365 |
- if you want all locks at a level use locking.ALL_SET as a value
|
366 |
|
367 |
If you need to share locks (rather than acquire them exclusively) at one
|
368 |
level you can modify self.share_locks, setting a true value (usually 1) for
|
369 |
that level. By default locks are not shared.
|
370 |
|
371 |
This function can also define a list of tasklets, which then will be
|
372 |
executed in order instead of the usual LU-level CheckPrereq and Exec
|
373 |
functions, if those are not defined by the LU.
|
374 |
|
375 |
Examples::
|
376 |
|
377 |
# Acquire all nodes and one instance
|
378 |
self.needed_locks = {
|
379 |
locking.LEVEL_NODE: locking.ALL_SET,
|
380 |
locking.LEVEL_INSTANCE: ['instance1.example.tld'],
|
381 |
}
|
382 |
# Acquire just two nodes
|
383 |
self.needed_locks = {
|
384 |
locking.LEVEL_NODE: ['node1.example.tld', 'node2.example.tld'],
|
385 |
}
|
386 |
# Acquire no locks
|
387 |
self.needed_locks = {} # No, you can't leave it to the default value None
|
388 |
|
389 |
"""
|
390 |
# The implementation of this method is mandatory only if the new LU is
|
391 |
# concurrent, so that old LUs don't need to be changed all at the same
|
392 |
# time.
|
393 |
if self.REQ_BGL: |
394 |
self.needed_locks = {} # Exclusive LUs don't need locks. |
395 |
else:
|
396 |
raise NotImplementedError |
397 |
|
398 |
def DeclareLocks(self, level): |
399 |
"""Declare LU locking needs for a level
|
400 |
|
401 |
While most LUs can just declare their locking needs at ExpandNames time,
|
402 |
sometimes there's the need to calculate some locks after having acquired
|
403 |
the ones before. This function is called just before acquiring locks at a
|
404 |
particular level, but after acquiring the ones at lower levels, and permits
|
405 |
such calculations. It can be used to modify self.needed_locks, and by
|
406 |
default it does nothing.
|
407 |
|
408 |
This function is only called if you have something already set in
|
409 |
self.needed_locks for the level.
|
410 |
|
411 |
@param level: Locking level which is going to be locked
|
412 |
@type level: member of ganeti.locking.LEVELS
|
413 |
|
414 |
"""
|
415 |
|
416 |
def CheckPrereq(self): |
417 |
"""Check prerequisites for this LU.
|
418 |
|
419 |
This method should check that the prerequisites for the execution
|
420 |
of this LU are fulfilled. It can do internode communication, but
|
421 |
it should be idempotent - no cluster or system changes are
|
422 |
allowed.
|
423 |
|
424 |
The method should raise errors.OpPrereqError in case something is
|
425 |
not fulfilled. Its return value is ignored.
|
426 |
|
427 |
This method should also update all the parameters of the opcode to
|
428 |
their canonical form if it hasn't been done by ExpandNames before.
|
429 |
|
430 |
"""
|
431 |
if self.tasklets is not None: |
432 |
for (idx, tl) in enumerate(self.tasklets): |
433 |
logging.debug("Checking prerequisites for tasklet %s/%s",
|
434 |
idx + 1, len(self.tasklets)) |
435 |
tl.CheckPrereq() |
436 |
else:
|
437 |
pass
|
438 |
|
439 |
def Exec(self, feedback_fn): |
440 |
"""Execute the LU.
|
441 |
|
442 |
This method should implement the actual work. It should raise
|
443 |
errors.OpExecError for failures that are somewhat dealt with in
|
444 |
code, or expected.
|
445 |
|
446 |
"""
|
447 |
if self.tasklets is not None: |
448 |
for (idx, tl) in enumerate(self.tasklets): |
449 |
logging.debug("Executing tasklet %s/%s", idx + 1, len(self.tasklets)) |
450 |
tl.Exec(feedback_fn) |
451 |
else:
|
452 |
raise NotImplementedError |
453 |
|
454 |
def BuildHooksEnv(self): |
455 |
"""Build hooks environment for this LU.
|
456 |
|
457 |
This method should return a three-node tuple consisting of: a dict
|
458 |
containing the environment that will be used for running the
|
459 |
specific hook for this LU, a list of node names on which the hook
|
460 |
should run before the execution, and a list of node names on which
|
461 |
the hook should run after the execution.
|
462 |
|
463 |
The keys of the dict must not have 'GANETI_' prefixed as this will
|
464 |
be handled in the hooks runner. Also note additional keys will be
|
465 |
added by the hooks runner. If the LU doesn't define any
|
466 |
environment, an empty dict (and not None) should be returned.
|
467 |
|
468 |
No nodes should be returned as an empty list (and not None).
|
469 |
|
470 |
Note that if the HPATH for a LU class is None, this function will
|
471 |
not be called.
|
472 |
|
473 |
"""
|
474 |
raise NotImplementedError |
475 |
|
476 |
def HooksCallBack(self, phase, hook_results, feedback_fn, lu_result): |
477 |
"""Notify the LU about the results of its hooks.
|
478 |
|
479 |
This method is called every time a hooks phase is executed, and notifies
|
480 |
the Logical Unit about the hooks' result. The LU can then use it to alter
|
481 |
its result based on the hooks. By default the method does nothing and the
|
482 |
previous result is passed back unchanged but any LU can define it if it
|
483 |
wants to use the local cluster hook-scripts somehow.
|
484 |
|
485 |
@param phase: one of L{constants.HOOKS_PHASE_POST} or
|
486 |
L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
|
487 |
@param hook_results: the results of the multi-node hooks rpc call
|
488 |
@param feedback_fn: function used send feedback back to the caller
|
489 |
@param lu_result: the previous Exec result this LU had, or None
|
490 |
in the PRE phase
|
491 |
@return: the new Exec result, based on the previous result
|
492 |
and hook results
|
493 |
|
494 |
"""
|
495 |
# API must be kept, thus we ignore the unused argument and could
|
496 |
# be a function warnings
|
497 |
# pylint: disable-msg=W0613,R0201
|
498 |
return lu_result
|
499 |
|
500 |
def _ExpandAndLockInstance(self): |
501 |
"""Helper function to expand and lock an instance.
|
502 |
|
503 |
Many LUs that work on an instance take its name in self.op.instance_name
|
504 |
and need to expand it and then declare the expanded name for locking. This
|
505 |
function does it, and then updates self.op.instance_name to the expanded
|
506 |
name. It also initializes needed_locks as a dict, if this hasn't been done
|
507 |
before.
|
508 |
|
509 |
"""
|
510 |
if self.needed_locks is None: |
511 |
self.needed_locks = {}
|
512 |
else:
|
513 |
assert locking.LEVEL_INSTANCE not in self.needed_locks, \ |
514 |
"_ExpandAndLockInstance called with instance-level locks set"
|
515 |
self.op.instance_name = _ExpandInstanceName(self.cfg, |
516 |
self.op.instance_name)
|
517 |
self.needed_locks[locking.LEVEL_INSTANCE] = self.op.instance_name |
518 |
|
519 |
def _LockInstancesNodes(self, primary_only=False): |
520 |
"""Helper function to declare instances' nodes for locking.
|
521 |
|
522 |
This function should be called after locking one or more instances to lock
|
523 |
their nodes. Its effect is populating self.needed_locks[locking.LEVEL_NODE]
|
524 |
with all primary or secondary nodes for instances already locked and
|
525 |
present in self.needed_locks[locking.LEVEL_INSTANCE].
|
526 |
|
527 |
It should be called from DeclareLocks, and for safety only works if
|
528 |
self.recalculate_locks[locking.LEVEL_NODE] is set.
|
529 |
|
530 |
In the future it may grow parameters to just lock some instance's nodes, or
|
531 |
to just lock primaries or secondary nodes, if needed.
|
532 |
|
533 |
If should be called in DeclareLocks in a way similar to::
|
534 |
|
535 |
if level == locking.LEVEL_NODE:
|
536 |
self._LockInstancesNodes()
|
537 |
|
538 |
@type primary_only: boolean
|
539 |
@param primary_only: only lock primary nodes of locked instances
|
540 |
|
541 |
"""
|
542 |
assert locking.LEVEL_NODE in self.recalculate_locks, \ |
543 |
"_LockInstancesNodes helper function called with no nodes to recalculate"
|
544 |
|
545 |
# TODO: check if we're really been called with the instance locks held
|
546 |
|
547 |
# For now we'll replace self.needed_locks[locking.LEVEL_NODE], but in the
|
548 |
# future we might want to have different behaviors depending on the value
|
549 |
# of self.recalculate_locks[locking.LEVEL_NODE]
|
550 |
wanted_nodes = [] |
551 |
for instance_name in self.acquired_locks[locking.LEVEL_INSTANCE]: |
552 |
instance = self.context.cfg.GetInstanceInfo(instance_name)
|
553 |
wanted_nodes.append(instance.primary_node) |
554 |
if not primary_only: |
555 |
wanted_nodes.extend(instance.secondary_nodes) |
556 |
|
557 |
if self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_REPLACE: |
558 |
self.needed_locks[locking.LEVEL_NODE] = wanted_nodes
|
559 |
elif self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_APPEND: |
560 |
self.needed_locks[locking.LEVEL_NODE].extend(wanted_nodes)
|
561 |
|
562 |
del self.recalculate_locks[locking.LEVEL_NODE] |
563 |
|
564 |
|
565 |
class NoHooksLU(LogicalUnit): # pylint: disable-msg=W0223 |
566 |
"""Simple LU which runs no hooks.
|
567 |
|
568 |
This LU is intended as a parent for other LogicalUnits which will
|
569 |
run no hooks, in order to reduce duplicate code.
|
570 |
|
571 |
"""
|
572 |
HPATH = None
|
573 |
HTYPE = None
|
574 |
|
575 |
def BuildHooksEnv(self): |
576 |
"""Empty BuildHooksEnv for NoHooksLu.
|
577 |
|
578 |
This just raises an error.
|
579 |
|
580 |
"""
|
581 |
assert False, "BuildHooksEnv called for NoHooksLUs" |
582 |
|
583 |
|
584 |
class Tasklet: |
585 |
"""Tasklet base class.
|
586 |
|
587 |
Tasklets are subcomponents for LUs. LUs can consist entirely of tasklets or
|
588 |
they can mix legacy code with tasklets. Locking needs to be done in the LU,
|
589 |
tasklets know nothing about locks.
|
590 |
|
591 |
Subclasses must follow these rules:
|
592 |
- Implement CheckPrereq
|
593 |
- Implement Exec
|
594 |
|
595 |
"""
|
596 |
def __init__(self, lu): |
597 |
self.lu = lu
|
598 |
|
599 |
# Shortcuts
|
600 |
self.cfg = lu.cfg
|
601 |
self.rpc = lu.rpc
|
602 |
|
603 |
def CheckPrereq(self): |
604 |
"""Check prerequisites for this tasklets.
|
605 |
|
606 |
This method should check whether the prerequisites for the execution of
|
607 |
this tasklet are fulfilled. It can do internode communication, but it
|
608 |
should be idempotent - no cluster or system changes are allowed.
|
609 |
|
610 |
The method should raise errors.OpPrereqError in case something is not
|
611 |
fulfilled. Its return value is ignored.
|
612 |
|
613 |
This method should also update all parameters to their canonical form if it
|
614 |
hasn't been done before.
|
615 |
|
616 |
"""
|
617 |
pass
|
618 |
|
619 |
def Exec(self, feedback_fn): |
620 |
"""Execute the tasklet.
|
621 |
|
622 |
This method should implement the actual work. It should raise
|
623 |
errors.OpExecError for failures that are somewhat dealt with in code, or
|
624 |
expected.
|
625 |
|
626 |
"""
|
627 |
raise NotImplementedError |
628 |
|
629 |
|
630 |
def _GetWantedNodes(lu, nodes): |
631 |
"""Returns list of checked and expanded node names.
|
632 |
|
633 |
@type lu: L{LogicalUnit}
|
634 |
@param lu: the logical unit on whose behalf we execute
|
635 |
@type nodes: list
|
636 |
@param nodes: list of node names or None for all nodes
|
637 |
@rtype: list
|
638 |
@return: the list of nodes, sorted
|
639 |
@raise errors.ProgrammerError: if the nodes parameter is wrong type
|
640 |
|
641 |
"""
|
642 |
if not nodes: |
643 |
raise errors.ProgrammerError("_GetWantedNodes should only be called with a" |
644 |
" non-empty list of nodes whose name is to be expanded.")
|
645 |
|
646 |
wanted = [_ExpandNodeName(lu.cfg, name) for name in nodes] |
647 |
return utils.NiceSort(wanted)
|
648 |
|
649 |
|
650 |
def _GetWantedInstances(lu, instances): |
651 |
"""Returns list of checked and expanded instance names.
|
652 |
|
653 |
@type lu: L{LogicalUnit}
|
654 |
@param lu: the logical unit on whose behalf we execute
|
655 |
@type instances: list
|
656 |
@param instances: list of instance names or None for all instances
|
657 |
@rtype: list
|
658 |
@return: the list of instances, sorted
|
659 |
@raise errors.OpPrereqError: if the instances parameter is wrong type
|
660 |
@raise errors.OpPrereqError: if any of the passed instances is not found
|
661 |
|
662 |
"""
|
663 |
if instances:
|
664 |
wanted = [_ExpandInstanceName(lu.cfg, name) for name in instances] |
665 |
else:
|
666 |
wanted = utils.NiceSort(lu.cfg.GetInstanceList()) |
667 |
return wanted
|
668 |
|
669 |
|
670 |
def _GetUpdatedParams(old_params, update_dict, |
671 |
use_default=True, use_none=False): |
672 |
"""Return the new version of a parameter dictionary.
|
673 |
|
674 |
@type old_params: dict
|
675 |
@param old_params: old parameters
|
676 |
@type update_dict: dict
|
677 |
@param update_dict: dict containing new parameter values, or
|
678 |
constants.VALUE_DEFAULT to reset the parameter to its default
|
679 |
value
|
680 |
@param use_default: boolean
|
681 |
@type use_default: whether to recognise L{constants.VALUE_DEFAULT}
|
682 |
values as 'to be deleted' values
|
683 |
@param use_none: boolean
|
684 |
@type use_none: whether to recognise C{None} values as 'to be
|
685 |
deleted' values
|
686 |
@rtype: dict
|
687 |
@return: the new parameter dictionary
|
688 |
|
689 |
"""
|
690 |
params_copy = copy.deepcopy(old_params) |
691 |
for key, val in update_dict.iteritems(): |
692 |
if ((use_default and val == constants.VALUE_DEFAULT) or |
693 |
(use_none and val is None)): |
694 |
try:
|
695 |
del params_copy[key]
|
696 |
except KeyError: |
697 |
pass
|
698 |
else:
|
699 |
params_copy[key] = val |
700 |
return params_copy
|
701 |
|
702 |
|
703 |
def _CheckOutputFields(static, dynamic, selected): |
704 |
"""Checks whether all selected fields are valid.
|
705 |
|
706 |
@type static: L{utils.FieldSet}
|
707 |
@param static: static fields set
|
708 |
@type dynamic: L{utils.FieldSet}
|
709 |
@param dynamic: dynamic fields set
|
710 |
|
711 |
"""
|
712 |
f = utils.FieldSet() |
713 |
f.Extend(static) |
714 |
f.Extend(dynamic) |
715 |
|
716 |
delta = f.NonMatching(selected) |
717 |
if delta:
|
718 |
raise errors.OpPrereqError("Unknown output fields selected: %s" |
719 |
% ",".join(delta), errors.ECODE_INVAL)
|
720 |
|
721 |
|
722 |
def _CheckGlobalHvParams(params): |
723 |
"""Validates that given hypervisor params are not global ones.
|
724 |
|
725 |
This will ensure that instances don't get customised versions of
|
726 |
global params.
|
727 |
|
728 |
"""
|
729 |
used_globals = constants.HVC_GLOBALS.intersection(params) |
730 |
if used_globals:
|
731 |
msg = ("The following hypervisor parameters are global and cannot"
|
732 |
" be customized at instance level, please modify them at"
|
733 |
" cluster level: %s" % utils.CommaJoin(used_globals))
|
734 |
raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
|
735 |
|
736 |
|
737 |
def _CheckNodeOnline(lu, node): |
738 |
"""Ensure that a given node is online.
|
739 |
|
740 |
@param lu: the LU on behalf of which we make the check
|
741 |
@param node: the node to check
|
742 |
@raise errors.OpPrereqError: if the node is offline
|
743 |
|
744 |
"""
|
745 |
if lu.cfg.GetNodeInfo(node).offline:
|
746 |
raise errors.OpPrereqError("Can't use offline node %s" % node, |
747 |
errors.ECODE_INVAL) |
748 |
|
749 |
|
750 |
def _CheckNodeNotDrained(lu, node): |
751 |
"""Ensure that a given node is not drained.
|
752 |
|
753 |
@param lu: the LU on behalf of which we make the check
|
754 |
@param node: the node to check
|
755 |
@raise errors.OpPrereqError: if the node is drained
|
756 |
|
757 |
"""
|
758 |
if lu.cfg.GetNodeInfo(node).drained:
|
759 |
raise errors.OpPrereqError("Can't use drained node %s" % node, |
760 |
errors.ECODE_INVAL) |
761 |
|
762 |
|
763 |
def _CheckNodeHasOS(lu, node, os_name, force_variant): |
764 |
"""Ensure that a node supports a given OS.
|
765 |
|
766 |
@param lu: the LU on behalf of which we make the check
|
767 |
@param node: the node to check
|
768 |
@param os_name: the OS to query about
|
769 |
@param force_variant: whether to ignore variant errors
|
770 |
@raise errors.OpPrereqError: if the node is not supporting the OS
|
771 |
|
772 |
"""
|
773 |
result = lu.rpc.call_os_get(node, os_name) |
774 |
result.Raise("OS '%s' not in supported OS list for node %s" %
|
775 |
(os_name, node), |
776 |
prereq=True, ecode=errors.ECODE_INVAL)
|
777 |
if not force_variant: |
778 |
_CheckOSVariant(result.payload, os_name) |
779 |
|
780 |
|
781 |
def _RequireFileStorage(): |
782 |
"""Checks that file storage is enabled.
|
783 |
|
784 |
@raise errors.OpPrereqError: when file storage is disabled
|
785 |
|
786 |
"""
|
787 |
if not constants.ENABLE_FILE_STORAGE: |
788 |
raise errors.OpPrereqError("File storage disabled at configure time", |
789 |
errors.ECODE_INVAL) |
790 |
|
791 |
|
792 |
def _CheckDiskTemplate(template): |
793 |
"""Ensure a given disk template is valid.
|
794 |
|
795 |
"""
|
796 |
if template not in constants.DISK_TEMPLATES: |
797 |
msg = ("Invalid disk template name '%s', valid templates are: %s" %
|
798 |
(template, utils.CommaJoin(constants.DISK_TEMPLATES))) |
799 |
raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
|
800 |
if template == constants.DT_FILE:
|
801 |
_RequireFileStorage() |
802 |
return True |
803 |
|
804 |
|
805 |
def _CheckStorageType(storage_type): |
806 |
"""Ensure a given storage type is valid.
|
807 |
|
808 |
"""
|
809 |
if storage_type not in constants.VALID_STORAGE_TYPES: |
810 |
raise errors.OpPrereqError("Unknown storage type: %s" % storage_type, |
811 |
errors.ECODE_INVAL) |
812 |
if storage_type == constants.ST_FILE:
|
813 |
_RequireFileStorage() |
814 |
return True |
815 |
|
816 |
|
817 |
def _GetClusterDomainSecret(): |
818 |
"""Reads the cluster domain secret.
|
819 |
|
820 |
"""
|
821 |
return utils.ReadOneLineFile(constants.CLUSTER_DOMAIN_SECRET_FILE,
|
822 |
strict=True)
|
823 |
|
824 |
|
825 |
def _CheckInstanceDown(lu, instance, reason): |
826 |
"""Ensure that an instance is not running."""
|
827 |
if instance.admin_up:
|
828 |
raise errors.OpPrereqError("Instance %s is marked to be up, %s" % |
829 |
(instance.name, reason), errors.ECODE_STATE) |
830 |
|
831 |
pnode = instance.primary_node |
832 |
ins_l = lu.rpc.call_instance_list([pnode], [instance.hypervisor])[pnode] |
833 |
ins_l.Raise("Can't contact node %s for instance information" % pnode,
|
834 |
prereq=True, ecode=errors.ECODE_ENVIRON)
|
835 |
|
836 |
if instance.name in ins_l.payload: |
837 |
raise errors.OpPrereqError("Instance %s is running, %s" % |
838 |
(instance.name, reason), errors.ECODE_STATE) |
839 |
|
840 |
|
841 |
def _ExpandItemName(fn, name, kind): |
842 |
"""Expand an item name.
|
843 |
|
844 |
@param fn: the function to use for expansion
|
845 |
@param name: requested item name
|
846 |
@param kind: text description ('Node' or 'Instance')
|
847 |
@return: the resolved (full) name
|
848 |
@raise errors.OpPrereqError: if the item is not found
|
849 |
|
850 |
"""
|
851 |
full_name = fn(name) |
852 |
if full_name is None: |
853 |
raise errors.OpPrereqError("%s '%s' not known" % (kind, name), |
854 |
errors.ECODE_NOENT) |
855 |
return full_name
|
856 |
|
857 |
|
858 |
def _ExpandNodeName(cfg, name): |
859 |
"""Wrapper over L{_ExpandItemName} for nodes."""
|
860 |
return _ExpandItemName(cfg.ExpandNodeName, name, "Node") |
861 |
|
862 |
|
863 |
def _ExpandInstanceName(cfg, name): |
864 |
"""Wrapper over L{_ExpandItemName} for instance."""
|
865 |
return _ExpandItemName(cfg.ExpandInstanceName, name, "Instance") |
866 |
|
867 |
|
868 |
def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status, |
869 |
memory, vcpus, nics, disk_template, disks, |
870 |
bep, hvp, hypervisor_name): |
871 |
"""Builds instance related env variables for hooks
|
872 |
|
873 |
This builds the hook environment from individual variables.
|
874 |
|
875 |
@type name: string
|
876 |
@param name: the name of the instance
|
877 |
@type primary_node: string
|
878 |
@param primary_node: the name of the instance's primary node
|
879 |
@type secondary_nodes: list
|
880 |
@param secondary_nodes: list of secondary nodes as strings
|
881 |
@type os_type: string
|
882 |
@param os_type: the name of the instance's OS
|
883 |
@type status: boolean
|
884 |
@param status: the should_run status of the instance
|
885 |
@type memory: string
|
886 |
@param memory: the memory size of the instance
|
887 |
@type vcpus: string
|
888 |
@param vcpus: the count of VCPUs the instance has
|
889 |
@type nics: list
|
890 |
@param nics: list of tuples (ip, mac, mode, link) representing
|
891 |
the NICs the instance has
|
892 |
@type disk_template: string
|
893 |
@param disk_template: the disk template of the instance
|
894 |
@type disks: list
|
895 |
@param disks: the list of (size, mode) pairs
|
896 |
@type bep: dict
|
897 |
@param bep: the backend parameters for the instance
|
898 |
@type hvp: dict
|
899 |
@param hvp: the hypervisor parameters for the instance
|
900 |
@type hypervisor_name: string
|
901 |
@param hypervisor_name: the hypervisor for the instance
|
902 |
@rtype: dict
|
903 |
@return: the hook environment for this instance
|
904 |
|
905 |
"""
|
906 |
if status:
|
907 |
str_status = "up"
|
908 |
else:
|
909 |
str_status = "down"
|
910 |
env = { |
911 |
"OP_TARGET": name,
|
912 |
"INSTANCE_NAME": name,
|
913 |
"INSTANCE_PRIMARY": primary_node,
|
914 |
"INSTANCE_SECONDARIES": " ".join(secondary_nodes), |
915 |
"INSTANCE_OS_TYPE": os_type,
|
916 |
"INSTANCE_STATUS": str_status,
|
917 |
"INSTANCE_MEMORY": memory,
|
918 |
"INSTANCE_VCPUS": vcpus,
|
919 |
"INSTANCE_DISK_TEMPLATE": disk_template,
|
920 |
"INSTANCE_HYPERVISOR": hypervisor_name,
|
921 |
} |
922 |
|
923 |
if nics:
|
924 |
nic_count = len(nics)
|
925 |
for idx, (ip, mac, mode, link) in enumerate(nics): |
926 |
if ip is None: |
927 |
ip = ""
|
928 |
env["INSTANCE_NIC%d_IP" % idx] = ip
|
929 |
env["INSTANCE_NIC%d_MAC" % idx] = mac
|
930 |
env["INSTANCE_NIC%d_MODE" % idx] = mode
|
931 |
env["INSTANCE_NIC%d_LINK" % idx] = link
|
932 |
if mode == constants.NIC_MODE_BRIDGED:
|
933 |
env["INSTANCE_NIC%d_BRIDGE" % idx] = link
|
934 |
else:
|
935 |
nic_count = 0
|
936 |
|
937 |
env["INSTANCE_NIC_COUNT"] = nic_count
|
938 |
|
939 |
if disks:
|
940 |
disk_count = len(disks)
|
941 |
for idx, (size, mode) in enumerate(disks): |
942 |
env["INSTANCE_DISK%d_SIZE" % idx] = size
|
943 |
env["INSTANCE_DISK%d_MODE" % idx] = mode
|
944 |
else:
|
945 |
disk_count = 0
|
946 |
|
947 |
env["INSTANCE_DISK_COUNT"] = disk_count
|
948 |
|
949 |
for source, kind in [(bep, "BE"), (hvp, "HV")]: |
950 |
for key, value in source.items(): |
951 |
env["INSTANCE_%s_%s" % (kind, key)] = value
|
952 |
|
953 |
return env
|
954 |
|
955 |
|
956 |
def _NICListToTuple(lu, nics): |
957 |
"""Build a list of nic information tuples.
|
958 |
|
959 |
This list is suitable to be passed to _BuildInstanceHookEnv or as a return
|
960 |
value in LUQueryInstanceData.
|
961 |
|
962 |
@type lu: L{LogicalUnit}
|
963 |
@param lu: the logical unit on whose behalf we execute
|
964 |
@type nics: list of L{objects.NIC}
|
965 |
@param nics: list of nics to convert to hooks tuples
|
966 |
|
967 |
"""
|
968 |
hooks_nics = [] |
969 |
cluster = lu.cfg.GetClusterInfo() |
970 |
for nic in nics: |
971 |
ip = nic.ip |
972 |
mac = nic.mac |
973 |
filled_params = cluster.SimpleFillNIC(nic.nicparams) |
974 |
mode = filled_params[constants.NIC_MODE] |
975 |
link = filled_params[constants.NIC_LINK] |
976 |
hooks_nics.append((ip, mac, mode, link)) |
977 |
return hooks_nics
|
978 |
|
979 |
|
980 |
def _BuildInstanceHookEnvByObject(lu, instance, override=None): |
981 |
"""Builds instance related env variables for hooks from an object.
|
982 |
|
983 |
@type lu: L{LogicalUnit}
|
984 |
@param lu: the logical unit on whose behalf we execute
|
985 |
@type instance: L{objects.Instance}
|
986 |
@param instance: the instance for which we should build the
|
987 |
environment
|
988 |
@type override: dict
|
989 |
@param override: dictionary with key/values that will override
|
990 |
our values
|
991 |
@rtype: dict
|
992 |
@return: the hook environment dictionary
|
993 |
|
994 |
"""
|
995 |
cluster = lu.cfg.GetClusterInfo() |
996 |
bep = cluster.FillBE(instance) |
997 |
hvp = cluster.FillHV(instance) |
998 |
args = { |
999 |
'name': instance.name,
|
1000 |
'primary_node': instance.primary_node,
|
1001 |
'secondary_nodes': instance.secondary_nodes,
|
1002 |
'os_type': instance.os,
|
1003 |
'status': instance.admin_up,
|
1004 |
'memory': bep[constants.BE_MEMORY],
|
1005 |
'vcpus': bep[constants.BE_VCPUS],
|
1006 |
'nics': _NICListToTuple(lu, instance.nics),
|
1007 |
'disk_template': instance.disk_template,
|
1008 |
'disks': [(disk.size, disk.mode) for disk in instance.disks], |
1009 |
'bep': bep,
|
1010 |
'hvp': hvp,
|
1011 |
'hypervisor_name': instance.hypervisor,
|
1012 |
} |
1013 |
if override:
|
1014 |
args.update(override) |
1015 |
return _BuildInstanceHookEnv(**args) # pylint: disable-msg=W0142 |
1016 |
|
1017 |
|
1018 |
def _AdjustCandidatePool(lu, exceptions): |
1019 |
"""Adjust the candidate pool after node operations.
|
1020 |
|
1021 |
"""
|
1022 |
mod_list = lu.cfg.MaintainCandidatePool(exceptions) |
1023 |
if mod_list:
|
1024 |
lu.LogInfo("Promoted nodes to master candidate role: %s",
|
1025 |
utils.CommaJoin(node.name for node in mod_list)) |
1026 |
for name in mod_list: |
1027 |
lu.context.ReaddNode(name) |
1028 |
mc_now, mc_max, _ = lu.cfg.GetMasterCandidateStats(exceptions) |
1029 |
if mc_now > mc_max:
|
1030 |
lu.LogInfo("Note: more nodes are candidates (%d) than desired (%d)" %
|
1031 |
(mc_now, mc_max)) |
1032 |
|
1033 |
|
1034 |
def _DecideSelfPromotion(lu, exceptions=None): |
1035 |
"""Decide whether I should promote myself as a master candidate.
|
1036 |
|
1037 |
"""
|
1038 |
cp_size = lu.cfg.GetClusterInfo().candidate_pool_size |
1039 |
mc_now, mc_should, _ = lu.cfg.GetMasterCandidateStats(exceptions) |
1040 |
# the new node will increase mc_max with one, so:
|
1041 |
mc_should = min(mc_should + 1, cp_size) |
1042 |
return mc_now < mc_should
|
1043 |
|
1044 |
|
1045 |
def _CheckNicsBridgesExist(lu, target_nics, target_node): |
1046 |
"""Check that the brigdes needed by a list of nics exist.
|
1047 |
|
1048 |
"""
|
1049 |
cluster = lu.cfg.GetClusterInfo() |
1050 |
paramslist = [cluster.SimpleFillNIC(nic.nicparams) for nic in target_nics] |
1051 |
brlist = [params[constants.NIC_LINK] for params in paramslist |
1052 |
if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED]
|
1053 |
if brlist:
|
1054 |
result = lu.rpc.call_bridges_exist(target_node, brlist) |
1055 |
result.Raise("Error checking bridges on destination node '%s'" %
|
1056 |
target_node, prereq=True, ecode=errors.ECODE_ENVIRON)
|
1057 |
|
1058 |
|
1059 |
def _CheckInstanceBridgesExist(lu, instance, node=None): |
1060 |
"""Check that the brigdes needed by an instance exist.
|
1061 |
|
1062 |
"""
|
1063 |
if node is None: |
1064 |
node = instance.primary_node |
1065 |
_CheckNicsBridgesExist(lu, instance.nics, node) |
1066 |
|
1067 |
|
1068 |
def _CheckOSVariant(os_obj, name): |
1069 |
"""Check whether an OS name conforms to the os variants specification.
|
1070 |
|
1071 |
@type os_obj: L{objects.OS}
|
1072 |
@param os_obj: OS object to check
|
1073 |
@type name: string
|
1074 |
@param name: OS name passed by the user, to check for validity
|
1075 |
|
1076 |
"""
|
1077 |
if not os_obj.supported_variants: |
1078 |
return
|
1079 |
try:
|
1080 |
variant = name.split("+", 1)[1] |
1081 |
except IndexError: |
1082 |
raise errors.OpPrereqError("OS name must include a variant", |
1083 |
errors.ECODE_INVAL) |
1084 |
|
1085 |
if variant not in os_obj.supported_variants: |
1086 |
raise errors.OpPrereqError("Unsupported OS variant", errors.ECODE_INVAL) |
1087 |
|
1088 |
|
1089 |
def _GetNodeInstancesInner(cfg, fn): |
1090 |
return [i for i in cfg.GetAllInstancesInfo().values() if fn(i)] |
1091 |
|
1092 |
|
1093 |
def _GetNodeInstances(cfg, node_name): |
1094 |
"""Returns a list of all primary and secondary instances on a node.
|
1095 |
|
1096 |
"""
|
1097 |
|
1098 |
return _GetNodeInstancesInner(cfg, lambda inst: node_name in inst.all_nodes) |
1099 |
|
1100 |
|
1101 |
def _GetNodePrimaryInstances(cfg, node_name): |
1102 |
"""Returns primary instances on a node.
|
1103 |
|
1104 |
"""
|
1105 |
return _GetNodeInstancesInner(cfg,
|
1106 |
lambda inst: node_name == inst.primary_node)
|
1107 |
|
1108 |
|
1109 |
def _GetNodeSecondaryInstances(cfg, node_name): |
1110 |
"""Returns secondary instances on a node.
|
1111 |
|
1112 |
"""
|
1113 |
return _GetNodeInstancesInner(cfg,
|
1114 |
lambda inst: node_name in inst.secondary_nodes) |
1115 |
|
1116 |
|
1117 |
def _GetStorageTypeArgs(cfg, storage_type): |
1118 |
"""Returns the arguments for a storage type.
|
1119 |
|
1120 |
"""
|
1121 |
# Special case for file storage
|
1122 |
if storage_type == constants.ST_FILE:
|
1123 |
# storage.FileStorage wants a list of storage directories
|
1124 |
return [[cfg.GetFileStorageDir()]]
|
1125 |
|
1126 |
return []
|
1127 |
|
1128 |
|
1129 |
def _FindFaultyInstanceDisks(cfg, rpc, instance, node_name, prereq): |
1130 |
faulty = [] |
1131 |
|
1132 |
for dev in instance.disks: |
1133 |
cfg.SetDiskID(dev, node_name) |
1134 |
|
1135 |
result = rpc.call_blockdev_getmirrorstatus(node_name, instance.disks) |
1136 |
result.Raise("Failed to get disk status from node %s" % node_name,
|
1137 |
prereq=prereq, ecode=errors.ECODE_ENVIRON) |
1138 |
|
1139 |
for idx, bdev_status in enumerate(result.payload): |
1140 |
if bdev_status and bdev_status.ldisk_status == constants.LDS_FAULTY: |
1141 |
faulty.append(idx) |
1142 |
|
1143 |
return faulty
|
1144 |
|
1145 |
|
1146 |
class LUPostInitCluster(LogicalUnit): |
1147 |
"""Logical unit for running hooks after cluster initialization.
|
1148 |
|
1149 |
"""
|
1150 |
HPATH = "cluster-init"
|
1151 |
HTYPE = constants.HTYPE_CLUSTER |
1152 |
|
1153 |
def BuildHooksEnv(self): |
1154 |
"""Build hooks env.
|
1155 |
|
1156 |
"""
|
1157 |
env = {"OP_TARGET": self.cfg.GetClusterName()} |
1158 |
mn = self.cfg.GetMasterNode()
|
1159 |
return env, [], [mn]
|
1160 |
|
1161 |
def Exec(self, feedback_fn): |
1162 |
"""Nothing to do.
|
1163 |
|
1164 |
"""
|
1165 |
return True |
1166 |
|
1167 |
|
1168 |
class LUDestroyCluster(LogicalUnit): |
1169 |
"""Logical unit for destroying the cluster.
|
1170 |
|
1171 |
"""
|
1172 |
HPATH = "cluster-destroy"
|
1173 |
HTYPE = constants.HTYPE_CLUSTER |
1174 |
|
1175 |
def BuildHooksEnv(self): |
1176 |
"""Build hooks env.
|
1177 |
|
1178 |
"""
|
1179 |
env = {"OP_TARGET": self.cfg.GetClusterName()} |
1180 |
return env, [], []
|
1181 |
|
1182 |
def CheckPrereq(self): |
1183 |
"""Check prerequisites.
|
1184 |
|
1185 |
This checks whether the cluster is empty.
|
1186 |
|
1187 |
Any errors are signaled by raising errors.OpPrereqError.
|
1188 |
|
1189 |
"""
|
1190 |
master = self.cfg.GetMasterNode()
|
1191 |
|
1192 |
nodelist = self.cfg.GetNodeList()
|
1193 |
if len(nodelist) != 1 or nodelist[0] != master: |
1194 |
raise errors.OpPrereqError("There are still %d node(s) in" |
1195 |
" this cluster." % (len(nodelist) - 1), |
1196 |
errors.ECODE_INVAL) |
1197 |
instancelist = self.cfg.GetInstanceList()
|
1198 |
if instancelist:
|
1199 |
raise errors.OpPrereqError("There are still %d instance(s) in" |
1200 |
" this cluster." % len(instancelist), |
1201 |
errors.ECODE_INVAL) |
1202 |
|
1203 |
def Exec(self, feedback_fn): |
1204 |
"""Destroys the cluster.
|
1205 |
|
1206 |
"""
|
1207 |
master = self.cfg.GetMasterNode()
|
1208 |
modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
|
1209 |
|
1210 |
# Run post hooks on master node before it's removed
|
1211 |
hm = self.proc.hmclass(self.rpc.call_hooks_runner, self) |
1212 |
try:
|
1213 |
hm.RunPhase(constants.HOOKS_PHASE_POST, [master]) |
1214 |
except:
|
1215 |
# pylint: disable-msg=W0702
|
1216 |
self.LogWarning("Errors occurred running hooks on %s" % master) |
1217 |
|
1218 |
result = self.rpc.call_node_stop_master(master, False) |
1219 |
result.Raise("Could not disable the master role")
|
1220 |
|
1221 |
if modify_ssh_setup:
|
1222 |
priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS) |
1223 |
utils.CreateBackup(priv_key) |
1224 |
utils.CreateBackup(pub_key) |
1225 |
|
1226 |
return master
|
1227 |
|
1228 |
|
1229 |
def _VerifyCertificate(filename): |
1230 |
"""Verifies a certificate for LUVerifyCluster.
|
1231 |
|
1232 |
@type filename: string
|
1233 |
@param filename: Path to PEM file
|
1234 |
|
1235 |
"""
|
1236 |
try:
|
1237 |
cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM, |
1238 |
utils.ReadFile(filename)) |
1239 |
except Exception, err: # pylint: disable-msg=W0703 |
1240 |
return (LUVerifyCluster.ETYPE_ERROR,
|
1241 |
"Failed to load X509 certificate %s: %s" % (filename, err))
|
1242 |
|
1243 |
(errcode, msg) = \ |
1244 |
utils.VerifyX509Certificate(cert, constants.SSL_CERT_EXPIRATION_WARN, |
1245 |
constants.SSL_CERT_EXPIRATION_ERROR) |
1246 |
|
1247 |
if msg:
|
1248 |
fnamemsg = "While verifying %s: %s" % (filename, msg)
|
1249 |
else:
|
1250 |
fnamemsg = None
|
1251 |
|
1252 |
if errcode is None: |
1253 |
return (None, fnamemsg) |
1254 |
elif errcode == utils.CERT_WARNING:
|
1255 |
return (LUVerifyCluster.ETYPE_WARNING, fnamemsg)
|
1256 |
elif errcode == utils.CERT_ERROR:
|
1257 |
return (LUVerifyCluster.ETYPE_ERROR, fnamemsg)
|
1258 |
|
1259 |
raise errors.ProgrammerError("Unhandled certificate error code %r" % errcode) |
1260 |
|
1261 |
|
1262 |
class LUVerifyCluster(LogicalUnit): |
1263 |
"""Verifies the cluster status.
|
1264 |
|
1265 |
"""
|
1266 |
HPATH = "cluster-verify"
|
1267 |
HTYPE = constants.HTYPE_CLUSTER |
1268 |
_OP_PARAMS = [ |
1269 |
("skip_checks", _EmptyList,
|
1270 |
_TListOf(_TElemOf(constants.VERIFY_OPTIONAL_CHECKS))), |
1271 |
("verbose", False, _TBool), |
1272 |
("error_codes", False, _TBool), |
1273 |
("debug_simulate_errors", False, _TBool), |
1274 |
] |
1275 |
REQ_BGL = False
|
1276 |
|
1277 |
TCLUSTER = "cluster"
|
1278 |
TNODE = "node"
|
1279 |
TINSTANCE = "instance"
|
1280 |
|
1281 |
ECLUSTERCFG = (TCLUSTER, "ECLUSTERCFG")
|
1282 |
ECLUSTERCERT = (TCLUSTER, "ECLUSTERCERT")
|
1283 |
EINSTANCEBADNODE = (TINSTANCE, "EINSTANCEBADNODE")
|
1284 |
EINSTANCEDOWN = (TINSTANCE, "EINSTANCEDOWN")
|
1285 |
EINSTANCELAYOUT = (TINSTANCE, "EINSTANCELAYOUT")
|
1286 |
EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
|
1287 |
EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
|
1288 |
EINSTANCEWRONGNODE = (TINSTANCE, "EINSTANCEWRONGNODE")
|
1289 |
ENODEDRBD = (TNODE, "ENODEDRBD")
|
1290 |
ENODEDRBDHELPER = (TNODE, "ENODEDRBDHELPER")
|
1291 |
ENODEFILECHECK = (TNODE, "ENODEFILECHECK")
|
1292 |
ENODEHOOKS = (TNODE, "ENODEHOOKS")
|
1293 |
ENODEHV = (TNODE, "ENODEHV")
|
1294 |
ENODELVM = (TNODE, "ENODELVM")
|
1295 |
ENODEN1 = (TNODE, "ENODEN1")
|
1296 |
ENODENET = (TNODE, "ENODENET")
|
1297 |
ENODEOS = (TNODE, "ENODEOS")
|
1298 |
ENODEORPHANINSTANCE = (TNODE, "ENODEORPHANINSTANCE")
|
1299 |
ENODEORPHANLV = (TNODE, "ENODEORPHANLV")
|
1300 |
ENODERPC = (TNODE, "ENODERPC")
|
1301 |
ENODESSH = (TNODE, "ENODESSH")
|
1302 |
ENODEVERSION = (TNODE, "ENODEVERSION")
|
1303 |
ENODESETUP = (TNODE, "ENODESETUP")
|
1304 |
ENODETIME = (TNODE, "ENODETIME")
|
1305 |
|
1306 |
ETYPE_FIELD = "code"
|
1307 |
ETYPE_ERROR = "ERROR"
|
1308 |
ETYPE_WARNING = "WARNING"
|
1309 |
|
1310 |
class NodeImage(object): |
1311 |
"""A class representing the logical and physical status of a node.
|
1312 |
|
1313 |
@type name: string
|
1314 |
@ivar name: the node name to which this object refers
|
1315 |
@ivar volumes: a structure as returned from
|
1316 |
L{ganeti.backend.GetVolumeList} (runtime)
|
1317 |
@ivar instances: a list of running instances (runtime)
|
1318 |
@ivar pinst: list of configured primary instances (config)
|
1319 |
@ivar sinst: list of configured secondary instances (config)
|
1320 |
@ivar sbp: diction of {secondary-node: list of instances} of all peers
|
1321 |
of this node (config)
|
1322 |
@ivar mfree: free memory, as reported by hypervisor (runtime)
|
1323 |
@ivar dfree: free disk, as reported by the node (runtime)
|
1324 |
@ivar offline: the offline status (config)
|
1325 |
@type rpc_fail: boolean
|
1326 |
@ivar rpc_fail: whether the RPC verify call was successfull (overall,
|
1327 |
not whether the individual keys were correct) (runtime)
|
1328 |
@type lvm_fail: boolean
|
1329 |
@ivar lvm_fail: whether the RPC call didn't return valid LVM data
|
1330 |
@type hyp_fail: boolean
|
1331 |
@ivar hyp_fail: whether the RPC call didn't return the instance list
|
1332 |
@type ghost: boolean
|
1333 |
@ivar ghost: whether this is a known node or not (config)
|
1334 |
@type os_fail: boolean
|
1335 |
@ivar os_fail: whether the RPC call didn't return valid OS data
|
1336 |
@type oslist: list
|
1337 |
@ivar oslist: list of OSes as diagnosed by DiagnoseOS
|
1338 |
|
1339 |
"""
|
1340 |
def __init__(self, offline=False, name=None): |
1341 |
self.name = name
|
1342 |
self.volumes = {}
|
1343 |
self.instances = []
|
1344 |
self.pinst = []
|
1345 |
self.sinst = []
|
1346 |
self.sbp = {}
|
1347 |
self.mfree = 0 |
1348 |
self.dfree = 0 |
1349 |
self.offline = offline
|
1350 |
self.rpc_fail = False |
1351 |
self.lvm_fail = False |
1352 |
self.hyp_fail = False |
1353 |
self.ghost = False |
1354 |
self.os_fail = False |
1355 |
self.oslist = {}
|
1356 |
|
1357 |
def ExpandNames(self): |
1358 |
self.needed_locks = {
|
1359 |
locking.LEVEL_NODE: locking.ALL_SET, |
1360 |
locking.LEVEL_INSTANCE: locking.ALL_SET, |
1361 |
} |
1362 |
self.share_locks = dict.fromkeys(locking.LEVELS, 1) |
1363 |
|
1364 |
def _Error(self, ecode, item, msg, *args, **kwargs): |
1365 |
"""Format an error message.
|
1366 |
|
1367 |
Based on the opcode's error_codes parameter, either format a
|
1368 |
parseable error code, or a simpler error string.
|
1369 |
|
1370 |
This must be called only from Exec and functions called from Exec.
|
1371 |
|
1372 |
"""
|
1373 |
ltype = kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR) |
1374 |
itype, etxt = ecode |
1375 |
# first complete the msg
|
1376 |
if args:
|
1377 |
msg = msg % args |
1378 |
# then format the whole message
|
1379 |
if self.op.error_codes: |
1380 |
msg = "%s:%s:%s:%s:%s" % (ltype, etxt, itype, item, msg)
|
1381 |
else:
|
1382 |
if item:
|
1383 |
item = " " + item
|
1384 |
else:
|
1385 |
item = ""
|
1386 |
msg = "%s: %s%s: %s" % (ltype, itype, item, msg)
|
1387 |
# and finally report it via the feedback_fn
|
1388 |
self._feedback_fn(" - %s" % msg) |
1389 |
|
1390 |
def _ErrorIf(self, cond, *args, **kwargs): |
1391 |
"""Log an error message if the passed condition is True.
|
1392 |
|
1393 |
"""
|
1394 |
cond = bool(cond) or self.op.debug_simulate_errors |
1395 |
if cond:
|
1396 |
self._Error(*args, **kwargs)
|
1397 |
# do not mark the operation as failed for WARN cases only
|
1398 |
if kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR) == self.ETYPE_ERROR: |
1399 |
self.bad = self.bad or cond |
1400 |
|
1401 |
def _VerifyNode(self, ninfo, nresult): |
1402 |
"""Run multiple tests against a node.
|
1403 |
|
1404 |
Test list:
|
1405 |
|
1406 |
- compares ganeti version
|
1407 |
- checks vg existence and size > 20G
|
1408 |
- checks config file checksum
|
1409 |
- checks ssh to other nodes
|
1410 |
|
1411 |
@type ninfo: L{objects.Node}
|
1412 |
@param ninfo: the node to check
|
1413 |
@param nresult: the results from the node
|
1414 |
@rtype: boolean
|
1415 |
@return: whether overall this call was successful (and we can expect
|
1416 |
reasonable values in the respose)
|
1417 |
|
1418 |
"""
|
1419 |
node = ninfo.name |
1420 |
_ErrorIf = self._ErrorIf # pylint: disable-msg=C0103 |
1421 |
|
1422 |
# main result, nresult should be a non-empty dict
|
1423 |
test = not nresult or not isinstance(nresult, dict) |
1424 |
_ErrorIf(test, self.ENODERPC, node,
|
1425 |
"unable to verify node: no data returned")
|
1426 |
if test:
|
1427 |
return False |
1428 |
|
1429 |
# compares ganeti version
|
1430 |
local_version = constants.PROTOCOL_VERSION |
1431 |
remote_version = nresult.get("version", None) |
1432 |
test = not (remote_version and |
1433 |
isinstance(remote_version, (list, tuple)) and |
1434 |
len(remote_version) == 2) |
1435 |
_ErrorIf(test, self.ENODERPC, node,
|
1436 |
"connection to node returned invalid data")
|
1437 |
if test:
|
1438 |
return False |
1439 |
|
1440 |
test = local_version != remote_version[0]
|
1441 |
_ErrorIf(test, self.ENODEVERSION, node,
|
1442 |
"incompatible protocol versions: master %s,"
|
1443 |
" node %s", local_version, remote_version[0]) |
1444 |
if test:
|
1445 |
return False |
1446 |
|
1447 |
# node seems compatible, we can actually try to look into its results
|
1448 |
|
1449 |
# full package version
|
1450 |
self._ErrorIf(constants.RELEASE_VERSION != remote_version[1], |
1451 |
self.ENODEVERSION, node,
|
1452 |
"software version mismatch: master %s, node %s",
|
1453 |
constants.RELEASE_VERSION, remote_version[1],
|
1454 |
code=self.ETYPE_WARNING)
|
1455 |
|
1456 |
hyp_result = nresult.get(constants.NV_HYPERVISOR, None)
|
1457 |
if isinstance(hyp_result, dict): |
1458 |
for hv_name, hv_result in hyp_result.iteritems(): |
1459 |
test = hv_result is not None |
1460 |
_ErrorIf(test, self.ENODEHV, node,
|
1461 |
"hypervisor %s verify failure: '%s'", hv_name, hv_result)
|
1462 |
|
1463 |
|
1464 |
test = nresult.get(constants.NV_NODESETUP, |
1465 |
["Missing NODESETUP results"])
|
1466 |
_ErrorIf(test, self.ENODESETUP, node, "node setup error: %s", |
1467 |
"; ".join(test))
|
1468 |
|
1469 |
return True |
1470 |
|
1471 |
def _VerifyNodeTime(self, ninfo, nresult, |
1472 |
nvinfo_starttime, nvinfo_endtime): |
1473 |
"""Check the node time.
|
1474 |
|
1475 |
@type ninfo: L{objects.Node}
|
1476 |
@param ninfo: the node to check
|
1477 |
@param nresult: the remote results for the node
|
1478 |
@param nvinfo_starttime: the start time of the RPC call
|
1479 |
@param nvinfo_endtime: the end time of the RPC call
|
1480 |
|
1481 |
"""
|
1482 |
node = ninfo.name |
1483 |
_ErrorIf = self._ErrorIf # pylint: disable-msg=C0103 |
1484 |
|
1485 |
ntime = nresult.get(constants.NV_TIME, None)
|
1486 |
try:
|
1487 |
ntime_merged = utils.MergeTime(ntime) |
1488 |
except (ValueError, TypeError): |
1489 |
_ErrorIf(True, self.ENODETIME, node, "Node returned invalid time") |
1490 |
return
|
1491 |
|
1492 |
if ntime_merged < (nvinfo_starttime - constants.NODE_MAX_CLOCK_SKEW):
|
1493 |
ntime_diff = "%.01fs" % abs(nvinfo_starttime - ntime_merged) |
1494 |
elif ntime_merged > (nvinfo_endtime + constants.NODE_MAX_CLOCK_SKEW):
|
1495 |
ntime_diff = "%.01fs" % abs(ntime_merged - nvinfo_endtime) |
1496 |
else:
|
1497 |
ntime_diff = None
|
1498 |
|
1499 |
_ErrorIf(ntime_diff is not None, self.ENODETIME, node, |
1500 |
"Node time diverges by at least %s from master node time",
|
1501 |
ntime_diff) |
1502 |
|
1503 |
def _VerifyNodeLVM(self, ninfo, nresult, vg_name): |
1504 |
"""Check the node time.
|
1505 |
|
1506 |
@type ninfo: L{objects.Node}
|
1507 |
@param ninfo: the node to check
|
1508 |
@param nresult: the remote results for the node
|
1509 |
@param vg_name: the configured VG name
|
1510 |
|
1511 |
"""
|
1512 |
if vg_name is None: |
1513 |
return
|
1514 |
|
1515 |
node = ninfo.name |
1516 |
_ErrorIf = self._ErrorIf # pylint: disable-msg=C0103 |
1517 |
|
1518 |
# checks vg existence and size > 20G
|
1519 |
vglist = nresult.get(constants.NV_VGLIST, None)
|
1520 |
test = not vglist
|
1521 |
_ErrorIf(test, self.ENODELVM, node, "unable to check volume groups") |
1522 |
if not test: |
1523 |
vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name, |
1524 |
constants.MIN_VG_SIZE) |
1525 |
_ErrorIf(vgstatus, self.ENODELVM, node, vgstatus)
|
1526 |
|
1527 |
# check pv names
|
1528 |
pvlist = nresult.get(constants.NV_PVLIST, None)
|
1529 |
test = pvlist is None |
1530 |
_ErrorIf(test, self.ENODELVM, node, "Can't get PV list from node") |
1531 |
if not test: |
1532 |
# check that ':' is not present in PV names, since it's a
|
1533 |
# special character for lvcreate (denotes the range of PEs to
|
1534 |
# use on the PV)
|
1535 |
for _, pvname, owner_vg in pvlist: |
1536 |
test = ":" in pvname |
1537 |
_ErrorIf(test, self.ENODELVM, node, "Invalid character ':' in PV" |
1538 |
" '%s' of VG '%s'", pvname, owner_vg)
|
1539 |
|
1540 |
def _VerifyNodeNetwork(self, ninfo, nresult): |
1541 |
"""Check the node time.
|
1542 |
|
1543 |
@type ninfo: L{objects.Node}
|
1544 |
@param ninfo: the node to check
|
1545 |
@param nresult: the remote results for the node
|
1546 |
|
1547 |
"""
|
1548 |
node = ninfo.name |
1549 |
_ErrorIf = self._ErrorIf # pylint: disable-msg=C0103 |
1550 |
|
1551 |
test = constants.NV_NODELIST not in nresult |
1552 |
_ErrorIf(test, self.ENODESSH, node,
|
1553 |
"node hasn't returned node ssh connectivity data")
|
1554 |
if not test: |
1555 |
if nresult[constants.NV_NODELIST]:
|
1556 |
for a_node, a_msg in nresult[constants.NV_NODELIST].items(): |
1557 |
_ErrorIf(True, self.ENODESSH, node, |
1558 |
"ssh communication with node '%s': %s", a_node, a_msg)
|
1559 |
|
1560 |
test = constants.NV_NODENETTEST not in nresult |
1561 |
_ErrorIf(test, self.ENODENET, node,
|
1562 |
"node hasn't returned node tcp connectivity data")
|
1563 |
if not test: |
1564 |
if nresult[constants.NV_NODENETTEST]:
|
1565 |
nlist = utils.NiceSort(nresult[constants.NV_NODENETTEST].keys()) |
1566 |
for anode in nlist: |
1567 |
_ErrorIf(True, self.ENODENET, node, |
1568 |
"tcp communication with node '%s': %s",
|
1569 |
anode, nresult[constants.NV_NODENETTEST][anode]) |
1570 |
|
1571 |
test = constants.NV_MASTERIP not in nresult |
1572 |
_ErrorIf(test, self.ENODENET, node,
|
1573 |
"node hasn't returned node master IP reachability data")
|
1574 |
if not test: |
1575 |
if not nresult[constants.NV_MASTERIP]: |
1576 |
if node == self.master_node: |
1577 |
msg = "the master node cannot reach the master IP (not configured?)"
|
1578 |
else:
|
1579 |
msg = "cannot reach the master IP"
|
1580 |
_ErrorIf(True, self.ENODENET, node, msg) |
1581 |
|
1582 |
|
1583 |
def _VerifyInstance(self, instance, instanceconfig, node_image): |
1584 |
"""Verify an instance.
|
1585 |
|
1586 |
This function checks to see if the required block devices are
|
1587 |
available on the instance's node.
|
1588 |
|
1589 |
"""
|
1590 |
_ErrorIf = self._ErrorIf # pylint: disable-msg=C0103 |
1591 |
node_current = instanceconfig.primary_node |
1592 |
|
1593 |
node_vol_should = {} |
1594 |
instanceconfig.MapLVsByNode(node_vol_should) |
1595 |
|
1596 |
for node in node_vol_should: |
1597 |
n_img = node_image[node] |
1598 |
if n_img.offline or n_img.rpc_fail or n_img.lvm_fail: |
1599 |
# ignore missing volumes on offline or broken nodes
|
1600 |
continue
|
1601 |
for volume in node_vol_should[node]: |
1602 |
test = volume not in n_img.volumes |
1603 |
_ErrorIf(test, self.EINSTANCEMISSINGDISK, instance,
|
1604 |
"volume %s missing on node %s", volume, node)
|
1605 |
|
1606 |
if instanceconfig.admin_up:
|
1607 |
pri_img = node_image[node_current] |
1608 |
test = instance not in pri_img.instances and not pri_img.offline |
1609 |
_ErrorIf(test, self.EINSTANCEDOWN, instance,
|
1610 |
"instance not running on its primary node %s",
|
1611 |
node_current) |
1612 |
|
1613 |
for node, n_img in node_image.items(): |
1614 |
if (not node == node_current): |
1615 |
test = instance in n_img.instances
|
1616 |
_ErrorIf(test, self.EINSTANCEWRONGNODE, instance,
|
1617 |
"instance should not run on node %s", node)
|
1618 |
|
1619 |
def _VerifyOrphanVolumes(self, node_vol_should, node_image): |
1620 |
"""Verify if there are any unknown volumes in the cluster.
|
1621 |
|
1622 |
The .os, .swap and backup volumes are ignored. All other volumes are
|
1623 |
reported as unknown.
|
1624 |
|
1625 |
"""
|
1626 |
for node, n_img in node_image.items(): |
1627 |
if n_img.offline or n_img.rpc_fail or n_img.lvm_fail: |
1628 |
# skip non-healthy nodes
|
1629 |
continue
|
1630 |
for volume in n_img.volumes: |
1631 |
test = (node not in node_vol_should or |
1632 |
volume not in node_vol_should[node]) |
1633 |
self._ErrorIf(test, self.ENODEORPHANLV, node, |
1634 |
"volume %s is unknown", volume)
|
1635 |
|
1636 |
def _VerifyOrphanInstances(self, instancelist, node_image): |
1637 |
"""Verify the list of running instances.
|
1638 |
|
1639 |
This checks what instances are running but unknown to the cluster.
|
1640 |
|
1641 |
"""
|
1642 |
for node, n_img in node_image.items(): |
1643 |
for o_inst in n_img.instances: |
1644 |
test = o_inst not in instancelist |
1645 |
self._ErrorIf(test, self.ENODEORPHANINSTANCE, node, |
1646 |
"instance %s on node %s should not exist", o_inst, node)
|
1647 |
|
1648 |
def _VerifyNPlusOneMemory(self, node_image, instance_cfg): |
1649 |
"""Verify N+1 Memory Resilience.
|
1650 |
|
1651 |
Check that if one single node dies we can still start all the
|
1652 |
instances it was primary for.
|
1653 |
|
1654 |
"""
|
1655 |
for node, n_img in node_image.items(): |
1656 |
# This code checks that every node which is now listed as
|
1657 |
# secondary has enough memory to host all instances it is
|
1658 |
# supposed to should a single other node in the cluster fail.
|
1659 |
# FIXME: not ready for failover to an arbitrary node
|
1660 |
# FIXME: does not support file-backed instances
|
1661 |
# WARNING: we currently take into account down instances as well
|
1662 |
# as up ones, considering that even if they're down someone
|
1663 |
# might want to start them even in the event of a node failure.
|
1664 |
for prinode, instances in n_img.sbp.items(): |
1665 |
needed_mem = 0
|
1666 |
for instance in instances: |
1667 |
bep = self.cfg.GetClusterInfo().FillBE(instance_cfg[instance])
|
1668 |
if bep[constants.BE_AUTO_BALANCE]:
|
1669 |
needed_mem += bep[constants.BE_MEMORY] |
1670 |
test = n_img.mfree < needed_mem |
1671 |
self._ErrorIf(test, self.ENODEN1, node, |
1672 |
"not enough memory on to accommodate"
|
1673 |
" failovers should peer node %s fail", prinode)
|
1674 |
|
1675 |
def _VerifyNodeFiles(self, ninfo, nresult, file_list, local_cksum, |
1676 |
master_files): |
1677 |
"""Verifies and computes the node required file checksums.
|
1678 |
|
1679 |
@type ninfo: L{objects.Node}
|
1680 |
@param ninfo: the node to check
|
1681 |
@param nresult: the remote results for the node
|
1682 |
@param file_list: required list of files
|
1683 |
@param local_cksum: dictionary of local files and their checksums
|
1684 |
@param master_files: list of files that only masters should have
|
1685 |
|
1686 |
"""
|
1687 |
node = ninfo.name |
1688 |
_ErrorIf = self._ErrorIf # pylint: disable-msg=C0103 |
1689 |
|
1690 |
remote_cksum = nresult.get(constants.NV_FILELIST, None)
|
1691 |
test = not isinstance(remote_cksum, dict) |
1692 |
_ErrorIf(test, self.ENODEFILECHECK, node,
|
1693 |
"node hasn't returned file checksum data")
|
1694 |
if test:
|
1695 |
return
|
1696 |
|
1697 |
for file_name in file_list: |
1698 |
node_is_mc = ninfo.master_candidate |
1699 |
must_have = (file_name not in master_files) or node_is_mc |
1700 |
# missing
|
1701 |
test1 = file_name not in remote_cksum |
1702 |
# invalid checksum
|
1703 |
test2 = not test1 and remote_cksum[file_name] != local_cksum[file_name] |
1704 |
# existing and good
|
1705 |
test3 = not test1 and remote_cksum[file_name] == local_cksum[file_name] |
1706 |
_ErrorIf(test1 and must_have, self.ENODEFILECHECK, node, |
1707 |
"file '%s' missing", file_name)
|
1708 |
_ErrorIf(test2 and must_have, self.ENODEFILECHECK, node, |
1709 |
"file '%s' has wrong checksum", file_name)
|
1710 |
# not candidate and this is not a must-have file
|
1711 |
_ErrorIf(test2 and not must_have, self.ENODEFILECHECK, node, |
1712 |
"file '%s' should not exist on non master"
|
1713 |
" candidates (and the file is outdated)", file_name)
|
1714 |
# all good, except non-master/non-must have combination
|
1715 |
_ErrorIf(test3 and not must_have, self.ENODEFILECHECK, node, |
1716 |
"file '%s' should not exist"
|
1717 |
" on non master candidates", file_name)
|
1718 |
|
1719 |
def _VerifyNodeDrbd(self, ninfo, nresult, instanceinfo, drbd_helper, |
1720 |
drbd_map): |
1721 |
"""Verifies and the node DRBD status.
|
1722 |
|
1723 |
@type ninfo: L{objects.Node}
|
1724 |
@param ninfo: the node to check
|
1725 |
@param nresult: the remote results for the node
|
1726 |
@param instanceinfo: the dict of instances
|
1727 |
@param drbd_helper: the configured DRBD usermode helper
|
1728 |
@param drbd_map: the DRBD map as returned by
|
1729 |
L{ganeti.config.ConfigWriter.ComputeDRBDMap}
|
1730 |
|
1731 |
"""
|
1732 |
node = ninfo.name |
1733 |
_ErrorIf = self._ErrorIf # pylint: disable-msg=C0103 |
1734 |
|
1735 |
if drbd_helper:
|
1736 |
helper_result = nresult.get(constants.NV_DRBDHELPER, None)
|
1737 |
test = (helper_result == None)
|
1738 |
_ErrorIf(test, self.ENODEDRBDHELPER, node,
|
1739 |
"no drbd usermode helper returned")
|
1740 |
if helper_result:
|
1741 |
status, payload = helper_result |
1742 |
test = not status
|
1743 |
_ErrorIf(test, self.ENODEDRBDHELPER, node,
|
1744 |
"drbd usermode helper check unsuccessful: %s", payload)
|
1745 |
test = status and (payload != drbd_helper)
|
1746 |
_ErrorIf(test, self.ENODEDRBDHELPER, node,
|
1747 |
"wrong drbd usermode helper: %s", payload)
|
1748 |
|
1749 |
# compute the DRBD minors
|
1750 |
node_drbd = {} |
1751 |
for minor, instance in drbd_map[node].items(): |
1752 |
test = instance not in instanceinfo |
1753 |
_ErrorIf(test, self.ECLUSTERCFG, None, |
1754 |
"ghost instance '%s' in temporary DRBD map", instance)
|
1755 |
# ghost instance should not be running, but otherwise we
|
1756 |
# don't give double warnings (both ghost instance and
|
1757 |
# unallocated minor in use)
|
1758 |
if test:
|
1759 |
node_drbd[minor] = (instance, False)
|
1760 |
else:
|
1761 |
instance = instanceinfo[instance] |
1762 |
node_drbd[minor] = (instance.name, instance.admin_up) |
1763 |
|
1764 |
# and now check them
|
1765 |
used_minors = nresult.get(constants.NV_DRBDLIST, []) |
1766 |
test = not isinstance(used_minors, (tuple, list)) |
1767 |
_ErrorIf(test, self.ENODEDRBD, node,
|
1768 |
"cannot parse drbd status file: %s", str(used_minors)) |
1769 |
if test:
|
1770 |
# we cannot check drbd status
|
1771 |
return
|
1772 |
|
1773 |
for minor, (iname, must_exist) in node_drbd.items(): |
1774 |
test = minor not in used_minors and must_exist |
1775 |
_ErrorIf(test, self.ENODEDRBD, node,
|
1776 |
"drbd minor %d of instance %s is not active", minor, iname)
|
1777 |
for minor in used_minors: |
1778 |
test = minor not in node_drbd |
1779 |
_ErrorIf(test, self.ENODEDRBD, node,
|
1780 |
"unallocated drbd minor %d is in use", minor)
|
1781 |
|
1782 |
def _UpdateNodeOS(self, ninfo, nresult, nimg): |
1783 |
"""Builds the node OS structures.
|
1784 |
|
1785 |
@type ninfo: L{objects.Node}
|
1786 |
@param ninfo: the node to check
|
1787 |
@param nresult: the remote results for the node
|
1788 |
@param nimg: the node image object
|
1789 |
|
1790 |
"""
|
1791 |
node = ninfo.name |
1792 |
_ErrorIf = self._ErrorIf # pylint: disable-msg=C0103 |
1793 |
|
1794 |
remote_os = nresult.get(constants.NV_OSLIST, None)
|
1795 |
test = (not isinstance(remote_os, list) or |
1796 |
not compat.all(isinstance(v, list) and len(v) == 7 |
1797 |
for v in remote_os)) |
1798 |
|
1799 |
_ErrorIf(test, self.ENODEOS, node,
|
1800 |
"node hasn't returned valid OS data")
|
1801 |
|
1802 |
nimg.os_fail = test |
1803 |
|
1804 |
if test:
|
1805 |
return
|
1806 |
|
1807 |
os_dict = {} |
1808 |
|
1809 |
for (name, os_path, status, diagnose,
|
1810 |
variants, parameters, api_ver) in nresult[constants.NV_OSLIST]:
|
1811 |
|
1812 |
if name not in os_dict: |
1813 |
os_dict[name] = [] |
1814 |
|
1815 |
# parameters is a list of lists instead of list of tuples due to
|
1816 |
# JSON lacking a real tuple type, fix it:
|
1817 |
parameters = [tuple(v) for v in parameters] |
1818 |
os_dict[name].append((os_path, status, diagnose, |
1819 |
set(variants), set(parameters), set(api_ver))) |
1820 |
|
1821 |
nimg.oslist = os_dict |
1822 |
|
1823 |
def _VerifyNodeOS(self, ninfo, nimg, base): |
1824 |
"""Verifies the node OS list.
|
1825 |
|
1826 |
@type ninfo: L{objects.Node}
|
1827 |
@param ninfo: the node to check
|
1828 |
@param nimg: the node image object
|
1829 |
@param base: the 'template' node we match against (e.g. from the master)
|
1830 |
|
1831 |
"""
|
1832 |
node = ninfo.name |
1833 |
_ErrorIf = self._ErrorIf # pylint: disable-msg=C0103 |
1834 |
|
1835 |
assert not nimg.os_fail, "Entered _VerifyNodeOS with failed OS rpc?" |
1836 |
|
1837 |
for os_name, os_data in nimg.oslist.items(): |
1838 |
assert os_data, "Empty OS status for OS %s?!" % os_name |
1839 |
f_path, f_status, f_diag, f_var, f_param, f_api = os_data[0]
|
1840 |
_ErrorIf(not f_status, self.ENODEOS, node, |
1841 |
"Invalid OS %s (located at %s): %s", os_name, f_path, f_diag)
|
1842 |
_ErrorIf(len(os_data) > 1, self.ENODEOS, node, |
1843 |
"OS '%s' has multiple entries (first one shadows the rest): %s",
|
1844 |
os_name, utils.CommaJoin([v[0] for v in os_data])) |
1845 |
# this will catched in backend too
|
1846 |
_ErrorIf(compat.any(v >= constants.OS_API_V15 for v in f_api) |
1847 |
and not f_var, self.ENODEOS, node, |
1848 |
"OS %s with API at least %d does not declare any variant",
|
1849 |
os_name, constants.OS_API_V15) |
1850 |
# comparisons with the 'base' image
|
1851 |
test = os_name not in base.oslist |
1852 |
_ErrorIf(test, self.ENODEOS, node,
|
1853 |
"Extra OS %s not present on reference node (%s)",
|
1854 |
os_name, base.name) |
1855 |
if test:
|
1856 |
continue
|
1857 |
assert base.oslist[os_name], "Base node has empty OS status?" |
1858 |
_, b_status, _, b_var, b_param, b_api = base.oslist[os_name][0]
|
1859 |
if not b_status: |
1860 |
# base OS is invalid, skipping
|
1861 |
continue
|
1862 |
for kind, a, b in [("API version", f_api, b_api), |
1863 |
("variants list", f_var, b_var),
|
1864 |
("parameters", f_param, b_param)]:
|
1865 |
_ErrorIf(a != b, self.ENODEOS, node,
|
1866 |
"OS %s %s differs from reference node %s: %s vs. %s",
|
1867 |
kind, os_name, base.name, |
1868 |
utils.CommaJoin(a), utils.CommaJoin(b)) |
1869 |
|
1870 |
# check any missing OSes
|
1871 |
missing = set(base.oslist.keys()).difference(nimg.oslist.keys())
|
1872 |
_ErrorIf(missing, self.ENODEOS, node,
|
1873 |
"OSes present on reference node %s but missing on this node: %s",
|
1874 |
base.name, utils.CommaJoin(missing)) |
1875 |
|
1876 |
def _UpdateNodeVolumes(self, ninfo, nresult, nimg, vg_name): |
1877 |
"""Verifies and updates the node volume data.
|
1878 |
|
1879 |
This function will update a L{NodeImage}'s internal structures
|
1880 |
with data from the remote call.
|
1881 |
|
1882 |
@type ninfo: L{objects.Node}
|
1883 |
@param ninfo: the node to check
|
1884 |
@param nresult: the remote results for the node
|
1885 |
@param nimg: the node image object
|
1886 |
@param vg_name: the configured VG name
|
1887 |
|
1888 |
"""
|
1889 |
node = ninfo.name |
1890 |
_ErrorIf = self._ErrorIf # pylint: disable-msg=C0103 |
1891 |
|
1892 |
nimg.lvm_fail = True
|
1893 |
lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
|
1894 |
if vg_name is None: |
1895 |
pass
|
1896 |
elif isinstance(lvdata, basestring): |
1897 |
_ErrorIf(True, self.ENODELVM, node, "LVM problem on node: %s", |
1898 |
utils.SafeEncode(lvdata)) |
1899 |
elif not isinstance(lvdata, dict): |
1900 |
_ErrorIf(True, self.ENODELVM, node, "rpc call to node failed (lvlist)") |
1901 |
else:
|
1902 |
nimg.volumes = lvdata |
1903 |
nimg.lvm_fail = False
|
1904 |
|
1905 |
def _UpdateNodeInstances(self, ninfo, nresult, nimg): |
1906 |
"""Verifies and updates the node instance list.
|
1907 |
|
1908 |
If the listing was successful, then updates this node's instance
|
1909 |
list. Otherwise, it marks the RPC call as failed for the instance
|
1910 |
list key.
|
1911 |
|
1912 |
@type ninfo: L{objects.Node}
|
1913 |
@param ninfo: the node to check
|
1914 |
@param nresult: the remote results for the node
|
1915 |
@param nimg: the node image object
|
1916 |
|
1917 |
"""
|
1918 |
idata = nresult.get(constants.NV_INSTANCELIST, None)
|
1919 |
test = not isinstance(idata, list) |
1920 |
self._ErrorIf(test, self.ENODEHV, ninfo.name, "rpc call to node failed" |
1921 |
" (instancelist): %s", utils.SafeEncode(str(idata))) |
1922 |
if test:
|
1923 |
nimg.hyp_fail = True
|
1924 |
else:
|
1925 |
nimg.instances = idata |
1926 |
|
1927 |
def _UpdateNodeInfo(self, ninfo, nresult, nimg, vg_name): |
1928 |
"""Verifies and computes a node information map
|
1929 |
|
1930 |
@type ninfo: L{objects.Node}
|
1931 |
@param ninfo: the node to check
|
1932 |
@param nresult: the remote results for the node
|
1933 |
@param nimg: the node image object
|
1934 |
@param vg_name: the configured VG name
|
1935 |
|
1936 |
"""
|
1937 |
node = ninfo.name |
1938 |
_ErrorIf = self._ErrorIf # pylint: disable-msg=C0103 |
1939 |
|
1940 |
# try to read free memory (from the hypervisor)
|
1941 |
hv_info = nresult.get(constants.NV_HVINFO, None)
|
1942 |
test = not isinstance(hv_info, dict) or "memory_free" not in hv_info |
1943 |
_ErrorIf(test, self.ENODEHV, node, "rpc call to node failed (hvinfo)") |
1944 |
if not test: |
1945 |
try:
|
1946 |
nimg.mfree = int(hv_info["memory_free"]) |
1947 |
except (ValueError, TypeError): |
1948 |
_ErrorIf(True, self.ENODERPC, node, |
1949 |
"node returned invalid nodeinfo, check hypervisor")
|
1950 |
|
1951 |
# FIXME: devise a free space model for file based instances as well
|
1952 |
if vg_name is not None: |
1953 |
test = (constants.NV_VGLIST not in nresult or |
1954 |
vg_name not in nresult[constants.NV_VGLIST]) |
1955 |
_ErrorIf(test, self.ENODELVM, node,
|
1956 |
"node didn't return data for the volume group '%s'"
|
1957 |
" - it is either missing or broken", vg_name)
|
1958 |
if not test: |
1959 |
try:
|
1960 |
nimg.dfree = int(nresult[constants.NV_VGLIST][vg_name])
|
1961 |
except (ValueError, TypeError): |
1962 |
_ErrorIf(True, self.ENODERPC, node, |
1963 |
"node returned invalid LVM info, check LVM status")
|
1964 |
|
1965 |
def BuildHooksEnv(self): |
1966 |
"""Build hooks env.
|
1967 |
|
1968 |
Cluster-Verify hooks just ran in the post phase and their failure makes
|
1969 |
the output be logged in the verify output and the verification to fail.
|
1970 |
|
1971 |
"""
|
1972 |
all_nodes = self.cfg.GetNodeList()
|
1973 |
env = { |
1974 |
"CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags()) |
1975 |
} |
1976 |
for node in self.cfg.GetAllNodesInfo().values(): |
1977 |
env["NODE_TAGS_%s" % node.name] = " ".join(node.GetTags()) |
1978 |
|
1979 |
return env, [], all_nodes
|
1980 |
|
1981 |
def Exec(self, feedback_fn): |
1982 |
"""Verify integrity of cluster, performing various test on nodes.
|
1983 |
|
1984 |
"""
|
1985 |
self.bad = False |
1986 |
_ErrorIf = self._ErrorIf # pylint: disable-msg=C0103 |
1987 |
verbose = self.op.verbose
|
1988 |
self._feedback_fn = feedback_fn
|
1989 |
feedback_fn("* Verifying global settings")
|
1990 |
for msg in self.cfg.VerifyConfig(): |
1991 |
_ErrorIf(True, self.ECLUSTERCFG, None, msg) |
1992 |
|
1993 |
# Check the cluster certificates
|
1994 |
for cert_filename in constants.ALL_CERT_FILES: |
1995 |
(errcode, msg) = _VerifyCertificate(cert_filename) |
1996 |
_ErrorIf(errcode, self.ECLUSTERCERT, None, msg, code=errcode) |
1997 |
|
1998 |
vg_name = self.cfg.GetVGName()
|
1999 |
drbd_helper = self.cfg.GetDRBDHelper()
|
2000 |
hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
|
2001 |
cluster = self.cfg.GetClusterInfo()
|
2002 |
nodelist = utils.NiceSort(self.cfg.GetNodeList())
|
2003 |
nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist] |
2004 |
instancelist = utils.NiceSort(self.cfg.GetInstanceList())
|
2005 |
instanceinfo = dict((iname, self.cfg.GetInstanceInfo(iname)) |
2006 |
for iname in instancelist) |
2007 |
i_non_redundant = [] # Non redundant instances
|
2008 |
i_non_a_balanced = [] # Non auto-balanced instances
|
2009 |
n_offline = 0 # Count of offline nodes |
2010 |
n_drained = 0 # Count of nodes being drained |
2011 |
node_vol_should = {} |
2012 |
|
2013 |
# FIXME: verify OS list
|
2014 |
# do local checksums
|
2015 |
master_files = [constants.CLUSTER_CONF_FILE] |
2016 |
master_node = self.master_node = self.cfg.GetMasterNode() |
2017 |
master_ip = self.cfg.GetMasterIP()
|
2018 |
|
2019 |
file_names = ssconf.SimpleStore().GetFileList() |
2020 |
file_names.extend(constants.ALL_CERT_FILES) |
2021 |
file_names.extend(master_files) |
2022 |
if cluster.modify_etc_hosts:
|
2023 |
file_names.append(constants.ETC_HOSTS) |
2024 |
|
2025 |
local_checksums = utils.FingerprintFiles(file_names) |
2026 |
|
2027 |
feedback_fn("* Gathering data (%d nodes)" % len(nodelist)) |
2028 |
node_verify_param = { |
2029 |
constants.NV_FILELIST: file_names, |
2030 |
constants.NV_NODELIST: [node.name for node in nodeinfo |
2031 |
if not node.offline], |
2032 |
constants.NV_HYPERVISOR: hypervisors, |
2033 |
constants.NV_NODENETTEST: [(node.name, node.primary_ip, |
2034 |
node.secondary_ip) for node in nodeinfo |
2035 |
if not node.offline], |
2036 |
constants.NV_INSTANCELIST: hypervisors, |
2037 |
constants.NV_VERSION: None,
|
2038 |
constants.NV_HVINFO: self.cfg.GetHypervisorType(),
|
2039 |
constants.NV_NODESETUP: None,
|
2040 |
constants.NV_TIME: None,
|
2041 |
constants.NV_MASTERIP: (master_node, master_ip), |
2042 |
constants.NV_OSLIST: None,
|
2043 |
} |
2044 |
|
2045 |
if vg_name is not None: |
2046 |
node_verify_param[constants.NV_VGLIST] = None
|
2047 |
node_verify_param[constants.NV_LVLIST] = vg_name |
2048 |
node_verify_param[constants.NV_PVLIST] = [vg_name] |
2049 |
node_verify_param[constants.NV_DRBDLIST] = None
|
2050 |
|
2051 |
if drbd_helper:
|
2052 |
node_verify_param[constants.NV_DRBDHELPER] = drbd_helper |
2053 |
|
2054 |
# Build our expected cluster state
|
2055 |
node_image = dict((node.name, self.NodeImage(offline=node.offline, |
2056 |
name=node.name)) |
2057 |
for node in nodeinfo) |
2058 |
|
2059 |
for instance in instancelist: |
2060 |
inst_config = instanceinfo[instance] |
2061 |
|
2062 |
for nname in inst_config.all_nodes: |
2063 |
if nname not in node_image: |
2064 |
# ghost node
|
2065 |
gnode = self.NodeImage(name=nname)
|
2066 |
gnode.ghost = True
|
2067 |
node_image[nname] = gnode |
2068 |
|
2069 |
inst_config.MapLVsByNode(node_vol_should) |
2070 |
|
2071 |
pnode = inst_config.primary_node |
2072 |
node_image[pnode].pinst.append(instance) |
2073 |
|
2074 |
for snode in inst_config.secondary_nodes: |
2075 |
nimg = node_image[snode] |
2076 |
nimg.sinst.append(instance) |
2077 |
if pnode not in nimg.sbp: |
2078 |
nimg.sbp[pnode] = [] |
2079 |
nimg.sbp[pnode].append(instance) |
2080 |
|
2081 |
# At this point, we have the in-memory data structures complete,
|
2082 |
# except for the runtime information, which we'll gather next
|
2083 |
|
2084 |
# Due to the way our RPC system works, exact response times cannot be
|
2085 |
# guaranteed (e.g. a broken node could run into a timeout). By keeping the
|
2086 |
# time before and after executing the request, we can at least have a time
|
2087 |
# window.
|
2088 |
nvinfo_starttime = time.time() |
2089 |
all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
|
2090 |
self.cfg.GetClusterName())
|
2091 |
nvinfo_endtime = time.time() |
2092 |
|
2093 |
all_drbd_map = self.cfg.ComputeDRBDMap()
|
2094 |
|
2095 |
feedback_fn("* Verifying node status")
|
2096 |
|
2097 |
refos_img = None
|
2098 |
|
2099 |
for node_i in nodeinfo: |
2100 |
node = node_i.name |
2101 |
nimg = node_image[node] |
2102 |
|
2103 |
if node_i.offline:
|
2104 |
if verbose:
|
2105 |
feedback_fn("* Skipping offline node %s" % (node,))
|
2106 |
n_offline += 1
|
2107 |
continue
|
2108 |
|
2109 |
if node == master_node:
|
2110 |
ntype = "master"
|
2111 |
elif node_i.master_candidate:
|
2112 |
ntype = "master candidate"
|
2113 |
elif node_i.drained:
|
2114 |
ntype = "drained"
|
2115 |
n_drained += 1
|
2116 |
else:
|
2117 |
ntype = "regular"
|
2118 |
if verbose:
|
2119 |
feedback_fn("* Verifying node %s (%s)" % (node, ntype))
|
2120 |
|
2121 |
msg = all_nvinfo[node].fail_msg |
2122 |
_ErrorIf(msg, self.ENODERPC, node, "while contacting node: %s", msg) |
2123 |
if msg:
|
2124 |
nimg.rpc_fail = True
|
2125 |
continue
|
2126 |
|
2127 |
nresult = all_nvinfo[node].payload |
2128 |
|
2129 |
nimg.call_ok = self._VerifyNode(node_i, nresult)
|
2130 |
self._VerifyNodeNetwork(node_i, nresult)
|
2131 |
self._VerifyNodeLVM(node_i, nresult, vg_name)
|
2132 |
self._VerifyNodeFiles(node_i, nresult, file_names, local_checksums,
|
2133 |
master_files) |
2134 |
self._VerifyNodeDrbd(node_i, nresult, instanceinfo, drbd_helper,
|
2135 |
all_drbd_map) |
2136 |
self._VerifyNodeTime(node_i, nresult, nvinfo_starttime, nvinfo_endtime)
|
2137 |
|
2138 |
self._UpdateNodeVolumes(node_i, nresult, nimg, vg_name)
|
2139 |
self._UpdateNodeInstances(node_i, nresult, nimg)
|
2140 |
self._UpdateNodeInfo(node_i, nresult, nimg, vg_name)
|
2141 |
self._UpdateNodeOS(node_i, nresult, nimg)
|
2142 |
if not nimg.os_fail: |
2143 |
if refos_img is None: |
2144 |
refos_img = nimg |
2145 |
self._VerifyNodeOS(node_i, nimg, refos_img)
|
2146 |
|
2147 |
feedback_fn("* Verifying instance status")
|
2148 |
for instance in instancelist: |
2149 |
if verbose:
|
2150 |
feedback_fn("* Verifying instance %s" % instance)
|
2151 |
inst_config = instanceinfo[instance] |
2152 |
self._VerifyInstance(instance, inst_config, node_image)
|
2153 |
inst_nodes_offline = [] |
2154 |
|
2155 |
pnode = inst_config.primary_node |
2156 |
pnode_img = node_image[pnode] |
2157 |
_ErrorIf(pnode_img.rpc_fail and not pnode_img.offline, |
2158 |
self.ENODERPC, pnode, "instance %s, connection to" |
2159 |
" primary node failed", instance)
|
2160 |
|
2161 |
if pnode_img.offline:
|
2162 |
inst_nodes_offline.append(pnode) |
2163 |
|
2164 |
# If the instance is non-redundant we cannot survive losing its primary
|
2165 |
# node, so we are not N+1 compliant. On the other hand we have no disk
|
2166 |
# templates with more than one secondary so that situation is not well
|
2167 |
# supported either.
|
2168 |
# FIXME: does not support file-backed instances
|
2169 |
if not inst_config.secondary_nodes: |
2170 |
i_non_redundant.append(instance) |
2171 |
_ErrorIf(len(inst_config.secondary_nodes) > 1, self.EINSTANCELAYOUT, |
2172 |
instance, "instance has multiple secondary nodes: %s",
|
2173 |
utils.CommaJoin(inst_config.secondary_nodes), |
2174 |
code=self.ETYPE_WARNING)
|
2175 |
|
2176 |
if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]: |
2177 |
i_non_a_balanced.append(instance) |
2178 |
|
2179 |
for snode in inst_config.secondary_nodes: |
2180 |
s_img = node_image[snode] |
2181 |
_ErrorIf(s_img.rpc_fail and not s_img.offline, self.ENODERPC, snode, |
2182 |
"instance %s, connection to secondary node failed", instance)
|
2183 |
|
2184 |
if s_img.offline:
|
2185 |
inst_nodes_offline.append(snode) |
2186 |
|
2187 |
# warn that the instance lives on offline nodes
|
2188 |
_ErrorIf(inst_nodes_offline, self.EINSTANCEBADNODE, instance,
|
2189 |
"instance lives on offline node(s) %s",
|
2190 |
utils.CommaJoin(inst_nodes_offline)) |
2191 |
# ... or ghost nodes
|
2192 |
for node in inst_config.all_nodes: |
2193 |
_ErrorIf(node_image[node].ghost, self.EINSTANCEBADNODE, instance,
|
2194 |
"instance lives on ghost node %s", node)
|
2195 |
|
2196 |
feedback_fn("* Verifying orphan volumes")
|
2197 |
self._VerifyOrphanVolumes(node_vol_should, node_image)
|
2198 |
|
2199 |
feedback_fn("* Verifying orphan instances")
|
2200 |
self._VerifyOrphanInstances(instancelist, node_image)
|
2201 |
|
2202 |
if constants.VERIFY_NPLUSONE_MEM not in self.op.skip_checks: |
2203 |
feedback_fn("* Verifying N+1 Memory redundancy")
|
2204 |
self._VerifyNPlusOneMemory(node_image, instanceinfo)
|
2205 |
|
2206 |
feedback_fn("* Other Notes")
|
2207 |
if i_non_redundant:
|
2208 |
feedback_fn(" - NOTICE: %d non-redundant instance(s) found."
|
2209 |
% len(i_non_redundant))
|
2210 |
|
2211 |
if i_non_a_balanced:
|
2212 |
feedback_fn(" - NOTICE: %d non-auto-balanced instance(s) found."
|
2213 |
% len(i_non_a_balanced))
|
2214 |
|
2215 |
if n_offline:
|
2216 |
feedback_fn(" - NOTICE: %d offline node(s) found." % n_offline)
|
2217 |
|
2218 |
if n_drained:
|
2219 |
feedback_fn(" - NOTICE: %d drained node(s) found." % n_drained)
|
2220 |
|
2221 |
return not self.bad |
2222 |
|
2223 |
def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result): |
2224 |
"""Analyze the post-hooks' result
|
2225 |
|
2226 |
This method analyses the hook result, handles it, and sends some
|
2227 |
nicely-formatted feedback back to the user.
|
2228 |
|
2229 |
@param phase: one of L{constants.HOOKS_PHASE_POST} or
|
2230 |
L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
|
2231 |
@param hooks_results: the results of the multi-node hooks rpc call
|
2232 |
@param feedback_fn: function used send feedback back to the caller
|
2233 |
@param lu_result: previous Exec result
|
2234 |
@return: the new Exec result, based on the previous result
|
2235 |
and hook results
|
2236 |
|
2237 |
"""
|
2238 |
# We only really run POST phase hooks, and are only interested in
|
2239 |
# their results
|
2240 |
if phase == constants.HOOKS_PHASE_POST:
|
2241 |
# Used to change hooks' output to proper indentation
|
2242 |
indent_re = re.compile('^', re.M)
|
2243 |
feedback_fn("* Hooks Results")
|
2244 |
assert hooks_results, "invalid result from hooks" |
2245 |
|
2246 |
for node_name in hooks_results: |
2247 |
res = hooks_results[node_name] |
2248 |
msg = res.fail_msg |
2249 |
test = msg and not res.offline |
2250 |
self._ErrorIf(test, self.ENODEHOOKS, node_name, |
2251 |
"Communication failure in hooks execution: %s", msg)
|
2252 |
if res.offline or msg: |
2253 |
# No need to investigate payload if node is offline or gave an error.
|
2254 |
# override manually lu_result here as _ErrorIf only
|
2255 |
# overrides self.bad
|
2256 |
lu_result = 1
|
2257 |
continue
|
2258 |
for script, hkr, output in res.payload: |
2259 |
test = hkr == constants.HKR_FAIL |
2260 |
self._ErrorIf(test, self.ENODEHOOKS, node_name, |
2261 |
"Script %s failed, output:", script)
|
2262 |
if test:
|
2263 |
output = indent_re.sub(' ', output)
|
2264 |
feedback_fn("%s" % output)
|
2265 |
lu_result = 0
|
2266 |
|
2267 |
return lu_result
|
2268 |
|
2269 |
|
2270 |
class LUVerifyDisks(NoHooksLU): |
2271 |
"""Verifies the cluster disks status.
|
2272 |
|
2273 |
"""
|
2274 |
REQ_BGL = False
|
2275 |
|
2276 |
def ExpandNames(self): |
2277 |
self.needed_locks = {
|
2278 |
locking.LEVEL_NODE: locking.ALL_SET, |
2279 |
locking.LEVEL_INSTANCE: locking.ALL_SET, |
2280 |
} |
2281 |
self.share_locks = dict.fromkeys(locking.LEVELS, 1) |
2282 |
|
2283 |
def Exec(self, feedback_fn): |
2284 |
"""Verify integrity of cluster disks.
|
2285 |
|
2286 |
@rtype: tuple of three items
|
2287 |
@return: a tuple of (dict of node-to-node_error, list of instances
|
2288 |
which need activate-disks, dict of instance: (node, volume) for
|
2289 |
missing volumes
|
2290 |
|
2291 |
"""
|
2292 |
result = res_nodes, res_instances, res_missing = {}, [], {} |
2293 |
|
2294 |
vg_name = self.cfg.GetVGName()
|
2295 |
nodes = utils.NiceSort(self.cfg.GetNodeList())
|
2296 |
instances = [self.cfg.GetInstanceInfo(name)
|
2297 |
for name in self.cfg.GetInstanceList()] |
2298 |
|
2299 |
nv_dict = {} |
2300 |
for inst in instances: |
2301 |
inst_lvs = {} |
2302 |
if (not inst.admin_up or |
2303 |
inst.disk_template not in constants.DTS_NET_MIRROR): |
2304 |
continue
|
2305 |
inst.MapLVsByNode(inst_lvs) |
2306 |
# transform { iname: {node: [vol,],},} to {(node, vol): iname}
|
2307 |
for node, vol_list in inst_lvs.iteritems(): |
2308 |
for vol in vol_list: |
2309 |
nv_dict[(node, vol)] = inst |
2310 |
|
2311 |
if not nv_dict: |
2312 |
return result
|
2313 |
|
2314 |
node_lvs = self.rpc.call_lv_list(nodes, vg_name)
|
2315 |
|
2316 |
for node in nodes: |
2317 |
# node_volume
|
2318 |
node_res = node_lvs[node] |
2319 |
if node_res.offline:
|
2320 |
continue
|
2321 |
msg = node_res.fail_msg |
2322 |
if msg:
|
2323 |
logging.warning("Error enumerating LVs on node %s: %s", node, msg)
|
2324 |
res_nodes[node] = msg |
2325 |
continue
|
2326 |
|
2327 |
lvs = node_res.payload |
2328 |
for lv_name, (_, _, lv_online) in lvs.items(): |
2329 |
inst = nv_dict.pop((node, lv_name), None)
|
2330 |
if (not lv_online and inst is not None |
2331 |
and inst.name not in res_instances): |
2332 |
res_instances.append(inst.name) |
2333 |
|
2334 |
# any leftover items in nv_dict are missing LVs, let's arrange the
|
2335 |
# data better
|
2336 |
for key, inst in nv_dict.iteritems(): |
2337 |
if inst.name not in res_missing: |
2338 |
res_missing[inst.name] = [] |
2339 |
res_missing[inst.name].append(key) |
2340 |
|
2341 |
return result
|
2342 |
|
2343 |
|
2344 |
class LURepairDiskSizes(NoHooksLU): |
2345 |
"""Verifies the cluster disks sizes.
|
2346 |
|
2347 |
"""
|
2348 |
_OP_PARAMS = [("instances", _EmptyList, _TListOf(_TNonEmptyString))]
|
2349 |
REQ_BGL = False
|
2350 |
|
2351 |
def ExpandNames(self): |
2352 |
if self.op.instances: |
2353 |
self.wanted_names = []
|
2354 |
for name in self.op.instances: |
2355 |
full_name = _ExpandInstanceName(self.cfg, name)
|
2356 |
self.wanted_names.append(full_name)
|
2357 |
self.needed_locks = {
|
2358 |
locking.LEVEL_NODE: [], |
2359 |
locking.LEVEL_INSTANCE: self.wanted_names,
|
2360 |
} |
2361 |
self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
|
2362 |
else:
|
2363 |
self.wanted_names = None |
2364 |
self.needed_locks = {
|
2365 |
locking.LEVEL_NODE: locking.ALL_SET, |
2366 |
locking.LEVEL_INSTANCE: locking.ALL_SET, |
2367 |
} |
2368 |
self.share_locks = dict(((i, 1) for i in locking.LEVELS)) |
2369 |
|
2370 |
def DeclareLocks(self, level): |
2371 |
if level == locking.LEVEL_NODE and self.wanted_names is not None: |
2372 |
self._LockInstancesNodes(primary_only=True) |
2373 |
|
2374 |
def CheckPrereq(self): |
2375 |
"""Check prerequisites.
|
2376 |
|
2377 |
This only checks the optional instance list against the existing names.
|
2378 |
|
2379 |
"""
|
2380 |
if self.wanted_names is None: |
2381 |
self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE] |
2382 |
|
2383 |
self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name |
2384 |
in self.wanted_names] |
2385 |
|
2386 |
def _EnsureChildSizes(self, disk): |
2387 |
"""Ensure children of the disk have the needed disk size.
|
2388 |
|
2389 |
This is valid mainly for DRBD8 and fixes an issue where the
|
2390 |
children have smaller disk size.
|
2391 |
|
2392 |
@param disk: an L{ganeti.objects.Disk} object
|
2393 |
|
2394 |
"""
|
2395 |
if disk.dev_type == constants.LD_DRBD8:
|
2396 |
assert disk.children, "Empty children for DRBD8?" |
2397 |
fchild = disk.children[0]
|
2398 |
mismatch = fchild.size < disk.size |
2399 |
if mismatch:
|
2400 |
self.LogInfo("Child disk has size %d, parent %d, fixing", |
2401 |
fchild.size, disk.size) |
2402 |
fchild.size = disk.size |
2403 |
|
2404 |
# and we recurse on this child only, not on the metadev
|
2405 |
return self._EnsureChildSizes(fchild) or mismatch |
2406 |
else:
|
2407 |
return False |
2408 |
|
2409 |
def Exec(self, feedback_fn): |
2410 |
"""Verify the size of cluster disks.
|
2411 |
|
2412 |
"""
|
2413 |
# TODO: check child disks too
|
2414 |
# TODO: check differences in size between primary/secondary nodes
|
2415 |
per_node_disks = {} |
2416 |
for instance in self.wanted_instances: |
2417 |
pnode = instance.primary_node |
2418 |
if pnode not in per_node_disks: |
2419 |
per_node_disks[pnode] = [] |
2420 |
for idx, disk in enumerate(instance.disks): |
2421 |
per_node_disks[pnode].append((instance, idx, disk)) |
2422 |
|
2423 |
changed = [] |
2424 |
for node, dskl in per_node_disks.items(): |
2425 |
newl = [v[2].Copy() for v in dskl] |
2426 |
for dsk in newl: |
2427 |
self.cfg.SetDiskID(dsk, node)
|
2428 |
result = self.rpc.call_blockdev_getsizes(node, newl)
|
2429 |
if result.fail_msg:
|
2430 |
self.LogWarning("Failure in blockdev_getsizes call to node" |
2431 |
" %s, ignoring", node)
|
2432 |
continue
|
2433 |
if len(result.data) != len(dskl): |
2434 |
self.LogWarning("Invalid result from node %s, ignoring node results", |
2435 |
node) |
2436 |
continue
|
2437 |
for ((instance, idx, disk), size) in zip(dskl, result.data): |
2438 |
if size is None: |
2439 |
self.LogWarning("Disk %d of instance %s did not return size" |
2440 |
" information, ignoring", idx, instance.name)
|
2441 |
continue
|
2442 |
if not isinstance(size, (int, long)): |
2443 |
self.LogWarning("Disk %d of instance %s did not return valid" |
2444 |
" size information, ignoring", idx, instance.name)
|
2445 |
continue
|
2446 |
size = size >> 20
|
2447 |
if size != disk.size:
|
2448 |
self.LogInfo("Disk %d of instance %s has mismatched size," |
2449 |
" correcting: recorded %d, actual %d", idx,
|
2450 |
instance.name, disk.size, size) |
2451 |
disk.size = size |
2452 |
self.cfg.Update(instance, feedback_fn)
|
2453 |
changed.append((instance.name, idx, size)) |
2454 |
if self._EnsureChildSizes(disk): |
2455 |
self.cfg.Update(instance, feedback_fn)
|
2456 |
changed.append((instance.name, idx, disk.size)) |
2457 |
return changed
|
2458 |
|
2459 |
|
2460 |
class LURenameCluster(LogicalUnit): |
2461 |
"""Rename the cluster.
|
2462 |
|
2463 |
"""
|
2464 |
HPATH = "cluster-rename"
|
2465 |
HTYPE = constants.HTYPE_CLUSTER |
2466 |
_OP_PARAMS = [("name", _NoDefault, _TNonEmptyString)]
|
2467 |
|
2468 |
def BuildHooksEnv(self): |
2469 |
"""Build hooks env.
|
2470 |
|
2471 |
"""
|
2472 |
env = { |
2473 |
"OP_TARGET": self.cfg.GetClusterName(), |
2474 |
"NEW_NAME": self.op.name, |
2475 |
} |
2476 |
mn = self.cfg.GetMasterNode()
|
2477 |
all_nodes = self.cfg.GetNodeList()
|
2478 |
return env, [mn], all_nodes
|
2479 |
|
2480 |
def CheckPrereq(self): |
2481 |
"""Verify that the passed name is a valid one.
|
2482 |
|
2483 |
"""
|
2484 |
hostname = utils.GetHostInfo(self.op.name)
|
2485 |
|
2486 |
new_name = hostname.name |
2487 |
self.ip = new_ip = hostname.ip
|
2488 |
old_name = self.cfg.GetClusterName()
|
2489 |
old_ip = self.cfg.GetMasterIP()
|
2490 |
if new_name == old_name and new_ip == old_ip: |
2491 |
raise errors.OpPrereqError("Neither the name nor the IP address of the" |
2492 |
" cluster has changed",
|
2493 |
errors.ECODE_INVAL) |
2494 |
if new_ip != old_ip:
|
2495 |
if utils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
|
2496 |
raise errors.OpPrereqError("The given cluster IP address (%s) is" |
2497 |
" reachable on the network. Aborting." %
|
2498 |
new_ip, errors.ECODE_NOTUNIQUE) |
2499 |
|
2500 |
self.op.name = new_name
|
2501 |
|
2502 |
def Exec(self, feedback_fn): |
2503 |
"""Rename the cluster.
|
2504 |
|
2505 |
"""
|
2506 |
clustername = self.op.name
|
2507 |
ip = self.ip
|
2508 |
|
2509 |
# shutdown the master IP
|
2510 |
master = self.cfg.GetMasterNode()
|
2511 |
result = self.rpc.call_node_stop_master(master, False) |
2512 |
result.Raise("Could not disable the master role")
|
2513 |
|
2514 |
try:
|
2515 |
cluster = self.cfg.GetClusterInfo()
|
2516 |
cluster.cluster_name = clustername |
2517 |
cluster.master_ip = ip |
2518 |
self.cfg.Update(cluster, feedback_fn)
|
2519 |
|
2520 |
# update the known hosts file
|
2521 |
ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
|
2522 |
node_list = self.cfg.GetNodeList()
|
2523 |
try:
|
2524 |
node_list.remove(master) |
2525 |
except ValueError: |
2526 |
pass
|
2527 |
result = self.rpc.call_upload_file(node_list,
|
2528 |
constants.SSH_KNOWN_HOSTS_FILE) |
2529 |
for to_node, to_result in result.iteritems(): |
2530 |
msg = to_result.fail_msg |
2531 |
if msg:
|
2532 |
msg = ("Copy of file %s to node %s failed: %s" %
|
2533 |
(constants.SSH_KNOWN_HOSTS_FILE, to_node, msg)) |
2534 |
self.proc.LogWarning(msg)
|
2535 |
|
2536 |
finally:
|
2537 |
result = self.rpc.call_node_start_master(master, False, False) |
2538 |
msg = result.fail_msg |
2539 |
if msg:
|
2540 |
self.LogWarning("Could not re-enable the master role on" |
2541 |
" the master, please restart manually: %s", msg)
|
2542 |
|
2543 |
|
2544 |
class LUSetClusterParams(LogicalUnit): |
2545 |
"""Change the parameters of the cluster.
|
2546 |
|
2547 |
"""
|
2548 |
HPATH = "cluster-modify"
|
2549 |
HTYPE = constants.HTYPE_CLUSTER |
2550 |
_OP_PARAMS = [ |
2551 |
("vg_name", None, _TMaybeString), |
2552 |
("enabled_hypervisors", None, |
2553 |
_TOr(_TAnd(_TListOf(_TElemOf(constants.HYPER_TYPES)), _TTrue), _TNone)), |
2554 |
("hvparams", None, _TOr(_TDictOf(_TNonEmptyString, _TDict), _TNone)), |
2555 |
("beparams", None, _TOr(_TDictOf(_TNonEmptyString, _TDict), _TNone)), |
2556 |
("os_hvp", None, _TOr(_TDictOf(_TNonEmptyString, _TDict), _TNone)), |
2557 |
("osparams", None, _TOr(_TDictOf(_TNonEmptyString, _TDict), _TNone)), |
2558 |
("candidate_pool_size", None, _TOr(_TStrictPositiveInt, _TNone)), |
2559 |
("uid_pool", None, _NoType), |
2560 |
("add_uids", None, _NoType), |
2561 |
("remove_uids", None, _NoType), |
2562 |
("maintain_node_health", None, _TMaybeBool), |
2563 |
("nicparams", None, _TOr(_TDict, _TNone)), |
2564 |
("drbd_helper", None, _TOr(_TString, _TNone)), |
2565 |
] |
2566 |
REQ_BGL = False
|
2567 |
|
2568 |
def CheckArguments(self): |
2569 |
"""Check parameters
|
2570 |
|
2571 |
"""
|
2572 |
if self.op.uid_pool: |
2573 |
uidpool.CheckUidPool(self.op.uid_pool)
|
2574 |
|
2575 |
if self.op.add_uids: |
2576 |
uidpool.CheckUidPool(self.op.add_uids)
|
2577 |
|
2578 |
if self.op.remove_uids: |
2579 |
uidpool.CheckUidPool(self.op.remove_uids)
|
2580 |
|
2581 |
def ExpandNames(self): |
2582 |
# FIXME: in the future maybe other cluster params won't require checking on
|
2583 |
# all nodes to be modified.
|
2584 |
self.needed_locks = {
|
2585 |
locking.LEVEL_NODE: locking.ALL_SET, |
2586 |
} |
2587 |
self.share_locks[locking.LEVEL_NODE] = 1 |
2588 |
|
2589 |
def BuildHooksEnv(self): |
2590 |
"""Build hooks env.
|
2591 |
|
2592 |
"""
|
2593 |
env = { |
2594 |
"OP_TARGET": self.cfg.GetClusterName(), |
2595 |
"NEW_VG_NAME": self.op.vg_name, |
2596 |
} |
2597 |
mn = self.cfg.GetMasterNode()
|
2598 |
return env, [mn], [mn]
|
2599 |
|
2600 |
def CheckPrereq(self): |
2601 |
"""Check prerequisites.
|
2602 |
|
2603 |
This checks whether the given params don't conflict and
|
2604 |
if the given volume group is valid.
|
2605 |
|
2606 |
"""
|
2607 |
if self.op.vg_name is not None and not self.op.vg_name: |
2608 |
if self.cfg.HasAnyDiskOfType(constants.LD_LV): |
2609 |
raise errors.OpPrereqError("Cannot disable lvm storage while lvm-based" |
2610 |
" instances exist", errors.ECODE_INVAL)
|
2611 |
|
2612 |
if self.op.drbd_helper is not None and not self.op.drbd_helper: |
2613 |
if self.cfg.HasAnyDiskOfType(constants.LD_DRBD8): |
2614 |
raise errors.OpPrereqError("Cannot disable drbd helper while" |
2615 |
" drbd-based instances exist",
|
2616 |
errors.ECODE_INVAL) |
2617 |
|
2618 |
node_list = self.acquired_locks[locking.LEVEL_NODE]
|
2619 |
|
2620 |
# if vg_name not None, checks given volume group on all nodes
|
2621 |
if self.op.vg_name: |
2622 |
vglist = self.rpc.call_vg_list(node_list)
|
2623 |
for node in node_list: |
2624 |
msg = vglist[node].fail_msg |
2625 |
if msg:
|
2626 |
# ignoring down node
|
2627 |
self.LogWarning("Error while gathering data on node %s" |
2628 |
" (ignoring node): %s", node, msg)
|
2629 |
continue
|
2630 |
vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload, |
2631 |
self.op.vg_name,
|
2632 |
constants.MIN_VG_SIZE) |
2633 |
if vgstatus:
|
2634 |
raise errors.OpPrereqError("Error on node '%s': %s" % |
2635 |
(node, vgstatus), errors.ECODE_ENVIRON) |
2636 |
|
2637 |
if self.op.drbd_helper: |
2638 |
# checks given drbd helper on all nodes
|
2639 |
helpers = self.rpc.call_drbd_helper(node_list)
|
2640 |
for node in node_list: |
2641 |
ninfo = self.cfg.GetNodeInfo(node)
|
2642 |
if ninfo.offline:
|
2643 |
self.LogInfo("Not checking drbd helper on offline node %s", node) |
2644 |
continue
|
2645 |
msg = helpers[node].fail_msg |
2646 |
if msg:
|
2647 |
raise errors.OpPrereqError("Error checking drbd helper on node" |
2648 |
" '%s': %s" % (node, msg),
|
2649 |
errors.ECODE_ENVIRON) |
2650 |
node_helper = helpers[node].payload |
2651 |
if node_helper != self.op.drbd_helper: |
2652 |
raise errors.OpPrereqError("Error on node '%s': drbd helper is %s" % |
2653 |
(node, node_helper), errors.ECODE_ENVIRON) |
2654 |
|
2655 |
self.cluster = cluster = self.cfg.GetClusterInfo() |
2656 |
# validate params changes
|
2657 |
if self.op.beparams: |
2658 |
utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
|
2659 |
self.new_beparams = cluster.SimpleFillBE(self.op.beparams) |
2660 |
|
2661 |
if self.op.nicparams: |
2662 |
utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
|
2663 |
self.new_nicparams = cluster.SimpleFillNIC(self.op.nicparams) |
2664 |
objects.NIC.CheckParameterSyntax(self.new_nicparams)
|
2665 |
nic_errors = [] |
2666 |
|
2667 |
# check all instances for consistency
|
2668 |
for instance in self.cfg.GetAllInstancesInfo().values(): |
2669 |
for nic_idx, nic in enumerate(instance.nics): |
2670 |
params_copy = copy.deepcopy(nic.nicparams) |
2671 |
params_filled = objects.FillDict(self.new_nicparams, params_copy)
|
2672 |
|
2673 |
# check parameter syntax
|
2674 |
try:
|
2675 |
objects.NIC.CheckParameterSyntax(params_filled) |
2676 |
except errors.ConfigurationError, err:
|
2677 |
nic_errors.append("Instance %s, nic/%d: %s" %
|
2678 |
(instance.name, nic_idx, err)) |
2679 |
|
2680 |
# if we're moving instances to routed, check that they have an ip
|
2681 |
target_mode = params_filled[constants.NIC_MODE] |
2682 |
if target_mode == constants.NIC_MODE_ROUTED and not nic.ip: |
2683 |
nic_errors.append("Instance %s, nic/%d: routed nick with no ip" %
|
2684 |
(instance.name, nic_idx)) |
2685 |
if nic_errors:
|
2686 |
raise errors.OpPrereqError("Cannot apply the change, errors:\n%s" % |
2687 |
"\n".join(nic_errors))
|
2688 |
|
2689 |
# hypervisor list/parameters
|
2690 |
self.new_hvparams = new_hvp = objects.FillDict(cluster.hvparams, {})
|
2691 |
if self.op.hvparams: |
2692 |
for hv_name, hv_dict in self.op.hvparams.items(): |
2693 |
if hv_name not in self.new_hvparams: |
2694 |
self.new_hvparams[hv_name] = hv_dict
|
2695 |
else:
|
2696 |
self.new_hvparams[hv_name].update(hv_dict)
|
2697 |
|
2698 |
# os hypervisor parameters
|
2699 |
self.new_os_hvp = objects.FillDict(cluster.os_hvp, {})
|
2700 |
if self.op.os_hvp: |
2701 |
for os_name, hvs in self.op.os_hvp.items(): |
2702 |
if os_name not in self.new_os_hvp: |
2703 |
self.new_os_hvp[os_name] = hvs
|
2704 |
else:
|
2705 |
for hv_name, hv_dict in hvs.items(): |
2706 |
if hv_name not in self.new_os_hvp[os_name]: |
2707 |
self.new_os_hvp[os_name][hv_name] = hv_dict
|
2708 |
else:
|
2709 |
self.new_os_hvp[os_name][hv_name].update(hv_dict)
|
2710 |
|
2711 |
# os parameters
|
2712 |
self.new_osp = objects.FillDict(cluster.osparams, {})
|
2713 |
if self.op.osparams: |
2714 |
for os_name, osp in self.op.osparams.items(): |
2715 |
if os_name not in self.new_osp: |
2716 |
self.new_osp[os_name] = {}
|
2717 |
|
2718 |
self.new_osp[os_name] = _GetUpdatedParams(self.new_osp[os_name], osp, |
2719 |
use_none=True)
|
2720 |
|
2721 |
if not self.new_osp[os_name]: |
2722 |
# we removed all parameters
|
2723 |
del self.new_osp[os_name] |
2724 |
else:
|
2725 |
# check the parameter validity (remote check)
|
2726 |
_CheckOSParams(self, False, [self.cfg.GetMasterNode()], |
2727 |
os_name, self.new_osp[os_name])
|
2728 |
|
2729 |
# changes to the hypervisor list
|
2730 |
if self.op.enabled_hypervisors is not None: |
2731 |
self.hv_list = self.op.enabled_hypervisors |
2732 |
for hv in self.hv_list: |
2733 |
# if the hypervisor doesn't already exist in the cluster
|
2734 |
# hvparams, we initialize it to empty, and then (in both
|
2735 |
# cases) we make sure to fill the defaults, as we might not
|
2736 |
# have a complete defaults list if the hypervisor wasn't
|
2737 |
# enabled before
|
2738 |
if hv not in new_hvp: |
2739 |
new_hvp[hv] = {} |
2740 |
new_hvp[hv] = objects.FillDict(constants.HVC_DEFAULTS[hv], new_hvp[hv]) |
2741 |
utils.ForceDictType(new_hvp[hv], constants.HVS_PARAMETER_TYPES) |
2742 |
else:
|
2743 |
self.hv_list = cluster.enabled_hypervisors
|
2744 |
|
2745 |
if self.op.hvparams or self.op.enabled_hypervisors is not None: |
2746 |
# either the enabled list has changed, or the parameters have, validate
|
2747 |
for hv_name, hv_params in self.new_hvparams.items(): |
2748 |
if ((self.op.hvparams and hv_name in self.op.hvparams) or |
2749 |
(self.op.enabled_hypervisors and |
2750 |
hv_name in self.op.enabled_hypervisors)): |
2751 |
# either this is a new hypervisor, or its parameters have changed
|
2752 |
hv_class = hypervisor.GetHypervisor(hv_name) |
2753 |
utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES) |
2754 |
hv_class.CheckParameterSyntax(hv_params) |
2755 |
_CheckHVParams(self, node_list, hv_name, hv_params)
|
2756 |
|
2757 |
if self.op.os_hvp: |
2758 |
# no need to check any newly-enabled hypervisors, since the
|
2759 |
# defaults have already been checked in the above code-block
|
2760 |
for os_name, os_hvp in self.new_os_hvp.items(): |
2761 |
for hv_name, hv_params in os_hvp.items(): |
2762 |
utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES) |
2763 |
# we need to fill in the new os_hvp on top of the actual hv_p
|
2764 |
cluster_defaults = self.new_hvparams.get(hv_name, {})
|
2765 |
new_osp = objects.FillDict(cluster_defaults, hv_params) |
2766 |
hv_class = hypervisor.GetHypervisor(hv_name) |
2767 |
hv_class.CheckParameterSyntax(new_osp) |
2768 |
_CheckHVParams(self, node_list, hv_name, new_osp)
|
2769 |
|
2770 |
|
2771 |
def Exec(self, feedback_fn): |
2772 |
"""Change the parameters of the cluster.
|
2773 |
|
2774 |
"""
|
2775 |
if self.op.vg_name is not None: |
2776 |
new_volume = self.op.vg_name
|
2777 |
if not new_volume: |
2778 |
new_volume = None
|
2779 |
if new_volume != self.cfg.GetVGName(): |
2780 |
self.cfg.SetVGName(new_volume)
|
2781 |
else:
|
2782 |
feedback_fn("Cluster LVM configuration already in desired"
|
2783 |
" state, not changing")
|
2784 |
if self.op.drbd_helper is not None: |
2785 |
new_helper = self.op.drbd_helper
|
2786 |
if not new_helper: |
2787 |
new_helper = None
|
2788 |
if new_helper != self.cfg.GetDRBDHelper(): |
2789 |
self.cfg.SetDRBDHelper(new_helper)
|
2790 |
else:
|
2791 |
feedback_fn("Cluster DRBD helper already in desired state,"
|
2792 |
" not changing")
|
2793 |
if self.op.hvparams: |
2794 |
self.cluster.hvparams = self.new_hvparams |
2795 |
if self.op.os_hvp: |
2796 |
self.cluster.os_hvp = self.new_os_hvp |
2797 |
if self.op.enabled_hypervisors is not None: |
2798 |
self.cluster.hvparams = self.new_hvparams |
2799 |
self.cluster.enabled_hypervisors = self.op.enabled_hypervisors |
2800 |
if self.op.beparams: |
2801 |
self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams |
2802 |
if self.op.nicparams: |
2803 |
self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams |
2804 |
if self.op.osparams: |
2805 |
self.cluster.osparams = self.new_osp |
2806 |
|
2807 |
if self.op.candidate_pool_size is not None: |
2808 |
self.cluster.candidate_pool_size = self.op.candidate_pool_size |
2809 |
# we need to update the pool size here, otherwise the save will fail
|
2810 |
_AdjustCandidatePool(self, [])
|
2811 |
|
2812 |
if self.op.maintain_node_health is not None: |
2813 |
self.cluster.maintain_node_health = self.op.maintain_node_health |
2814 |
|
2815 |
if self.op.add_uids is not None: |
2816 |
uidpool.AddToUidPool(self.cluster.uid_pool, self.op.add_uids) |
2817 |
|
2818 |
if self.op.remove_uids is not None: |
2819 |
uidpool.RemoveFromUidPool(self.cluster.uid_pool, self.op.remove_uids) |
2820 |
|
2821 |
if self.op.uid_pool is not None: |
2822 |
self.cluster.uid_pool = self.op.uid_pool |
2823 |
|
2824 |
self.cfg.Update(self.cluster, feedback_fn) |
2825 |
|
2826 |
|
2827 |
def _RedistributeAncillaryFiles(lu, additional_nodes=None): |
2828 |
"""Distribute additional files which are part of the cluster configuration.
|
2829 |
|
2830 |
ConfigWriter takes care of distributing the config and ssconf files, but
|
2831 |
there are more files which should be distributed to all nodes. This function
|
2832 |
makes sure those are copied.
|
2833 |
|
2834 |
@param lu: calling logical unit
|
2835 |
@param additional_nodes: list of nodes not in the config to distribute to
|
2836 |
|
2837 |
"""
|
2838 |
# 1. Gather target nodes
|
2839 |
myself = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode()) |
2840 |
dist_nodes = lu.cfg.GetOnlineNodeList() |
2841 |
if additional_nodes is not None: |
2842 |
dist_nodes.extend(additional_nodes) |
2843 |
if myself.name in dist_nodes: |
2844 |
dist_nodes.remove(myself.name) |
2845 |
|
2846 |
# 2. Gather files to distribute
|
2847 |
dist_files = set([constants.ETC_HOSTS,
|
2848 |
constants.SSH_KNOWN_HOSTS_FILE, |
2849 |
constants.RAPI_CERT_FILE, |
2850 |
constants.RAPI_USERS_FILE, |
2851 |
constants.CONFD_HMAC_KEY, |
2852 |
constants.CLUSTER_DOMAIN_SECRET_FILE, |
2853 |
]) |
2854 |
|
2855 |
enabled_hypervisors = lu.cfg.GetClusterInfo().enabled_hypervisors |
2856 |
for hv_name in enabled_hypervisors: |
2857 |
hv_class = hypervisor.GetHypervisor(hv_name) |
2858 |
dist_files.update(hv_class.GetAncillaryFiles()) |
2859 |
|
2860 |
# 3. Perform the files upload
|
2861 |
for fname in dist_files: |
2862 |
if os.path.exists(fname):
|
2863 |
result = lu.rpc.call_upload_file(dist_nodes, fname) |
2864 |
for to_node, to_result in result.items(): |
2865 |
msg = to_result.fail_msg |
2866 |
if msg:
|
2867 |
msg = ("Copy of file %s to node %s failed: %s" %
|
2868 |
(fname, to_node, msg)) |
2869 |
lu.proc.LogWarning(msg) |
2870 |
|
2871 |
|
2872 |
class LURedistributeConfig(NoHooksLU): |
2873 |
"""Force the redistribution of cluster configuration.
|
2874 |
|
2875 |
This is a very simple LU.
|
2876 |
|
2877 |
"""
|
2878 |
REQ_BGL = False
|
2879 |
|
2880 |
def ExpandNames(self): |
2881 |
self.needed_locks = {
|
2882 |
locking.LEVEL_NODE: locking.ALL_SET, |
2883 |
} |
2884 |
self.share_locks[locking.LEVEL_NODE] = 1 |
2885 |
|
2886 |
def Exec(self, feedback_fn): |
2887 |
"""Redistribute the configuration.
|
2888 |
|
2889 |
"""
|
2890 |
self.cfg.Update(self.cfg.GetClusterInfo(), feedback_fn) |
2891 |
_RedistributeAncillaryFiles(self)
|
2892 |
|
2893 |
|
2894 |
def _WaitForSync(lu, instance, disks=None, oneshot=False): |
2895 |
"""Sleep and poll for an instance's disk to sync.
|
2896 |
|
2897 |
"""
|
2898 |
if not instance.disks or disks is not None and not disks: |
2899 |
return True |
2900 |
|
2901 |
disks = _ExpandCheckDisks(instance, disks) |
2902 |
|
2903 |
if not oneshot: |
2904 |
lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
|
2905 |
|
2906 |
node = instance.primary_node |
2907 |
|
2908 |
for dev in disks: |
2909 |
lu.cfg.SetDiskID(dev, node) |
2910 |
|
2911 |
# TODO: Convert to utils.Retry
|
2912 |
|
2913 |
retries = 0
|
2914 |
degr_retries = 10 # in seconds, as we sleep 1 second each time |
2915 |
while True: |
2916 |
max_time = 0
|
2917 |
done = True
|
2918 |
cumul_degraded = False
|
2919 |
rstats = lu.rpc.call_blockdev_getmirrorstatus(node, disks) |
2920 |
msg = rstats.fail_msg |
2921 |
if msg:
|
2922 |
lu.LogWarning("Can't get any data from node %s: %s", node, msg)
|
2923 |
retries += 1
|
2924 |
if retries >= 10: |
2925 |
raise errors.RemoteError("Can't contact node %s for mirror data," |
2926 |
" aborting." % node)
|
2927 |
time.sleep(6)
|
2928 |
continue
|
2929 |
rstats = rstats.payload |
2930 |
retries = 0
|
2931 |
for i, mstat in enumerate(rstats): |
2932 |
if mstat is None: |
2933 |
lu.LogWarning("Can't compute data for node %s/%s",
|
2934 |
node, disks[i].iv_name) |
2935 |
continue
|
2936 |
|
2937 |
cumul_degraded = (cumul_degraded or
|
2938 |
(mstat.is_degraded and mstat.sync_percent is None)) |
2939 |
if mstat.sync_percent is not None: |
2940 |
done = False
|
2941 |
if mstat.estimated_time is not None: |
2942 |
rem_time = ("%s remaining (estimated)" %
|
2943 |
utils.FormatSeconds(mstat.estimated_time)) |
2944 |
max_time = mstat.estimated_time |
2945 |
else:
|
2946 |
rem_time = "no time estimate"
|
2947 |
lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
|
2948 |
(disks[i].iv_name, mstat.sync_percent, rem_time)) |
2949 |
|
2950 |
# if we're done but degraded, let's do a few small retries, to
|
2951 |
# make sure we see a stable and not transient situation; therefore
|
2952 |
# we force restart of the loop
|
2953 |
if (done or oneshot) and cumul_degraded and degr_retries > 0: |
2954 |
logging.info("Degraded disks found, %d retries left", degr_retries)
|
2955 |
degr_retries -= 1
|
2956 |
time.sleep(1)
|
2957 |
continue
|
2958 |
|
2959 |
if done or oneshot: |
2960 |
break
|
2961 |
|
2962 |
time.sleep(min(60, max_time)) |
2963 |
|
2964 |
if done:
|
2965 |
lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
|
2966 |
return not cumul_degraded |
2967 |
|
2968 |
|
2969 |
def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False): |
2970 |
"""Check that mirrors are not degraded.
|
2971 |
|
2972 |
The ldisk parameter, if True, will change the test from the
|
2973 |
is_degraded attribute (which represents overall non-ok status for
|
2974 |
the device(s)) to the ldisk (representing the local storage status).
|
2975 |
|
2976 |
"""
|
2977 |
lu.cfg.SetDiskID(dev, node) |
2978 |
|
2979 |
result = True
|
2980 |
|
2981 |
if on_primary or dev.AssembleOnSecondary(): |
2982 |
rstats = lu.rpc.call_blockdev_find(node, dev) |
2983 |
msg = rstats.fail_msg |
2984 |
if msg:
|
2985 |
lu.LogWarning("Can't find disk on node %s: %s", node, msg)
|
2986 |
result = False
|
2987 |
elif not rstats.payload: |
2988 |
lu.LogWarning("Can't find disk on node %s", node)
|
2989 |
result = False
|
2990 |
else:
|
2991 |
if ldisk:
|
2992 |
result = result and rstats.payload.ldisk_status == constants.LDS_OKAY
|
2993 |
else:
|
2994 |
result = result and not rstats.payload.is_degraded |
2995 |
|
2996 |
if dev.children:
|
2997 |
for child in dev.children: |
2998 |
result = result and _CheckDiskConsistency(lu, child, node, on_primary)
|
2999 |
|
3000 |
return result
|
3001 |
|
3002 |
|
3003 |
class LUDiagnoseOS(NoHooksLU): |
3004 |
"""Logical unit for OS diagnose/query.
|
3005 |
|
3006 |
"""
|
3007 |
_OP_PARAMS = [ |
3008 |
_POutputFields, |
3009 |
("names", _EmptyList, _TListOf(_TNonEmptyString)),
|
3010 |
] |
3011 |
REQ_BGL = False
|
3012 |
_FIELDS_STATIC = utils.FieldSet() |
3013 |
_FIELDS_DYNAMIC = utils.FieldSet("name", "valid", "node_status", "variants", |
3014 |
"parameters", "api_versions") |
3015 |
|
3016 |
def CheckArguments(self): |
3017 |
if self.op.names: |
3018 |
raise errors.OpPrereqError("Selective OS query not supported", |
3019 |
errors.ECODE_INVAL) |
3020 |
|
3021 |
_CheckOutputFields(static=self._FIELDS_STATIC,
|
3022 |
dynamic=self._FIELDS_DYNAMIC,
|
3023 |
selected=self.op.output_fields)
|
3024 |
|
3025 |
def ExpandNames(self): |
3026 |
# Lock all nodes, in shared mode
|
3027 |
# Temporary removal of locks, should be reverted later
|
3028 |
# TODO: reintroduce locks when they are lighter-weight
|
3029 |
self.needed_locks = {}
|
3030 |
#self.share_locks[locking.LEVEL_NODE] = 1
|
3031 |
#self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
|
3032 |
|
3033 |
@staticmethod
|
3034 |
def _DiagnoseByOS(rlist): |
3035 |
"""Remaps a per-node return list into an a per-os per-node dictionary
|
3036 |
|
3037 |
@param rlist: a map with node names as keys and OS objects as values
|
3038 |
|
3039 |
@rtype: dict
|
3040 |
@return: a dictionary with osnames as keys and as value another
|
3041 |
map, with nodes as keys and tuples of (path, status, diagnose,
|
3042 |
variants, parameters, api_versions) as values, eg::
|
3043 |
|
3044 |
{"debian-etch": {"node1": [(/usr/lib/..., True, "", [], []),
|
3045 |
(/srv/..., False, "invalid api")],
|
3046 |
"node2": [(/srv/..., True, "", [], [])]}
|
3047 |
}
|
3048 |
|
3049 |
"""
|
3050 |
all_os = {} |
3051 |
# we build here the list of nodes that didn't fail the RPC (at RPC
|
3052 |
# level), so that nodes with a non-responding node daemon don't
|
3053 |
# make all OSes invalid
|
3054 |
good_nodes = [node_name for node_name in rlist |
3055 |
if not rlist[node_name].fail_msg] |
3056 |
for node_name, nr in rlist.items(): |
3057 |
if nr.fail_msg or not nr.payload: |
3058 |
continue
|
3059 |
for (name, path, status, diagnose, variants,
|
3060 |
params, api_versions) in nr.payload:
|
3061 |
if name not in all_os: |
3062 |
# build a list of nodes for this os containing empty lists
|
3063 |
# for each node in node_list
|
3064 |
all_os[name] = {} |
3065 |
for nname in good_nodes: |
3066 |
all_os[name][nname] = [] |
3067 |
# convert params from [name, help] to (name, help)
|
3068 |
params = [tuple(v) for v in params] |
3069 |
all_os[name][node_name].append((path, status, diagnose, |
3070 |
variants, params, api_versions)) |
3071 |
return all_os
|
3072 |
|
3073 |
def Exec(self, feedback_fn): |
3074 |
"""Compute the list of OSes.
|
3075 |
|
3076 |
"""
|
3077 |
valid_nodes = [node for node in self.cfg.GetOnlineNodeList()] |
3078 |
node_data = self.rpc.call_os_diagnose(valid_nodes)
|
3079 |
pol = self._DiagnoseByOS(node_data)
|
3080 |
output = [] |
3081 |
|
3082 |
for os_name, os_data in pol.items(): |
3083 |
row = [] |
3084 |
valid = True
|
3085 |
(variants, params, api_versions) = null_state = (set(), set(), set()) |
3086 |
for idx, osl in enumerate(os_data.values()): |
3087 |
valid = bool(valid and osl and osl[0][1]) |
3088 |
if not valid: |
3089 |
(variants, params, api_versions) = null_state |
3090 |
break
|
3091 |
node_variants, node_params, node_api = osl[0][3:6] |
3092 |
if idx == 0: # first entry |
3093 |
variants = set(node_variants)
|
3094 |
params = set(node_params)
|
3095 |
api_versions = set(node_api)
|
3096 |
else: # keep consistency |
3097 |
variants.intersection_update(node_variants) |
3098 |
params.intersection_update(node_params) |
3099 |
api_versions.intersection_update(node_api) |
3100 |
|
3101 |
for field in self.op.output_fields: |
3102 |
if field == "name": |
3103 |
val = os_name |
3104 |
elif field == "valid": |
3105 |
val = valid |
3106 |
elif field == "node_status": |
3107 |
# this is just a copy of the dict
|
3108 |
val = {} |
3109 |
for node_name, nos_list in os_data.items(): |
3110 |
val[node_name] = nos_list |
3111 |
elif field == "variants": |
3112 |
val = list(variants)
|
3113 |
elif field == "parameters": |
3114 |
val = list(params)
|
3115 |
elif field == "api_versions": |
3116 |
val = list(api_versions)
|
3117 |
else:
|
3118 |
raise errors.ParameterError(field)
|
3119 |
row.append(val) |
3120 |
output.append(row) |
3121 |
|
3122 |
return output
|
3123 |
|
3124 |
|
3125 |
class LURemoveNode(LogicalUnit): |
3126 |
"""Logical unit for removing a node.
|
3127 |
|
3128 |
"""
|
3129 |
HPATH = "node-remove"
|
3130 |
HTYPE = constants.HTYPE_NODE |
3131 |
_OP_PARAMS = [ |
3132 |
_PNodeName, |
3133 |
] |
3134 |
|
3135 |
def BuildHooksEnv(self): |
3136 |
"""Build hooks env.
|
3137 |
|
3138 |
This doesn't run on the target node in the pre phase as a failed
|
3139 |
node would then be impossible to remove.
|
3140 |
|
3141 |
"""
|
3142 |
env = { |
3143 |
"OP_TARGET": self.op.node_name, |
3144 |
"NODE_NAME": self.op.node_name, |
3145 |
} |
3146 |
all_nodes = self.cfg.GetNodeList()
|
3147 |
try:
|
3148 |
all_nodes.remove(self.op.node_name)
|
3149 |
except ValueError: |
3150 |
logging.warning("Node %s which is about to be removed not found"
|
3151 |
" in the all nodes list", self.op.node_name) |
3152 |
return env, all_nodes, all_nodes
|
3153 |
|
3154 |
def CheckPrereq(self): |
3155 |
"""Check prerequisites.
|
3156 |
|
3157 |
This checks:
|
3158 |
- the node exists in the configuration
|
3159 |
- it does not have primary or secondary instances
|
3160 |
- it's not the master
|
3161 |
|
3162 |
Any errors are signaled by raising errors.OpPrereqError.
|
3163 |
|
3164 |
"""
|
3165 |
self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name) |
3166 |
node = self.cfg.GetNodeInfo(self.op.node_name) |
3167 |
assert node is not None |
3168 |
|
3169 |
instance_list = self.cfg.GetInstanceList()
|
3170 |
|
3171 |
masternode = self.cfg.GetMasterNode()
|
3172 |
if node.name == masternode:
|
3173 |
raise errors.OpPrereqError("Node is the master node," |
3174 |
" you need to failover first.",
|
3175 |
errors.ECODE_INVAL) |
3176 |
|
3177 |
for instance_name in instance_list: |
3178 |
instance = self.cfg.GetInstanceInfo(instance_name)
|
3179 |
if node.name in instance.all_nodes: |
3180 |
raise errors.OpPrereqError("Instance %s is still running on the node," |
3181 |
" please remove first." % instance_name,
|
3182 |
errors.ECODE_INVAL) |
3183 |
self.op.node_name = node.name
|
3184 |
self.node = node
|
3185 |
|
3186 |
def Exec(self, feedback_fn): |
3187 |
"""Removes the node from the cluster.
|
3188 |
|
3189 |
"""
|
3190 |
node = self.node
|
3191 |
logging.info("Stopping the node daemon and removing configs from node %s",
|
3192 |
node.name) |
3193 |
|
3194 |
modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
|
3195 |
|
3196 |
# Promote nodes to master candidate as needed
|
3197 |
_AdjustCandidatePool(self, exceptions=[node.name])
|
3198 |
self.context.RemoveNode(node.name)
|
3199 |
|
3200 |
# Run post hooks on the node before it's removed
|
3201 |
hm = self.proc.hmclass(self.rpc.call_hooks_runner, self) |
3202 |
try:
|
3203 |
hm.RunPhase(constants.HOOKS_PHASE_POST, [node.name]) |
3204 |
except:
|
3205 |
# pylint: disable-msg=W0702
|
3206 |
self.LogWarning("Errors occurred running hooks on %s" % node.name) |
3207 |
|
3208 |
result = self.rpc.call_node_leave_cluster(node.name, modify_ssh_setup)
|
3209 |
msg = result.fail_msg |
3210 |
if msg:
|
3211 |
self.LogWarning("Errors encountered on the remote node while leaving" |
3212 |
" the cluster: %s", msg)
|
3213 |
|
3214 |
# Remove node from our /etc/hosts
|
3215 |
if self.cfg.GetClusterInfo().modify_etc_hosts: |
3216 |
# FIXME: this should be done via an rpc call to node daemon
|
3217 |
utils.RemoveHostFromEtcHosts(node.name) |
3218 |
_RedistributeAncillaryFiles(self)
|
3219 |
|
3220 |
|
3221 |
class LUQueryNodes(NoHooksLU): |
3222 |
"""Logical unit for querying nodes.
|
3223 |
|
3224 |
"""
|
3225 |
# pylint: disable-msg=W0142
|
3226 |
_OP_PARAMS = [ |
3227 |
_POutputFields, |
3228 |
("names", _EmptyList, _TListOf(_TNonEmptyString)),
|
3229 |
("use_locking", False, _TBool), |
3230 |
] |
3231 |
REQ_BGL = False
|
3232 |
|
3233 |
_SIMPLE_FIELDS = ["name", "serial_no", "ctime", "mtime", "uuid", |
3234 |
"master_candidate", "offline", "drained"] |
3235 |
|
3236 |
_FIELDS_DYNAMIC = utils.FieldSet( |
3237 |
"dtotal", "dfree", |
3238 |
"mtotal", "mnode", "mfree", |
3239 |
"bootid",
|
3240 |
"ctotal", "cnodes", "csockets", |
3241 |
) |
3242 |
|
3243 |
_FIELDS_STATIC = utils.FieldSet(*[ |
3244 |
"pinst_cnt", "sinst_cnt", |
3245 |
"pinst_list", "sinst_list", |
3246 |
"pip", "sip", "tags", |
3247 |
"master",
|
3248 |
"role"] + _SIMPLE_FIELDS
|
3249 |
) |
3250 |
|
3251 |
def CheckArguments(self): |
3252 |
_CheckOutputFields(static=self._FIELDS_STATIC,
|
3253 |
dynamic=self._FIELDS_DYNAMIC,
|
3254 |
selected=self.op.output_fields)
|
3255 |
|
3256 |
def ExpandNames(self): |
3257 |
self.needed_locks = {}
|
3258 |
self.share_locks[locking.LEVEL_NODE] = 1 |
3259 |
|
3260 |
if self.op.names: |
3261 |
self.wanted = _GetWantedNodes(self, self.op.names) |
3262 |
else:
|
3263 |
self.wanted = locking.ALL_SET
|
3264 |
|
3265 |
self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields) |
3266 |
self.do_locking = self.do_node_query and self.op.use_locking |
3267 |
if self.do_locking: |
3268 |
# if we don't request only static fields, we need to lock the nodes
|
3269 |
self.needed_locks[locking.LEVEL_NODE] = self.wanted |
3270 |
|
3271 |
def Exec(self, feedback_fn): |
3272 |
"""Computes the list of nodes and their attributes.
|
3273 |
|
3274 |
"""
|
3275 |
all_info = self.cfg.GetAllNodesInfo()
|
3276 |
if self.do_locking: |
3277 |
nodenames = self.acquired_locks[locking.LEVEL_NODE]
|
3278 |
elif self.wanted != locking.ALL_SET: |
3279 |
nodenames = self.wanted
|
3280 |
missing = set(nodenames).difference(all_info.keys())
|
3281 |
if missing:
|
3282 |
raise errors.OpExecError(
|
3283 |
"Some nodes were removed before retrieving their data: %s" % missing)
|
3284 |
else:
|
3285 |
nodenames = all_info.keys() |
3286 |
|
3287 |
nodenames = utils.NiceSort(nodenames) |
3288 |
nodelist = [all_info[name] for name in nodenames] |
3289 |
|
3290 |
# begin data gathering
|
3291 |
|
3292 |
if self.do_node_query: |
3293 |
live_data = {} |
3294 |
node_data = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(), |
3295 |
self.cfg.GetHypervisorType())
|
3296 |
for name in nodenames: |
3297 |
nodeinfo = node_data[name] |
3298 |
if not nodeinfo.fail_msg and nodeinfo.payload: |
3299 |
nodeinfo = nodeinfo.payload |
3300 |
fn = utils.TryConvert |
3301 |
live_data[name] = { |
3302 |
"mtotal": fn(int, nodeinfo.get('memory_total', None)), |
3303 |
"mnode": fn(int, nodeinfo.get('memory_dom0', None)), |
3304 |
"mfree": fn(int, nodeinfo.get('memory_free', None)), |
3305 |
"dtotal": fn(int, nodeinfo.get('vg_size', None)), |
3306 |
"dfree": fn(int, nodeinfo.get('vg_free', None)), |
3307 |
"ctotal": fn(int, nodeinfo.get('cpu_total', None)), |
3308 |
"bootid": nodeinfo.get('bootid', None), |
3309 |
"cnodes": fn(int, nodeinfo.get('cpu_nodes', None)), |
3310 |
"csockets": fn(int, nodeinfo.get('cpu_sockets', None)), |
3311 |
} |
3312 |
else:
|
3313 |
live_data[name] = {} |
3314 |
else:
|
3315 |
live_data = dict.fromkeys(nodenames, {})
|
3316 |
|
3317 |
node_to_primary = dict([(name, set()) for name in nodenames]) |
3318 |
node_to_secondary = dict([(name, set()) for name in nodenames]) |
3319 |
|
3320 |
inst_fields = frozenset(("pinst_cnt", "pinst_list", |
3321 |
"sinst_cnt", "sinst_list")) |
3322 |
if inst_fields & frozenset(self.op.output_fields): |
3323 |
inst_data = self.cfg.GetAllInstancesInfo()
|
3324 |
|
3325 |
for inst in inst_data.values(): |
3326 |
if inst.primary_node in node_to_primary: |
3327 |
node_to_primary[inst.primary_node].add(inst.name) |
3328 |
for secnode in inst.secondary_nodes: |
3329 |
if secnode in node_to_secondary: |
3330 |
node_to_secondary[secnode].add(inst.name) |
3331 |
|
3332 |
master_node = self.cfg.GetMasterNode()
|
3333 |
|
3334 |
# end data gathering
|
3335 |
|
3336 |
output = [] |
3337 |
for node in nodelist: |
3338 |
node_output = [] |
3339 |
for field in self.op.output_fields: |
3340 |
if field in self._SIMPLE_FIELDS: |
3341 |
val = getattr(node, field)
|
3342 |
elif field == "pinst_list": |
3343 |
val = list(node_to_primary[node.name])
|
3344 |
elif field == "sinst_list": |
3345 |
val = list(node_to_secondary[node.name])
|
3346 |
elif field == "pinst_cnt": |
3347 |
val = len(node_to_primary[node.name])
|
3348 |
elif field == "sinst_cnt": |
3349 |
val = len(node_to_secondary[node.name])
|
3350 |
elif field == "pip": |
3351 |
val = node.primary_ip |
3352 |
elif field == "sip": |
3353 |
val = node.secondary_ip |
3354 |
elif field == "tags": |
3355 |
val = list(node.GetTags())
|
3356 |
elif field == "master": |
3357 |
val = node.name == master_node |
3358 |
elif self._FIELDS_DYNAMIC.Matches(field): |
3359 |
val = live_data[node.name].get(field, None)
|
3360 |
elif field == "role": |
3361 |
if node.name == master_node:
|
3362 |
val = "M"
|
3363 |
elif node.master_candidate:
|
3364 |
val = "C"
|
3365 |
elif node.drained:
|
3366 |
val = "D"
|
3367 |
elif node.offline:
|
3368 |
val = "O"
|
3369 |
else:
|
3370 |
val = "R"
|
3371 |
else:
|
3372 |
raise errors.ParameterError(field)
|
3373 |
node_output.append(val) |
3374 |
output.append(node_output) |
3375 |
|
3376 |
return output
|
3377 |
|
3378 |
|
3379 |
class LUQueryNodeVolumes(NoHooksLU): |
3380 |
"""Logical unit for getting volumes on node(s).
|
3381 |
|
3382 |
"""
|
3383 |
_OP_PARAMS = [ |
3384 |
("nodes", _EmptyList, _TListOf(_TNonEmptyString)),
|
3385 |
("output_fields", _NoDefault, _TListOf(_TNonEmptyString)),
|
3386 |
] |
3387 |
REQ_BGL = False
|
3388 |
_FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance") |
3389 |
_FIELDS_STATIC = utils.FieldSet("node")
|
3390 |
|
3391 |
def CheckArguments(self): |
3392 |
_CheckOutputFields(static=self._FIELDS_STATIC,
|
3393 |
dynamic=self._FIELDS_DYNAMIC,
|
3394 |
selected=self.op.output_fields)
|
3395 |
|
3396 |
def ExpandNames(self): |
3397 |
self.needed_locks = {}
|
3398 |
self.share_locks[locking.LEVEL_NODE] = 1 |
3399 |
if not self.op.nodes: |
3400 |
self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
|
3401 |
else:
|
3402 |
self.needed_locks[locking.LEVEL_NODE] = \
|
3403 |
_GetWantedNodes(self, self.op.nodes) |
3404 |
|
3405 |
def Exec(self, feedback_fn): |
3406 |
"""Computes the list of nodes and their attributes.
|
3407 |
|
3408 |
"""
|
3409 |
nodenames = self.acquired_locks[locking.LEVEL_NODE]
|
3410 |
volumes = self.rpc.call_node_volumes(nodenames)
|
3411 |
|
3412 |
ilist = [self.cfg.GetInstanceInfo(iname) for iname |
3413 |
in self.cfg.GetInstanceList()] |
3414 |
|
3415 |
lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist]) |
3416 |
|
3417 |
output = [] |
3418 |
for node in nodenames: |
3419 |
nresult = volumes[node] |
3420 |
if nresult.offline:
|
3421 |
continue
|
3422 |
msg = nresult.fail_msg |
3423 |
if msg:
|
3424 |
self.LogWarning("Can't compute volume data on node %s: %s", node, msg) |
3425 |
continue
|
3426 |
|
3427 |
node_vols = nresult.payload[:] |
3428 |
node_vols.sort(key=lambda vol: vol['dev']) |
3429 |
|
3430 |
for vol in node_vols: |
3431 |
node_output = [] |
3432 |
for field in self.op.output_fields: |
3433 |
if field == "node": |
3434 |
val = node |
3435 |
elif field == "phys": |
3436 |
val = vol['dev']
|
3437 |
elif field == "vg": |
3438 |
val = vol['vg']
|
3439 |
elif field == "name": |
3440 |
val = vol['name']
|
3441 |
elif field == "size": |
3442 |
val = int(float(vol['size'])) |
3443 |
elif field == "instance": |
3444 |
for inst in ilist: |
3445 |
if node not in lv_by_node[inst]: |
3446 |
continue
|
3447 |
if vol['name'] in lv_by_node[inst][node]: |
3448 |
val = inst.name |
3449 |
break
|
3450 |
else:
|
3451 |
val = '-'
|
3452 |
else:
|
3453 |
raise errors.ParameterError(field)
|
3454 |
node_output.append(str(val))
|
3455 |
|
3456 |
output.append(node_output) |
3457 |
|
3458 |
return output
|
3459 |
|
3460 |
|
3461 |
class LUQueryNodeStorage(NoHooksLU): |
3462 |
"""Logical unit for getting information on storage units on node(s).
|
3463 |
|
3464 |
"""
|
3465 |
_FIELDS_STATIC = utils.FieldSet(constants.SF_NODE) |
3466 |
_OP_PARAMS = [ |
3467 |
("nodes", _EmptyList, _TListOf(_TNonEmptyString)),
|
3468 |
("storage_type", _NoDefault, _CheckStorageType),
|
3469 |
("output_fields", _NoDefault, _TListOf(_TNonEmptyString)),
|
3470 |
("name", None, _TMaybeString), |
3471 |
] |
3472 |
REQ_BGL = False
|
3473 |
|
3474 |
def CheckArguments(self): |
3475 |
_CheckOutputFields(static=self._FIELDS_STATIC,
|
3476 |
dynamic=utils.FieldSet(*constants.VALID_STORAGE_FIELDS), |
3477 |
selected=self.op.output_fields)
|
3478 |
|
3479 |
def ExpandNames(self): |
3480 |
self.needed_locks = {}
|
3481 |
self.share_locks[locking.LEVEL_NODE] = 1 |
3482 |
|
3483 |
if self.op.nodes: |
3484 |
self.needed_locks[locking.LEVEL_NODE] = \
|
3485 |
_GetWantedNodes(self, self.op.nodes) |
3486 |
else:
|
3487 |
self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
|
3488 |
|
3489 |
def Exec(self, feedback_fn): |
3490 |
"""Computes the list of nodes and their attributes.
|
3491 |
|
3492 |
"""
|
3493 |
self.nodes = self.acquired_locks[locking.LEVEL_NODE] |
3494 |
|
3495 |
# Always get name to sort by
|
3496 |
if constants.SF_NAME in self.op.output_fields: |
3497 |
fields = self.op.output_fields[:]
|
3498 |
else:
|
3499 |
fields = [constants.SF_NAME] + self.op.output_fields
|
3500 |
|
3501 |
# Never ask for node or type as it's only known to the LU
|
3502 |
for extra in [constants.SF_NODE, constants.SF_TYPE]: |
3503 |
while extra in fields: |
3504 |
fields.remove(extra) |
3505 |
|
3506 |
field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)]) |
3507 |
name_idx = field_idx[constants.SF_NAME] |
3508 |
|
3509 |
st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type) |
3510 |
data = self.rpc.call_storage_list(self.nodes, |
3511 |
self.op.storage_type, st_args,
|
3512 |
self.op.name, fields)
|
3513 |
|
3514 |
result = [] |
3515 |
|
3516 |
for node in utils.NiceSort(self.nodes): |
3517 |
nresult = data[node] |
3518 |
if nresult.offline:
|
3519 |
continue
|
3520 |
|
3521 |
msg = nresult.fail_msg |
3522 |
if msg:
|
3523 |
self.LogWarning("Can't get storage data from node %s: %s", node, msg) |
3524 |
continue
|
3525 |
|
3526 |
rows = dict([(row[name_idx], row) for row in nresult.payload]) |
3527 |
|
3528 |
for name in utils.NiceSort(rows.keys()): |
3529 |
row = rows[name] |
3530 |
|
3531 |
out = [] |
3532 |
|
3533 |
for field in self.op.output_fields: |
3534 |
if field == constants.SF_NODE:
|
3535 |
val = node |
3536 |
elif field == constants.SF_TYPE:
|
3537 |
val = self.op.storage_type
|
3538 |
elif field in field_idx: |
3539 |
val = row[field_idx[field]] |
3540 |
else:
|
3541 |
raise errors.ParameterError(field)
|
3542 |
|
3543 |
out.append(val) |
3544 |
|
3545 |
result.append(out) |
3546 |
|
3547 |
return result
|
3548 |
|
3549 |
|
3550 |
class LUModifyNodeStorage(NoHooksLU): |
3551 |
"""Logical unit for modifying a storage volume on a node.
|
3552 |
|
3553 |
"""
|
3554 |
_OP_PARAMS = [ |
3555 |
_PNodeName, |
3556 |
("storage_type", _NoDefault, _CheckStorageType),
|
3557 |
("name", _NoDefault, _TNonEmptyString),
|
3558 |
("changes", _NoDefault, _TDict),
|
3559 |
] |
3560 |
REQ_BGL = False
|
3561 |
|
3562 |
def CheckArguments(self): |
3563 |
self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name) |
3564 |
|
3565 |
storage_type = self.op.storage_type
|
3566 |
|
3567 |
try:
|
3568 |
modifiable = constants.MODIFIABLE_STORAGE_FIELDS[storage_type] |
3569 |
except KeyError: |
3570 |
raise errors.OpPrereqError("Storage units of type '%s' can not be" |
3571 |
" modified" % storage_type,
|
3572 |
errors.ECODE_INVAL) |
3573 |
|
3574 |
diff = set(self.op.changes.keys()) - modifiable |
3575 |
if diff:
|
3576 |
raise errors.OpPrereqError("The following fields can not be modified for" |
3577 |
" storage units of type '%s': %r" %
|
3578 |
(storage_type, list(diff)),
|
3579 |
errors.ECODE_INVAL) |
3580 |
|
3581 |
def ExpandNames(self): |
3582 |
self.needed_locks = {
|
3583 |
locking.LEVEL_NODE: self.op.node_name,
|
3584 |
} |
3585 |
|
3586 |
def Exec(self, feedback_fn): |
3587 |
"""Computes the list of nodes and their attributes.
|
3588 |
|
3589 |
"""
|
3590 |
st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type) |
3591 |
result = self.rpc.call_storage_modify(self.op.node_name, |
3592 |
self.op.storage_type, st_args,
|
3593 |
self.op.name, self.op.changes) |
3594 |
result.Raise("Failed to modify storage unit '%s' on %s" %
|
3595 |
(self.op.name, self.op.node_name)) |
3596 |
|
3597 |
|
3598 |
class LUAddNode(LogicalUnit): |
3599 |
"""Logical unit for adding node to the cluster.
|
3600 |
|
3601 |
"""
|
3602 |
HPATH = "node-add"
|
3603 |
HTYPE = constants.HTYPE_NODE |
3604 |
_OP_PARAMS = [ |
3605 |
_PNodeName, |
3606 |
("primary_ip", None, _NoType), |
3607 |
("secondary_ip", None, _TMaybeString), |
3608 |
("readd", False, _TBool), |
3609 |
] |
3610 |
|
3611 |
def CheckArguments(self): |
3612 |
# validate/normalize the node name
|
3613 |
self.op.node_name = utils.HostInfo.NormalizeName(self.op.node_name) |
3614 |
|
3615 |
def BuildHooksEnv(self): |
3616 |
"""Build hooks env.
|
3617 |
|
3618 |
This will run on all nodes before, and on all nodes + the new node after.
|
3619 |
|
3620 |
"""
|
3621 |
env = { |
3622 |
"OP_TARGET": self.op.node_name, |
3623 |
"NODE_NAME": self.op.node_name, |
3624 |
"NODE_PIP": self.op.primary_ip, |
3625 |
"NODE_SIP": self.op.secondary_ip, |
3626 |
} |
3627 |
nodes_0 = self.cfg.GetNodeList()
|
3628 |
nodes_1 = nodes_0 + [self.op.node_name, ]
|
3629 |
return env, nodes_0, nodes_1
|
3630 |
|
3631 |
def CheckPrereq(self): |
3632 |
"""Check prerequisites.
|
3633 |
|
3634 |
This checks:
|
3635 |
- the new node is not already in the config
|
3636 |
- it is resolvable
|
3637 |
- its parameters (single/dual homed) matches the cluster
|
3638 |
|
3639 |
Any errors are signaled by raising errors.OpPrereqError.
|
3640 |
|
3641 |
"""
|
3642 |
node_name = self.op.node_name
|
3643 |
cfg = self.cfg
|
3644 |
|
3645 |
dns_data = utils.GetHostInfo(node_name) |
3646 |
|
3647 |
node = dns_data.name |
3648 |
primary_ip = self.op.primary_ip = dns_data.ip
|
3649 |
if self.op.secondary_ip is None: |
3650 |
self.op.secondary_ip = primary_ip
|
3651 |
if not utils.IsValidIP4(self.op.secondary_ip): |
3652 |
raise errors.OpPrereqError("Invalid secondary IP given", |
3653 |
errors.ECODE_INVAL) |
3654 |
secondary_ip = self.op.secondary_ip
|
3655 |
|
3656 |
node_list = cfg.GetNodeList() |
3657 |
if not self.op.readd and node in node_list: |
3658 |
raise errors.OpPrereqError("Node %s is already in the configuration" % |
3659 |
node, errors.ECODE_EXISTS) |
3660 |
elif self.op.readd and node not in node_list: |
3661 |
raise errors.OpPrereqError("Node %s is not in the configuration" % node, |
3662 |
errors.ECODE_NOENT) |
3663 |
|
3664 |
self.changed_primary_ip = False |
3665 |
|
3666 |
for existing_node_name in node_list: |
3667 |
existing_node = cfg.GetNodeInfo(existing_node_name) |
3668 |
|
3669 |
if self.op.readd and node == existing_node_name: |
3670 |
if existing_node.secondary_ip != secondary_ip:
|
3671 |
raise errors.OpPrereqError("Readded node doesn't have the same IP" |
3672 |
" address configuration as before",
|
3673 |
errors.ECODE_INVAL) |
3674 |
if existing_node.primary_ip != primary_ip:
|
3675 |
self.changed_primary_ip = True |
3676 |
|
3677 |
continue
|
3678 |
|
3679 |
if (existing_node.primary_ip == primary_ip or |
3680 |
existing_node.secondary_ip == primary_ip or
|
3681 |
existing_node.primary_ip == secondary_ip or
|
3682 |
existing_node.secondary_ip == secondary_ip): |
3683 |
raise errors.OpPrereqError("New node ip address(es) conflict with" |
3684 |
" existing node %s" % existing_node.name,
|
3685 |
errors.ECODE_NOTUNIQUE) |
3686 |
|
3687 |
# check that the type of the node (single versus dual homed) is the
|
3688 |
# same as for the master
|
3689 |
myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
|
3690 |
master_singlehomed = myself.secondary_ip == myself.primary_ip |
3691 |
newbie_singlehomed = secondary_ip == primary_ip |
3692 |
if master_singlehomed != newbie_singlehomed:
|
3693 |
if master_singlehomed:
|
3694 |
raise errors.OpPrereqError("The master has no private ip but the" |
3695 |
" new node has one",
|
3696 |
errors.ECODE_INVAL) |
3697 |
else:
|
3698 |
raise errors.OpPrereqError("The master has a private ip but the" |
3699 |
" new node doesn't have one",
|
3700 |
errors.ECODE_INVAL) |
3701 |
|
3702 |
# checks reachability
|
3703 |
if not utils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT): |
3704 |
raise errors.OpPrereqError("Node not reachable by ping", |
3705 |
errors.ECODE_ENVIRON) |
3706 |
|
3707 |
if not newbie_singlehomed: |
3708 |
# check reachability from my secondary ip to newbie's secondary ip
|
3709 |
if not utils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT, |
3710 |
source=myself.secondary_ip): |
3711 |
raise errors.OpPrereqError("Node secondary ip not reachable by TCP" |
3712 |
" based ping to noded port",
|
3713 |
errors.ECODE_ENVIRON) |
3714 |
|
3715 |
if self.op.readd: |
3716 |
exceptions = [node] |
3717 |
else:
|
3718 |
exceptions = [] |
3719 |
|
3720 |
self.master_candidate = _DecideSelfPromotion(self, exceptions=exceptions) |
3721 |
|
3722 |
if self.op.readd: |
3723 |
self.new_node = self.cfg.GetNodeInfo(node) |
3724 |
assert self.new_node is not None, "Can't retrieve locked node %s" % node |
3725 |
else:
|
3726 |
self.new_node = objects.Node(name=node,
|
3727 |
primary_ip=primary_ip, |
3728 |
secondary_ip=secondary_ip, |
3729 |
master_candidate=self.master_candidate,
|
3730 |
offline=False, drained=False) |
3731 |
|
3732 |
def Exec(self, feedback_fn): |
3733 |
"""Adds the new node to the cluster.
|
3734 |
|
3735 |
"""
|
3736 |
new_node = self.new_node
|
3737 |
node = new_node.name |
3738 |
|
3739 |
# for re-adds, reset the offline/drained/master-candidate flags;
|
3740 |
# we need to reset here, otherwise offline would prevent RPC calls
|
3741 |
# later in the procedure; this also means that if the re-add
|
3742 |
# fails, we are left with a non-offlined, broken node
|
3743 |
if self.op.readd: |
3744 |
new_node.drained = new_node.offline = False # pylint: disable-msg=W0201 |
3745 |
self.LogInfo("Readding a node, the offline/drained flags were reset") |
3746 |
# if we demote the node, we do cleanup later in the procedure
|
3747 |
new_node.master_candidate = self.master_candidate
|
3748 |
if self.changed_primary_ip: |
3749 |
new_node.primary_ip = self.op.primary_ip
|
3750 |
|
3751 |
# notify the user about any possible mc promotion
|
3752 |
if new_node.master_candidate:
|
3753 |
self.LogInfo("Node will be a master candidate") |
3754 |
|
3755 |
# check connectivity
|
3756 |
result = self.rpc.call_version([node])[node]
|
3757 |
result.Raise("Can't get version information from node %s" % node)
|
3758 |
if constants.PROTOCOL_VERSION == result.payload:
|
3759 |
logging.info("Communication to node %s fine, sw version %s match",
|
3760 |
node, result.payload) |
3761 |
else:
|
3762 |
raise errors.OpExecError("Version mismatch master version %s," |
3763 |
" node version %s" %
|
3764 |
(constants.PROTOCOL_VERSION, result.payload)) |
3765 |
|
3766 |
# setup ssh on node
|
3767 |
if self.cfg.GetClusterInfo().modify_ssh_setup: |
3768 |
logging.info("Copy ssh key to node %s", node)
|
3769 |
priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS) |
3770 |
keyarray = [] |
3771 |
keyfiles = [constants.SSH_HOST_DSA_PRIV, constants.SSH_HOST_DSA_PUB, |
3772 |
constants.SSH_HOST_RSA_PRIV, constants.SSH_HOST_RSA_PUB, |
3773 |
priv_key, pub_key] |
3774 |
|
3775 |
for i in keyfiles: |
3776 |
keyarray.append(utils.ReadFile(i)) |
3777 |
|
3778 |
result = self.rpc.call_node_add(node, keyarray[0], keyarray[1], |
3779 |
keyarray[2], keyarray[3], keyarray[4], |
3780 |
keyarray[5])
|
3781 |
result.Raise("Cannot transfer ssh keys to the new node")
|
3782 |
|
3783 |
# Add node to our /etc/hosts, and add key to known_hosts
|
3784 |
if self.cfg.GetClusterInfo().modify_etc_hosts: |
3785 |
# FIXME: this should be done via an rpc call to node daemon
|
3786 |
utils.AddHostToEtcHosts(new_node.name) |
3787 |
|
3788 |
if new_node.secondary_ip != new_node.primary_ip:
|
3789 |
result = self.rpc.call_node_has_ip_address(new_node.name,
|
3790 |
new_node.secondary_ip) |
3791 |
result.Raise("Failure checking secondary ip on node %s" % new_node.name,
|
3792 |
prereq=True, ecode=errors.ECODE_ENVIRON)
|
3793 |
if not result.payload: |
3794 |
raise errors.OpExecError("Node claims it doesn't have the secondary ip" |
3795 |
" you gave (%s). Please fix and re-run this"
|
3796 |
" command." % new_node.secondary_ip)
|
3797 |
|
3798 |
node_verify_list = [self.cfg.GetMasterNode()]
|
3799 |
node_verify_param = { |
3800 |
constants.NV_NODELIST: [node], |
3801 |
# TODO: do a node-net-test as well?
|
3802 |
} |
3803 |
|
3804 |
result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
|
3805 |
self.cfg.GetClusterName())
|
3806 |
for verifier in node_verify_list: |
3807 |
result[verifier].Raise("Cannot communicate with node %s" % verifier)
|
3808 |
nl_payload = result[verifier].payload[constants.NV_NODELIST] |
3809 |
if nl_payload:
|
3810 |
for failed in nl_payload: |
3811 |
feedback_fn("ssh/hostname verification failed"
|
3812 |
" (checking from %s): %s" %
|
3813 |
(verifier, nl_payload[failed])) |
3814 |
raise errors.OpExecError("ssh/hostname verification failed.") |
3815 |
|
3816 |
if self.op.readd: |
3817 |
_RedistributeAncillaryFiles(self)
|
3818 |
self.context.ReaddNode(new_node)
|
3819 |
# make sure we redistribute the config
|
3820 |
self.cfg.Update(new_node, feedback_fn)
|
3821 |
# and make sure the new node will not have old files around
|
3822 |
if not new_node.master_candidate: |
3823 |
result = self.rpc.call_node_demote_from_mc(new_node.name)
|
3824 |
msg = result.fail_msg |
3825 |
if msg:
|
3826 |
self.LogWarning("Node failed to demote itself from master" |
3827 |
" candidate status: %s" % msg)
|
3828 |
else:
|
3829 |
_RedistributeAncillaryFiles(self, additional_nodes=[node])
|
3830 |
self.context.AddNode(new_node, self.proc.GetECId()) |
3831 |
|
3832 |
|
3833 |
class LUSetNodeParams(LogicalUnit): |
3834 |
"""Modifies the parameters of a node.
|
3835 |
|
3836 |
"""
|
3837 |
HPATH = "node-modify"
|
3838 |
HTYPE = constants.HTYPE_NODE |
3839 |
_OP_PARAMS = [ |
3840 |
_PNodeName, |
3841 |
("master_candidate", None, _TMaybeBool), |
3842 |
("offline", None, _TMaybeBool), |
3843 |
("drained", None, _TMaybeBool), |
3844 |
("auto_promote", False, _TBool), |
3845 |
_PForce, |
3846 |
] |
3847 |
REQ_BGL = False
|
3848 |
|
3849 |
def CheckArguments(self): |
3850 |
self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name) |
3851 |
all_mods = [self.op.offline, self.op.master_candidate, self.op.drained] |
3852 |
if all_mods.count(None) == 3: |
3853 |
raise errors.OpPrereqError("Please pass at least one modification", |
3854 |
errors.ECODE_INVAL) |
3855 |
if all_mods.count(True) > 1: |
3856 |
raise errors.OpPrereqError("Can't set the node into more than one" |
3857 |
" state at the same time",
|
3858 |
errors.ECODE_INVAL) |
3859 |
|
3860 |
# Boolean value that tells us whether we're offlining or draining the node
|
3861 |
self.offline_or_drain = (self.op.offline == True or |
3862 |
self.op.drained == True) |
3863 |
self.deoffline_or_drain = (self.op.offline == False or |
3864 |
self.op.drained == False) |
3865 |
self.might_demote = (self.op.master_candidate == False or |
3866 |
self.offline_or_drain)
|
3867 |
|
3868 |
self.lock_all = self.op.auto_promote and self.might_demote |
3869 |
|
3870 |
|
3871 |
def ExpandNames(self): |
3872 |
if self.lock_all: |
3873 |
self.needed_locks = {locking.LEVEL_NODE: locking.ALL_SET}
|
3874 |
else:
|
3875 |
self.needed_locks = {locking.LEVEL_NODE: self.op.node_name} |
3876 |
|
3877 |
def BuildHooksEnv(self): |
3878 |
"""Build hooks env.
|
3879 |
|
3880 |
This runs on the master node.
|
3881 |
|
3882 |
"""
|
3883 |
env = { |
3884 |
"OP_TARGET": self.op.node_name, |
3885 |
"MASTER_CANDIDATE": str(self.op.master_candidate), |
3886 |
"OFFLINE": str(self.op.offline), |
3887 |
"DRAINED": str(self.op.drained), |
3888 |
} |
3889 |
nl = [self.cfg.GetMasterNode(),
|
3890 |
self.op.node_name]
|
3891 |
return env, nl, nl
|
3892 |
|
3893 |
def CheckPrereq(self): |
3894 |
"""Check prerequisites.
|
3895 |
|
3896 |
This only checks the instance list against the existing names.
|
3897 |
|
3898 |
"""
|
3899 |
node = self.node = self.cfg.GetNodeInfo(self.op.node_name) |
3900 |
|
3901 |
if (self.op.master_candidate is not None or |
3902 |
self.op.drained is not None or |
3903 |
self.op.offline is not None): |
3904 |
# we can't change the master's node flags
|
3905 |
if self.op.node_name == self.cfg.GetMasterNode(): |
3906 |
raise errors.OpPrereqError("The master role can be changed" |
3907 |
" only via masterfailover",
|
3908 |
errors.ECODE_INVAL) |
3909 |
|
3910 |
|
3911 |
if node.master_candidate and self.might_demote and not self.lock_all: |
3912 |
assert not self.op.auto_promote, "auto-promote set but lock_all not" |
3913 |
# check if after removing the current node, we're missing master
|
3914 |
# candidates
|
3915 |
(mc_remaining, mc_should, _) = \ |
3916 |
self.cfg.GetMasterCandidateStats(exceptions=[node.name])
|
3917 |
if mc_remaining < mc_should:
|
3918 |
raise errors.OpPrereqError("Not enough master candidates, please" |
3919 |
" pass auto_promote to allow promotion",
|
3920 |
errors.ECODE_INVAL) |
3921 |
|
3922 |
if (self.op.master_candidate == True and |
3923 |
((node.offline and not self.op.offline == False) or |
3924 |
(node.drained and not self.op.drained == False))): |
3925 |
raise errors.OpPrereqError("Node '%s' is offline or drained, can't set" |
3926 |
" to master_candidate" % node.name,
|
3927 |
errors.ECODE_INVAL) |
3928 |
|
3929 |
# If we're being deofflined/drained, we'll MC ourself if needed
|
3930 |
if (self.deoffline_or_drain and not self.offline_or_drain and not |
3931 |
self.op.master_candidate == True and not node.master_candidate): |
3932 |
self.op.master_candidate = _DecideSelfPromotion(self) |
3933 |
if self.op.master_candidate: |
3934 |
self.LogInfo("Autopromoting node to master candidate") |
3935 |
|
3936 |
return
|
3937 |
|
3938 |
def Exec(self, feedback_fn): |
3939 |
"""Modifies a node.
|
3940 |
|
3941 |
"""
|
3942 |
node = self.node
|
3943 |
|
3944 |
result = [] |
3945 |
changed_mc = False
|
3946 |
|
3947 |
if self.op.offline is not None: |
3948 |
node.offline = self.op.offline
|
3949 |
result.append(("offline", str(self.op.offline))) |
3950 |
if self.op.offline == True: |
3951 |
if node.master_candidate:
|
3952 |
node.master_candidate = False
|
3953 |
changed_mc = True
|
3954 |
result.append(("master_candidate", "auto-demotion due to offline")) |
3955 |
if node.drained:
|
3956 |
node.drained = False
|
3957 |
result.append(("drained", "clear drained status due to offline")) |
3958 |
|
3959 |
if self.op.master_candidate is not None: |
3960 |
node.master_candidate = self.op.master_candidate
|
3961 |
changed_mc = True
|
3962 |
result.append(("master_candidate", str(self.op.master_candidate))) |
3963 |
if self.op.master_candidate == False: |
3964 |
rrc = self.rpc.call_node_demote_from_mc(node.name)
|
3965 |
msg = rrc.fail_msg |
3966 |
if msg:
|
3967 |
self.LogWarning("Node failed to demote itself: %s" % msg) |
3968 |
|
3969 |
if self.op.drained is not None: |
3970 |
node.drained = self.op.drained
|
3971 |
result.append(("drained", str(self.op.drained))) |
3972 |
if self.op.drained == True: |
3973 |
if node.master_candidate:
|
3974 |
node.master_candidate = False
|
3975 |
changed_mc = True
|
3976 |
result.append(("master_candidate", "auto-demotion due to drain")) |
3977 |
rrc = self.rpc.call_node_demote_from_mc(node.name)
|
3978 |
msg = rrc.fail_msg |
3979 |
if msg:
|
3980 |
self.LogWarning("Node failed to demote itself: %s" % msg) |
3981 |
if node.offline:
|
3982 |
node.offline = False
|
3983 |
result.append(("offline", "clear offline status due to drain")) |
3984 |
|
3985 |
# we locked all nodes, we adjust the CP before updating this node
|
3986 |
if self.lock_all: |
3987 |
_AdjustCandidatePool(self, [node.name])
|
3988 |
|
3989 |
# this will trigger configuration file update, if needed
|
3990 |
self.cfg.Update(node, feedback_fn)
|
3991 |
|
3992 |
# this will trigger job queue propagation or cleanup
|
3993 |
if changed_mc:
|
3994 |
self.context.ReaddNode(node)
|
3995 |
|
3996 |
return result
|
3997 |
|
3998 |
|
3999 |
class LUPowercycleNode(NoHooksLU): |
4000 |
"""Powercycles a node.
|
4001 |
|
4002 |
"""
|
4003 |
_OP_PARAMS = [ |
4004 |
_PNodeName, |
4005 |
_PForce, |
4006 |
] |
4007 |
REQ_BGL = False
|
4008 |
|
4009 |
def CheckArguments(self): |
4010 |
self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name) |
4011 |
if self.op.node_name == self.cfg.GetMasterNode() and not self.op.force: |
4012 |
raise errors.OpPrereqError("The node is the master and the force" |
4013 |
" parameter was not set",
|
4014 |
errors.ECODE_INVAL) |
4015 |
|
4016 |
def ExpandNames(self): |
4017 |
"""Locking for PowercycleNode.
|
4018 |
|
4019 |
This is a last-resort option and shouldn't block on other
|
4020 |
jobs. Therefore, we grab no locks.
|
4021 |
|
4022 |
"""
|
4023 |
self.needed_locks = {}
|
4024 |
|
4025 |
def Exec(self, feedback_fn): |
4026 |
"""Reboots a node.
|
4027 |
|
4028 |
"""
|
4029 |
result = self.rpc.call_node_powercycle(self.op.node_name, |
4030 |
self.cfg.GetHypervisorType())
|
4031 |
result.Raise("Failed to schedule the reboot")
|
4032 |
return result.payload
|
4033 |
|
4034 |
|
4035 |
class LUQueryClusterInfo(NoHooksLU): |
4036 |
"""Query cluster configuration.
|
4037 |
|
4038 |
"""
|
4039 |
REQ_BGL = False
|
4040 |
|
4041 |
def ExpandNames(self): |
4042 |
self.needed_locks = {}
|
4043 |
|
4044 |
def Exec(self, feedback_fn): |
4045 |
"""Return cluster config.
|
4046 |
|
4047 |
"""
|
4048 |
cluster = self.cfg.GetClusterInfo()
|
4049 |
os_hvp = {} |
4050 |
|
4051 |
# Filter just for enabled hypervisors
|
4052 |
for os_name, hv_dict in cluster.os_hvp.items(): |
4053 |
os_hvp[os_name] = {} |
4054 |
for hv_name, hv_params in hv_dict.items(): |
4055 |
if hv_name in cluster.enabled_hypervisors: |
4056 |
os_hvp[os_name][hv_name] = hv_params |
4057 |
|
4058 |
result = { |
4059 |
"software_version": constants.RELEASE_VERSION,
|
4060 |
"protocol_version": constants.PROTOCOL_VERSION,
|
4061 |
"config_version": constants.CONFIG_VERSION,
|
4062 |
"os_api_version": max(constants.OS_API_VERSIONS), |
4063 |
"export_version": constants.EXPORT_VERSION,
|
4064 |
"architecture": (platform.architecture()[0], platform.machine()), |
4065 |
"name": cluster.cluster_name,
|
4066 |
"master": cluster.master_node,
|
4067 |
"default_hypervisor": cluster.enabled_hypervisors[0], |
4068 |
"enabled_hypervisors": cluster.enabled_hypervisors,
|
4069 |
"hvparams": dict([(hypervisor_name, cluster.hvparams[hypervisor_name]) |
4070 |
for hypervisor_name in cluster.enabled_hypervisors]), |
4071 |
"os_hvp": os_hvp,
|
4072 |
"beparams": cluster.beparams,
|
4073 |
"osparams": cluster.osparams,
|
4074 |
"nicparams": cluster.nicparams,
|
4075 |
"candidate_pool_size": cluster.candidate_pool_size,
|
4076 |
"master_netdev": cluster.master_netdev,
|
4077 |
"volume_group_name": cluster.volume_group_name,
|
4078 |
"file_storage_dir": cluster.file_storage_dir,
|
4079 |
"maintain_node_health": cluster.maintain_node_health,
|
4080 |
"ctime": cluster.ctime,
|
4081 |
"mtime": cluster.mtime,
|
4082 |
"uuid": cluster.uuid,
|
4083 |
"tags": list(cluster.GetTags()), |
4084 |
"uid_pool": cluster.uid_pool,
|
4085 |
} |
4086 |
|
4087 |
return result
|
4088 |
|
4089 |
|
4090 |
class LUQueryConfigValues(NoHooksLU): |
4091 |
"""Return configuration values.
|
4092 |
|
4093 |
"""
|
4094 |
_OP_PARAMS = [_POutputFields] |
4095 |
REQ_BGL = False
|
4096 |
_FIELDS_DYNAMIC = utils.FieldSet() |
4097 |
_FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag", |
4098 |
"watcher_pause")
|
4099 |
|
4100 |
def CheckArguments(self): |
4101 |
_CheckOutputFields(static=self._FIELDS_STATIC,
|
4102 |
dynamic=self._FIELDS_DYNAMIC,
|
4103 |
selected=self.op.output_fields)
|
4104 |
|
4105 |
def ExpandNames(self): |
4106 |
self.needed_locks = {}
|
4107 |
|
4108 |
def Exec(self, feedback_fn): |
4109 |
"""Dump a representation of the cluster config to the standard output.
|
4110 |
|
4111 |
"""
|
4112 |
values = [] |
4113 |
for field in self.op.output_fields: |
4114 |
if field == "cluster_name": |
4115 |
entry = self.cfg.GetClusterName()
|
4116 |
elif field == "master_node": |
4117 |
entry = self.cfg.GetMasterNode()
|
4118 |
elif field == "drain_flag": |
4119 |
entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE) |
4120 |
elif field == "watcher_pause": |
4121 |
entry = utils.ReadWatcherPauseFile(constants.WATCHER_PAUSEFILE) |
4122 |
else:
|
4123 |
raise errors.ParameterError(field)
|
4124 |
values.append(entry) |
4125 |
return values
|
4126 |
|
4127 |
|
4128 |
class LUActivateInstanceDisks(NoHooksLU): |
4129 |
"""Bring up an instance's disks.
|
4130 |
|
4131 |
"""
|
4132 |
_OP_PARAMS = [ |
4133 |
_PInstanceName, |
4134 |
("ignore_size", False, _TBool), |
4135 |
] |
4136 |
REQ_BGL = False
|
4137 |
|
4138 |
def ExpandNames(self): |
4139 |
self._ExpandAndLockInstance()
|
4140 |
self.needed_locks[locking.LEVEL_NODE] = []
|
4141 |
self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
|
4142 |
|
4143 |
def DeclareLocks(self, level): |
4144 |
if level == locking.LEVEL_NODE:
|
4145 |
self._LockInstancesNodes()
|
4146 |
|
4147 |
def CheckPrereq(self): |
4148 |
"""Check prerequisites.
|
4149 |
|
4150 |
This checks that the instance is in the cluster.
|
4151 |
|
4152 |
"""
|
4153 |
self.instance = self.cfg.GetInstanceInfo(self.op.instance_name) |
4154 |
assert self.instance is not None, \ |
4155 |
"Cannot retrieve locked instance %s" % self.op.instance_name |
4156 |
_CheckNodeOnline(self, self.instance.primary_node) |
4157 |
|
4158 |
def Exec(self, feedback_fn): |
4159 |
"""Activate the disks.
|
4160 |
|
4161 |
"""
|
4162 |
disks_ok, disks_info = \ |
4163 |
_AssembleInstanceDisks(self, self.instance, |
4164 |
ignore_size=self.op.ignore_size)
|
4165 |
if not disks_ok: |
4166 |
raise errors.OpExecError("Cannot activate block devices") |
4167 |
|
4168 |
return disks_info
|
4169 |
|
4170 |
|
4171 |
def _AssembleInstanceDisks(lu, instance, disks=None, ignore_secondaries=False, |
4172 |
ignore_size=False):
|
4173 |
"""Prepare the block devices for an instance.
|
4174 |
|
4175 |
This sets up the block devices on all nodes.
|
4176 |
|
4177 |
@type lu: L{LogicalUnit}
|
4178 |
@param lu: the logical unit on whose behalf we execute
|
4179 |
@type instance: L{objects.Instance}
|
4180 |
@param instance: the instance for whose disks we assemble
|
4181 |
@type disks: list of L{objects.Disk} or None
|
4182 |
@param disks: which disks to assemble (or all, if None)
|
4183 |
@type ignore_secondaries: boolean
|
4184 |
@param ignore_secondaries: if true, errors on secondary nodes
|
4185 |
won't result in an error return from the function
|
4186 |
@type ignore_size: boolean
|
4187 |
@param ignore_size: if true, the current known size of the disk
|
4188 |
will not be used during the disk activation, useful for cases
|
4189 |
when the size is wrong
|
4190 |
@return: False if the operation failed, otherwise a list of
|
4191 |
(host, instance_visible_name, node_visible_name)
|
4192 |
with the mapping from node devices to instance devices
|
4193 |
|
4194 |
"""
|
4195 |
device_info = [] |
4196 |
disks_ok = True
|
4197 |
iname = instance.name |
4198 |
disks = _ExpandCheckDisks(instance, disks) |
4199 |
|
4200 |
# With the two passes mechanism we try to reduce the window of
|
4201 |
# opportunity for the race condition of switching DRBD to primary
|
4202 |
# before handshaking occured, but we do not eliminate it
|
4203 |
|
4204 |
# The proper fix would be to wait (with some limits) until the
|
4205 |
# connection has been made and drbd transitions from WFConnection
|
4206 |
# into any other network-connected state (Connected, SyncTarget,
|
4207 |
# SyncSource, etc.)
|
4208 |
|
4209 |
# 1st pass, assemble on all nodes in secondary mode
|
4210 |
for inst_disk in disks: |
4211 |
for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node): |
4212 |
if ignore_size:
|
4213 |
node_disk = node_disk.Copy() |
4214 |
node_disk.UnsetSize() |
4215 |
lu.cfg.SetDiskID(node_disk, node) |
4216 |
result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False)
|
4217 |
msg = result.fail_msg |
4218 |
if msg:
|
4219 |
lu.proc.LogWarning("Could not prepare block device %s on node %s"
|
4220 |
" (is_primary=False, pass=1): %s",
|
4221 |
inst_disk.iv_name, node, msg) |
4222 |
if not ignore_secondaries: |
4223 |
disks_ok = False
|
4224 |
|
4225 |
# FIXME: race condition on drbd migration to primary
|
4226 |
|
4227 |
# 2nd pass, do only the primary node
|
4228 |
for inst_disk in disks: |
4229 |
dev_path = None
|
4230 |
|
4231 |
for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node): |
4232 |
if node != instance.primary_node:
|
4233 |
continue
|
4234 |
if ignore_size:
|
4235 |
node_disk = node_disk.Copy() |
4236 |
node_disk.UnsetSize() |
4237 |
lu.cfg.SetDiskID(node_disk, node) |
4238 |
result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True)
|
4239 |
msg = result.fail_msg |
4240 |
if msg:
|
4241 |
lu.proc.LogWarning("Could not prepare block device %s on node %s"
|
4242 |
" (is_primary=True, pass=2): %s",
|
4243 |
inst_disk.iv_name, node, msg) |
4244 |
disks_ok = False
|
4245 |
else:
|
4246 |
dev_path = result.payload |
4247 |
|
4248 |
device_info.append((instance.primary_node, inst_disk.iv_name, dev_path)) |
4249 |
|
4250 |
# leave the disks configured for the primary node
|
4251 |
# this is a workaround that would be fixed better by
|
4252 |
# improving the logical/physical id handling
|
4253 |
for disk in disks: |
4254 |
lu.cfg.SetDiskID(disk, instance.primary_node) |
4255 |
|
4256 |
return disks_ok, device_info
|
4257 |
|
4258 |
|
4259 |
def _StartInstanceDisks(lu, instance, force): |
4260 |
"""Start the disks of an instance.
|
4261 |
|
4262 |
"""
|
4263 |
disks_ok, _ = _AssembleInstanceDisks(lu, instance, |
4264 |
ignore_secondaries=force) |
4265 |
if not disks_ok: |
4266 |
_ShutdownInstanceDisks(lu, instance) |
4267 |
if force is not None and not force: |
4268 |
lu.proc.LogWarning("", hint="If the message above refers to a" |
4269 |
" secondary node,"
|
4270 |
" you can retry the operation using '--force'.")
|
4271 |
raise errors.OpExecError("Disk consistency error") |
4272 |
|
4273 |
|
4274 |
class LUDeactivateInstanceDisks(NoHooksLU): |
4275 |
"""Shutdown an instance's disks.
|
4276 |
|
4277 |
"""
|
4278 |
_OP_PARAMS = [ |
4279 |
_PInstanceName, |
4280 |
] |
4281 |
REQ_BGL = False
|
4282 |
|
4283 |
def ExpandNames(self): |
4284 |
self._ExpandAndLockInstance()
|
4285 |
self.needed_locks[locking.LEVEL_NODE] = []
|
4286 |
self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
|
4287 |
|
4288 |
def DeclareLocks(self, level): |
4289 |
if level == locking.LEVEL_NODE:
|
4290 |
self._LockInstancesNodes()
|
4291 |
|
4292 |
def CheckPrereq(self): |
4293 |
"""Check prerequisites.
|
4294 |
|
4295 |
This checks that the instance is in the cluster.
|
4296 |
|
4297 |
"""
|
4298 |
self.instance = self.cfg.GetInstanceInfo(self.op.instance_name) |
4299 |
assert self.instance is not None, \ |
4300 |
"Cannot retrieve locked instance %s" % self.op.instance_name |
4301 |
|
4302 |
def Exec(self, feedback_fn): |
4303 |
"""Deactivate the disks
|
4304 |
|
4305 |
"""
|
4306 |
instance = self.instance
|
4307 |
_SafeShutdownInstanceDisks(self, instance)
|
4308 |
|
4309 |
|
4310 |
def _SafeShutdownInstanceDisks(lu, instance, disks=None): |
4311 |
"""Shutdown block devices of an instance.
|
4312 |
|
4313 |
This function checks if an instance is running, before calling
|
4314 |
_ShutdownInstanceDisks.
|
4315 |
|
4316 |
"""
|
4317 |
_CheckInstanceDown(lu, instance, "cannot shutdown disks")
|
4318 |
_ShutdownInstanceDisks(lu, instance, disks=disks) |
4319 |
|
4320 |
|
4321 |
def _ExpandCheckDisks(instance, disks): |
4322 |
"""Return the instance disks selected by the disks list
|
4323 |
|
4324 |
@type disks: list of L{objects.Disk} or None
|
4325 |
@param disks: selected disks
|
4326 |
@rtype: list of L{objects.Disk}
|
4327 |
@return: selected instance disks to act on
|
4328 |
|
4329 |
"""
|
4330 |
if disks is None: |
4331 |
return instance.disks
|
4332 |
else:
|
4333 |
if not set(disks).issubset(instance.disks): |
4334 |
raise errors.ProgrammerError("Can only act on disks belonging to the" |
4335 |
" target instance")
|
4336 |
return disks
|
4337 |
|
4338 |
|
4339 |
def _ShutdownInstanceDisks(lu, instance, disks=None, ignore_primary=False): |
4340 |
"""Shutdown block devices of an instance.
|
4341 |
|
4342 |
This does the shutdown on all nodes of the instance.
|
4343 |
|
4344 |
If the ignore_primary is false, errors on the primary node are
|
4345 |
ignored.
|
4346 |
|
4347 |
"""
|
4348 |
all_result = True
|
4349 |
disks = _ExpandCheckDisks(instance, disks) |
4350 |
|
4351 |
for disk in disks: |
4352 |
for node, top_disk in disk.ComputeNodeTree(instance.primary_node): |
4353 |
lu.cfg.SetDiskID(top_disk, node) |
4354 |
result = lu.rpc.call_blockdev_shutdown(node, top_disk) |
4355 |
msg = result.fail_msg |
4356 |
if msg:
|
4357 |
lu.LogWarning("Could not shutdown block device %s on node %s: %s",
|
4358 |
disk.iv_name, node, msg) |
4359 |
if not ignore_primary or node != instance.primary_node: |
4360 |
all_result = False
|
4361 |
return all_result
|
4362 |
|
4363 |
|
4364 |
def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name): |
4365 |
"""Checks if a node has enough free memory.
|
4366 |
|
4367 |
This function check if a given node has the needed amount of free
|
4368 |
memory. In case the node has less memory or we cannot get the
|
4369 |
information from the node, this function raise an OpPrereqError
|
4370 |
exception.
|
4371 |
|
4372 |
@type lu: C{LogicalUnit}
|
4373 |
@param lu: a logical unit from which we get configuration data
|
4374 |
@type node: C{str}
|
4375 |
@param node: the node to check
|
4376 |
@type reason: C{str}
|
4377 |
@param reason: string to use in the error message
|
4378 |
@type requested: C{int}
|
4379 |
@param requested: the amount of memory in MiB to check for
|
4380 |
@type hypervisor_name: C{str}
|
4381 |
@param hypervisor_name: the hypervisor to ask for memory stats
|
4382 |
@raise errors.OpPrereqError: if the node doesn't have enough memory, or
|
4383 |
we cannot check the node
|
4384 |
|
4385 |
"""
|
4386 |
nodeinfo = lu.rpc.call_node_info([node], lu.cfg.GetVGName(), hypervisor_name) |
4387 |
nodeinfo[node].Raise("Can't get data from node %s" % node,
|
4388 |
prereq=True, ecode=errors.ECODE_ENVIRON)
|
4389 |
free_mem = nodeinfo[node].payload.get('memory_free', None) |
4390 |
if not isinstance(free_mem, int): |
4391 |
raise errors.OpPrereqError("Can't compute free memory on node %s, result" |
4392 |
" was '%s'" % (node, free_mem),
|
4393 |
errors.ECODE_ENVIRON) |
4394 |
if requested > free_mem:
|
4395 |
raise errors.OpPrereqError("Not enough memory on node %s for %s:" |
4396 |
" needed %s MiB, available %s MiB" %
|
4397 |
(node, reason, requested, free_mem), |
4398 |
errors.ECODE_NORES) |
4399 |
|
4400 |
|
4401 |
def _CheckNodesFreeDisk(lu, nodenames, requested): |
4402 |
"""Checks if nodes have enough free disk space in the default VG.
|
4403 |
|
4404 |
This function check if all given nodes have the needed amount of
|
4405 |
free disk. In case any node has less disk or we cannot get the
|
4406 |
information from the node, this function raise an OpPrereqError
|
4407 |
exception.
|
4408 |
|
4409 |
@type lu: C{LogicalUnit}
|
4410 |
@param lu: a logical unit from which we get configuration data
|
4411 |
@type nodenames: C{list}
|
4412 |
@param nodenames: the list of node names to check
|
4413 |
@type requested: C{int}
|
4414 |
@param requested: the amount of disk in MiB to check for
|
4415 |
@raise errors.OpPrereqError: if the node doesn't have enough disk, or
|
4416 |
we cannot check the node
|
4417 |
|
4418 |
"""
|
4419 |
nodeinfo = lu.rpc.call_node_info(nodenames, lu.cfg.GetVGName(), |
4420 |
lu.cfg.GetHypervisorType()) |
4421 |
for node in nodenames: |
4422 |
info = nodeinfo[node] |
4423 |
info.Raise("Cannot get current information from node %s" % node,
|
4424 |
prereq=True, ecode=errors.ECODE_ENVIRON)
|
4425 |
vg_free = info.payload.get("vg_free", None) |
4426 |
if not isinstance(vg_free, int): |
4427 |
raise errors.OpPrereqError("Can't compute free disk space on node %s," |
4428 |
" result was '%s'" % (node, vg_free),
|
4429 |
errors.ECODE_ENVIRON) |
4430 |
if requested > vg_free:
|
4431 |
raise errors.OpPrereqError("Not enough disk space on target node %s:" |
4432 |
" required %d MiB, available %d MiB" %
|
4433 |
(node, requested, vg_free), |
4434 |
errors.ECODE_NORES) |
4435 |
|
4436 |
|
4437 |
class LUStartupInstance(LogicalUnit): |
4438 |
"""Starts an instance.
|
4439 |
|
4440 |
"""
|
4441 |
HPATH = "instance-start"
|
4442 |
HTYPE = constants.HTYPE_INSTANCE |
4443 |
_OP_PARAMS = [ |
4444 |
_PInstanceName, |
4445 |
_PForce, |
4446 |
("hvparams", _EmptyDict, _TDict),
|
4447 |
("beparams", _EmptyDict, _TDict),
|
4448 |
] |
4449 |
REQ_BGL = False
|
4450 |
|
4451 |
def CheckArguments(self): |
4452 |
# extra beparams
|
4453 |
if self.op.beparams: |
4454 |
# fill the beparams dict
|
4455 |
utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
|
4456 |
|
4457 |
def ExpandNames(self): |
4458 |
self._ExpandAndLockInstance()
|
4459 |
|
4460 |
def BuildHooksEnv(self): |
4461 |
"""Build hooks env.
|
4462 |
|
4463 |
This runs on master, primary and secondary nodes of the instance.
|
4464 |
|
4465 |
"""
|
4466 |
env = { |
4467 |
"FORCE": self.op.force, |
4468 |
} |
4469 |
env.update(_BuildInstanceHookEnvByObject(self, self.instance)) |
4470 |
nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes) |
4471 |
return env, nl, nl
|
4472 |
|
4473 |
def CheckPrereq(self): |
4474 |
"""Check prerequisites.
|
4475 |
|
4476 |
This checks that the instance is in the cluster.
|
4477 |
|
4478 |
"""
|
4479 |
self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name) |
4480 |
assert self.instance is not None, \ |
4481 |
"Cannot retrieve locked instance %s" % self.op.instance_name |
4482 |
|
4483 |
# extra hvparams
|
4484 |
if self.op.hvparams: |
4485 |
# check hypervisor parameter syntax (locally)
|
4486 |
cluster = self.cfg.GetClusterInfo()
|
4487 |
utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
|
4488 |
filled_hvp = cluster.FillHV(instance) |
4489 |
filled_hvp.update(self.op.hvparams)
|
4490 |
hv_type = hypervisor.GetHypervisor(instance.hypervisor) |
4491 |
hv_type.CheckParameterSyntax(filled_hvp) |
4492 |
_CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp)
|
4493 |
|
4494 |
_CheckNodeOnline(self, instance.primary_node)
|
4495 |
|
4496 |
bep = self.cfg.GetClusterInfo().FillBE(instance)
|
4497 |
# check bridges existence
|
4498 |
_CheckInstanceBridgesExist(self, instance)
|
4499 |
|
4500 |
remote_info = self.rpc.call_instance_info(instance.primary_node,
|
4501 |
instance.name, |
4502 |
instance.hypervisor) |
4503 |
remote_info.Raise("Error checking node %s" % instance.primary_node,
|
4504 |
prereq=True, ecode=errors.ECODE_ENVIRON)
|
4505 |
if not remote_info.payload: # not running already |
4506 |
_CheckNodeFreeMemory(self, instance.primary_node,
|
4507 |
"starting instance %s" % instance.name,
|
4508 |
bep[constants.BE_MEMORY], instance.hypervisor) |
4509 |
|
4510 |
def Exec(self, feedback_fn): |
4511 |
"""Start the instance.
|
4512 |
|
4513 |
"""
|
4514 |
instance = self.instance
|
4515 |
force = self.op.force
|
4516 |
|
4517 |
self.cfg.MarkInstanceUp(instance.name)
|
4518 |
|
4519 |
node_current = instance.primary_node |
4520 |
|
4521 |
_StartInstanceDisks(self, instance, force)
|
4522 |
|
4523 |
result = self.rpc.call_instance_start(node_current, instance,
|
4524 |
self.op.hvparams, self.op.beparams) |
4525 |
msg = result.fail_msg |
4526 |
if msg:
|
4527 |
_ShutdownInstanceDisks(self, instance)
|
4528 |
raise errors.OpExecError("Could not start instance: %s" % msg) |
4529 |
|
4530 |
|
4531 |
class LURebootInstance(LogicalUnit): |
4532 |
"""Reboot an instance.
|
4533 |
|
4534 |
"""
|
4535 |
HPATH = "instance-reboot"
|
4536 |
HTYPE = constants.HTYPE_INSTANCE |
4537 |
_OP_PARAMS = [ |
4538 |
_PInstanceName, |
4539 |
("ignore_secondaries", False, _TBool), |
4540 |
("reboot_type", _NoDefault, _TElemOf(constants.REBOOT_TYPES)),
|
4541 |
_PShutdownTimeout, |
4542 |
] |
4543 |
REQ_BGL = False
|
4544 |
|
4545 |
def ExpandNames(self): |
4546 |
self._ExpandAndLockInstance()
|
4547 |
|
4548 |
def BuildHooksEnv(self): |
4549 |
"""Build hooks env.
|
4550 |
|
4551 |
This runs on master, primary and secondary nodes of the instance.
|
4552 |
|
4553 |
"""
|
4554 |
env = { |
4555 |
"IGNORE_SECONDARIES": self.op.ignore_secondaries, |
4556 |
"REBOOT_TYPE": self.op.reboot_type, |
4557 |
"SHUTDOWN_TIMEOUT": self.op.shutdown_timeout, |
4558 |
} |
4559 |
env.update(_BuildInstanceHookEnvByObject(self, self.instance)) |
4560 |
nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes) |
4561 |
return env, nl, nl
|
4562 |
|
4563 |
def CheckPrereq(self): |
4564 |
"""Check prerequisites.
|
4565 |
|
4566 |
This checks that the instance is in the cluster.
|
4567 |
|
4568 |
"""
|
4569 |
self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name) |
4570 |
assert self.instance is not None, \ |
4571 |
"Cannot retrieve locked instance %s" % self.op.instance_name |
4572 |
|
4573 |
_CheckNodeOnline(self, instance.primary_node)
|
4574 |
|
4575 |
# check bridges existence
|
4576 |
_CheckInstanceBridgesExist(self, instance)
|
4577 |
|
4578 |
def Exec(self, feedback_fn): |
4579 |
"""Reboot the instance.
|
4580 |
|
4581 |
"""
|
4582 |
instance = self.instance
|
4583 |
ignore_secondaries = self.op.ignore_secondaries
|
4584 |
reboot_type = self.op.reboot_type
|
4585 |
|
4586 |
node_current = instance.primary_node |
4587 |
|
4588 |
if reboot_type in [constants.INSTANCE_REBOOT_SOFT, |
4589 |
constants.INSTANCE_REBOOT_HARD]: |
4590 |
for disk in instance.disks: |
4591 |
self.cfg.SetDiskID(disk, node_current)
|
4592 |
result = self.rpc.call_instance_reboot(node_current, instance,
|
4593 |
reboot_type, |
4594 |
self.op.shutdown_timeout)
|
4595 |
result.Raise("Could not reboot instance")
|
4596 |
else:
|
4597 |
result = self.rpc.call_instance_shutdown(node_current, instance,
|
4598 |
self.op.shutdown_timeout)
|
4599 |
result.Raise("Could not shutdown instance for full reboot")
|
4600 |
_ShutdownInstanceDisks(self, instance)
|
4601 |
_StartInstanceDisks(self, instance, ignore_secondaries)
|
4602 |
result = self.rpc.call_instance_start(node_current, instance, None, None) |
4603 |
msg = result.fail_msg |
4604 |
if msg:
|
4605 |
_ShutdownInstanceDisks(self, instance)
|
4606 |
raise errors.OpExecError("Could not start instance for" |
4607 |
" full reboot: %s" % msg)
|
4608 |
|
4609 |
self.cfg.MarkInstanceUp(instance.name)
|
4610 |
|
4611 |
|
4612 |
class LUShutdownInstance(LogicalUnit): |
4613 |
"""Shutdown an instance.
|
4614 |
|
4615 |
"""
|
4616 |
HPATH = "instance-stop"
|
4617 |
HTYPE = constants.HTYPE_INSTANCE |
4618 |
_OP_PARAMS = [ |
4619 |
_PInstanceName, |
4620 |
("timeout", constants.DEFAULT_SHUTDOWN_TIMEOUT, _TPositiveInt),
|
4621 |
] |
4622 |
REQ_BGL = False
|
4623 |
|
4624 |
def ExpandNames(self): |
4625 |
self._ExpandAndLockInstance()
|
4626 |
|
4627 |
def BuildHooksEnv(self): |
4628 |
"""Build hooks env.
|
4629 |
|
4630 |
This runs on master, primary and secondary nodes of the instance.
|
4631 |
|
4632 |
"""
|
4633 |
env = _BuildInstanceHookEnvByObject(self, self.instance) |
4634 |
env["TIMEOUT"] = self.op.timeout |
4635 |
nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes) |
4636 |
return env, nl, nl
|
4637 |
|
4638 |
def CheckPrereq(self): |
4639 |
"""Check prerequisites.
|
4640 |
|
4641 |
This checks that the instance is in the cluster.
|
4642 |
|
4643 |
"""
|
4644 |
self.instance = self.cfg.GetInstanceInfo(self.op.instance_name) |
4645 |
assert self.instance is not None, \ |
4646 |
"Cannot retrieve locked instance %s" % self.op.instance_name |
4647 |
_CheckNodeOnline(self, self.instance.primary_node) |
4648 |
|
4649 |
def Exec(self, feedback_fn): |
4650 |
"""Shutdown the instance.
|
4651 |
|
4652 |
"""
|
4653 |
instance = self.instance
|
4654 |
node_current = instance.primary_node |
4655 |
timeout = self.op.timeout
|
4656 |
self.cfg.MarkInstanceDown(instance.name)
|
4657 |
result = self.rpc.call_instance_shutdown(node_current, instance, timeout)
|
4658 |
msg = result.fail_msg |
4659 |
if msg:
|
4660 |
self.proc.LogWarning("Could not shutdown instance: %s" % msg) |
4661 |
|
4662 |
_ShutdownInstanceDisks(self, instance)
|
4663 |
|
4664 |
|
4665 |
class LUReinstallInstance(LogicalUnit): |
4666 |
"""Reinstall an instance.
|
4667 |
|
4668 |
"""
|
4669 |
HPATH = "instance-reinstall"
|
4670 |
HTYPE = constants.HTYPE_INSTANCE |
4671 |
_OP_PARAMS = [ |
4672 |
_PInstanceName, |
4673 |
("os_type", None, _TMaybeString), |
4674 |
("force_variant", False, _TBool), |
4675 |
] |
4676 |
REQ_BGL = False
|
4677 |
|
4678 |
def ExpandNames(self): |
4679 |
self._ExpandAndLockInstance()
|
4680 |
|
4681 |
def BuildHooksEnv(self): |
4682 |
"""Build hooks env.
|
4683 |
|
4684 |
This runs on master, primary and secondary nodes of the instance.
|
4685 |
|
4686 |
"""
|
4687 |
env = _BuildInstanceHookEnvByObject(self, self.instance) |
4688 |
nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes) |
4689 |
return env, nl, nl
|
4690 |
|
4691 |
def CheckPrereq(self): |
4692 |
"""Check prerequisites.
|
4693 |
|
4694 |
This checks that the instance is in the cluster and is not running.
|
4695 |
|
4696 |
"""
|
4697 |
instance = self.cfg.GetInstanceInfo(self.op.instance_name) |
4698 |
assert instance is not None, \ |
4699 |
"Cannot retrieve locked instance %s" % self.op.instance_name |
4700 |
_CheckNodeOnline(self, instance.primary_node)
|
4701 |
|
4702 |
if instance.disk_template == constants.DT_DISKLESS:
|
4703 |
raise errors.OpPrereqError("Instance '%s' has no disks" % |
4704 |
self.op.instance_name,
|
4705 |
errors.ECODE_INVAL) |
4706 |
_CheckInstanceDown(self, instance, "cannot reinstall") |
4707 |
|
4708 |
if self.op.os_type is not None: |
4709 |
# OS verification
|
4710 |
pnode = _ExpandNodeName(self.cfg, instance.primary_node)
|
4711 |
_CheckNodeHasOS(self, pnode, self.op.os_type, self.op.force_variant) |
4712 |
|
4713 |
self.instance = instance
|
4714 |
|
4715 |
def Exec(self, feedback_fn): |
4716 |
"""Reinstall the instance.
|
4717 |
|
4718 |
"""
|
4719 |
inst = self.instance
|
4720 |
|
4721 |
if self.op.os_type is not None: |
4722 |
feedback_fn("Changing OS to '%s'..." % self.op.os_type) |
4723 |
inst.os = self.op.os_type
|
4724 |
self.cfg.Update(inst, feedback_fn)
|
4725 |
|
4726 |
_StartInstanceDisks(self, inst, None) |
4727 |
try:
|
4728 |
feedback_fn("Running the instance OS create scripts...")
|
4729 |
# FIXME: pass debug option from opcode to backend
|
4730 |
result = self.rpc.call_instance_os_add(inst.primary_node, inst, True, |
4731 |
self.op.debug_level)
|
4732 |
result.Raise("Could not install OS for instance %s on node %s" %
|
4733 |
(inst.name, inst.primary_node)) |
4734 |
finally:
|
4735 |
_ShutdownInstanceDisks(self, inst)
|
4736 |
|
4737 |
|
4738 |
class LURecreateInstanceDisks(LogicalUnit): |
4739 |
"""Recreate an instance's missing disks.
|
4740 |
|
4741 |
"""
|
4742 |
HPATH = "instance-recreate-disks"
|
4743 |
HTYPE = constants.HTYPE_INSTANCE |
4744 |
_OP_PARAMS = [ |
4745 |
_PInstanceName, |
4746 |
("disks", _EmptyList, _TListOf(_TPositiveInt)),
|
4747 |
] |
4748 |
REQ_BGL = False
|
4749 |
|
4750 |
def ExpandNames(self): |
4751 |
self._ExpandAndLockInstance()
|
4752 |
|
4753 |
def BuildHooksEnv(self): |
4754 |
"""Build hooks env.
|
4755 |
|
4756 |
This runs on master, primary and secondary nodes of the instance.
|
4757 |
|
4758 |
"""
|
4759 |
env = _BuildInstanceHookEnvByObject(self, self.instance) |
4760 |
nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes) |
4761 |
return env, nl, nl
|
4762 |
|
4763 |
def CheckPrereq(self): |
4764 |
"""Check prerequisites.
|
4765 |
|
4766 |
This checks that the instance is in the cluster and is not running.
|
4767 |
|
4768 |
"""
|
4769 |
instance = self.cfg.GetInstanceInfo(self.op.instance_name) |
4770 |
assert instance is not None, \ |
4771 |
"Cannot retrieve locked instance %s" % self.op.instance_name |
4772 |
_CheckNodeOnline(self, instance.primary_node)
|
4773 |
|
4774 |
if instance.disk_template == constants.DT_DISKLESS:
|
4775 |
raise errors.OpPrereqError("Instance '%s' has no disks" % |
4776 |
self.op.instance_name, errors.ECODE_INVAL)
|
4777 |
_CheckInstanceDown(self, instance, "cannot recreate disks") |
4778 |
|
4779 |
if not self.op.disks: |
4780 |
self.op.disks = range(len(instance.disks)) |
4781 |
else:
|
4782 |
for idx in self.op.disks: |
4783 |
if idx >= len(instance.disks): |
4784 |
raise errors.OpPrereqError("Invalid disk index passed '%s'" % idx, |
4785 |
errors.ECODE_INVAL) |
4786 |
|
4787 |
self.instance = instance
|
4788 |
|
4789 |
def Exec(self, feedback_fn): |
4790 |
"""Recreate the disks.
|
4791 |
|
4792 |
"""
|
4793 |
to_skip = [] |
4794 |
for idx, _ in enumerate(self.instance.disks): |
4795 |
if idx not in self.op.disks: # disk idx has not been passed in |
4796 |
to_skip.append(idx) |
4797 |
continue
|
4798 |
|
4799 |
_CreateDisks(self, self.instance, to_skip=to_skip) |
4800 |
|
4801 |
|
4802 |
class LURenameInstance(LogicalUnit): |
4803 |
"""Rename an instance.
|
4804 |
|
4805 |
"""
|
4806 |
HPATH = "instance-rename"
|
4807 |
HTYPE = constants.HTYPE_INSTANCE |
4808 |
_OP_PARAMS = [ |
4809 |
_PInstanceName, |
4810 |
("new_name", _NoDefault, _TNonEmptyString),
|
4811 |
("ignore_ip", False, _TBool), |
4812 |
("check_name", True, _TBool), |
4813 |
] |
4814 |
|
4815 |
def BuildHooksEnv(self): |
4816 |
"""Build hooks env.
|
4817 |
|
4818 |
This runs on master, primary and secondary nodes of the instance.
|
4819 |
|
4820 |
"""
|
4821 |
env = _BuildInstanceHookEnvByObject(self, self.instance) |
4822 |
env["INSTANCE_NEW_NAME"] = self.op.new_name |
4823 |
nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes) |
4824 |
return env, nl, nl
|
4825 |
|
4826 |
def CheckPrereq(self): |
4827 |
"""Check prerequisites.
|
4828 |
|
4829 |
This checks that the instance is in the cluster and is not running.
|
4830 |
|
4831 |
"""
|
4832 |
self.op.instance_name = _ExpandInstanceName(self.cfg, |
4833 |
self.op.instance_name)
|
4834 |
instance = self.cfg.GetInstanceInfo(self.op.instance_name) |
4835 |
assert instance is not None |
4836 |
_CheckNodeOnline(self, instance.primary_node)
|
4837 |
_CheckInstanceDown(self, instance, "cannot rename") |
4838 |
self.instance = instance
|
4839 |
|
4840 |
# new name verification
|
4841 |
if self.op.check_name: |
4842 |
name_info = utils.GetHostInfo(self.op.new_name)
|
4843 |
self.op.new_name = name_info.name
|
4844 |
|
4845 |
new_name = self.op.new_name
|
4846 |
|
4847 |
instance_list = self.cfg.GetInstanceList()
|
4848 |
if new_name in instance_list: |
4849 |
raise errors.OpPrereqError("Instance '%s' is already in the cluster" % |
4850 |
new_name, errors.ECODE_EXISTS) |
4851 |
|
4852 |
if not self.op.ignore_ip: |
4853 |
if utils.TcpPing(name_info.ip, constants.DEFAULT_NODED_PORT):
|
4854 |
raise errors.OpPrereqError("IP %s of instance %s already in use" % |
4855 |
(name_info.ip, new_name), |
4856 |
errors.ECODE_NOTUNIQUE) |
4857 |
|
4858 |
def Exec(self, feedback_fn): |
4859 |
"""Reinstall the instance.
|
4860 |
|
4861 |
"""
|
4862 |
inst = self.instance
|
4863 |
old_name = inst.name |
4864 |
|
4865 |
if inst.disk_template == constants.DT_FILE:
|
4866 |
old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1]) |
4867 |
|
4868 |
self.cfg.RenameInstance(inst.name, self.op.new_name) |
4869 |
# Change the instance lock. This is definitely safe while we hold the BGL
|
4870 |
self.context.glm.remove(locking.LEVEL_INSTANCE, old_name)
|
4871 |
self.context.glm.add(locking.LEVEL_INSTANCE, self.op.new_name) |
4872 |
|
4873 |
# re-read the instance from the configuration after rename
|
4874 |
inst = self.cfg.GetInstanceInfo(self.op.new_name) |
4875 |
|
4876 |
if inst.disk_template == constants.DT_FILE:
|
4877 |
new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1]) |
4878 |
result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
|
4879 |
old_file_storage_dir, |
4880 |
new_file_storage_dir) |
4881 |
result.Raise("Could not rename on node %s directory '%s' to '%s'"
|
4882 |
" (but the instance has been renamed in Ganeti)" %
|
4883 |
(inst.primary_node, old_file_storage_dir, |
4884 |
new_file_storage_dir)) |
4885 |
|
4886 |
_StartInstanceDisks(self, inst, None) |
4887 |
try:
|
4888 |
result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
|
4889 |
old_name, self.op.debug_level)
|
4890 |
msg = result.fail_msg |
4891 |
if msg:
|
4892 |
msg = ("Could not run OS rename script for instance %s on node %s"
|
4893 |
" (but the instance has been renamed in Ganeti): %s" %
|
4894 |
(inst.name, inst.primary_node, msg)) |
4895 |
self.proc.LogWarning(msg)
|
4896 |
finally:
|
4897 |
_ShutdownInstanceDisks(self, inst)
|
4898 |
|
4899 |
|
4900 |
class LURemoveInstance(LogicalUnit): |
4901 |
"""Remove an instance.
|
4902 |
|
4903 |
"""
|
4904 |
HPATH = "instance-remove"
|
4905 |
HTYPE = constants.HTYPE_INSTANCE |
4906 |
_OP_PARAMS = [ |
4907 |
_PInstanceName, |
4908 |
("ignore_failures", False, _TBool), |
4909 |
_PShutdownTimeout, |
4910 |
] |
4911 |
REQ_BGL = False
|
4912 |
|
4913 |
def ExpandNames(self): |
4914 |
self._ExpandAndLockInstance()
|
4915 |
self.needed_locks[locking.LEVEL_NODE] = []
|
4916 |
self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
|
4917 |
|
4918 |
def DeclareLocks(self, level): |
4919 |
if level == locking.LEVEL_NODE:
|
4920 |
self._LockInstancesNodes()
|
4921 |
|
4922 |
def BuildHooksEnv(self): |
4923 |
"""Build hooks env.
|
4924 |
|
4925 |
This runs on master, primary and secondary nodes of the instance.
|
4926 |
|
4927 |
"""
|
4928 |
env = _BuildInstanceHookEnvByObject(self, self.instance) |
4929 |
env["SHUTDOWN_TIMEOUT"] = self.op.shutdown_timeout |
4930 |
nl = [self.cfg.GetMasterNode()]
|
4931 |
nl_post = list(self.instance.all_nodes) + nl |
4932 |
return env, nl, nl_post
|
4933 |
|
4934 |
def CheckPrereq(self): |
4935 |
"""Check prerequisites.
|
4936 |
|
4937 |
This checks that the instance is in the cluster.
|
4938 |
|
4939 |
"""
|
4940 |
self.instance = self.cfg.GetInstanceInfo(self.op.instance_name) |
4941 |
assert self.instance is not None, \ |
4942 |
"Cannot retrieve locked instance %s" % self.op.instance_name |
4943 |
|
4944 |
def Exec(self, feedback_fn): |
4945 |
"""Remove the instance.
|
4946 |
|
4947 |
"""
|
4948 |
instance = self.instance
|
4949 |
logging.info("Shutting down instance %s on node %s",
|
4950 |
instance.name, instance.primary_node) |
4951 |
|
4952 |
result = self.rpc.call_instance_shutdown(instance.primary_node, instance,
|
4953 |
self.op.shutdown_timeout)
|
4954 |
msg = result.fail_msg |
4955 |
if msg:
|
4956 |
if self.op.ignore_failures: |
4957 |
feedback_fn("Warning: can't shutdown instance: %s" % msg)
|
4958 |
else:
|
4959 |
raise errors.OpExecError("Could not shutdown instance %s on" |
4960 |
" node %s: %s" %
|
4961 |
(instance.name, instance.primary_node, msg)) |
4962 |
|
4963 |
_RemoveInstance(self, feedback_fn, instance, self.op.ignore_failures) |
4964 |
|
4965 |
|
4966 |
def _RemoveInstance(lu, feedback_fn, instance, ignore_failures): |
4967 |
"""Utility function to remove an instance.
|
4968 |
|
4969 |
"""
|
4970 |
logging.info("Removing block devices for instance %s", instance.name)
|
4971 |
|
4972 |
if not _RemoveDisks(lu, instance): |
4973 |
if not ignore_failures: |
4974 |
raise errors.OpExecError("Can't remove instance's disks") |
4975 |
feedback_fn("Warning: can't remove instance's disks")
|
4976 |
|
4977 |
logging.info("Removing instance %s out of cluster config", instance.name)
|
4978 |
|
4979 |
lu.cfg.RemoveInstance(instance.name) |
4980 |
|
4981 |
assert not lu.remove_locks.get(locking.LEVEL_INSTANCE), \ |
4982 |
"Instance lock removal conflict"
|
4983 |
|
4984 |
# Remove lock for the instance
|
4985 |
lu.remove_locks[locking.LEVEL_INSTANCE] = instance.name |
4986 |
|
4987 |
|
4988 |
class LUQueryInstances(NoHooksLU): |
4989 |
"""Logical unit for querying instances.
|
4990 |
|
4991 |
"""
|
4992 |
# pylint: disable-msg=W0142
|
4993 |
_OP_PARAMS = [ |
4994 |
("output_fields", _NoDefault, _TListOf(_TNonEmptyString)),
|
4995 |
("names", _EmptyList, _TListOf(_TNonEmptyString)),
|
4996 |
("use_locking", False, _TBool), |
4997 |
] |
4998 |
REQ_BGL = False
|
4999 |
_SIMPLE_FIELDS = ["name", "os", "network_port", "hypervisor", |
5000 |
"serial_no", "ctime", "mtime", "uuid"] |
5001 |
_FIELDS_STATIC = utils.FieldSet(*["name", "os", "pnode", "snodes", |
5002 |
"admin_state",
|
5003 |
"disk_template", "ip", "mac", "bridge", |
5004 |
"nic_mode", "nic_link", |
5005 |
"sda_size", "sdb_size", "vcpus", "tags", |
5006 |
"network_port", "beparams", |
5007 |
r"(disk)\.(size)/([0-9]+)",
|
5008 |
r"(disk)\.(sizes)", "disk_usage", |
5009 |
r"(nic)\.(mac|ip|mode|link)/([0-9]+)",
|
5010 |
r"(nic)\.(bridge)/([0-9]+)",
|
5011 |
r"(nic)\.(macs|ips|modes|links|bridges)",
|
5012 |
r"(disk|nic)\.(count)",
|
5013 |
"hvparams",
|
5014 |
] + _SIMPLE_FIELDS + |
5015 |
["hv/%s" % name
|
5016 |
for name in constants.HVS_PARAMETERS |
5017 |
if name not in constants.HVC_GLOBALS] + |
5018 |
["be/%s" % name
|
5019 |
for name in constants.BES_PARAMETERS]) |
5020 |
_FIELDS_DYNAMIC = utils.FieldSet("oper_state", "oper_ram", "status") |
5021 |
|
5022 |
|
5023 |
def CheckArguments(self): |
5024 |
_CheckOutputFields(static=self._FIELDS_STATIC,
|
5025 |
dynamic=self._FIELDS_DYNAMIC,
|
5026 |
selected=self.op.output_fields)
|
5027 |
|
5028 |
def ExpandNames(self): |
5029 |
self.needed_locks = {}
|
5030 |
self.share_locks[locking.LEVEL_INSTANCE] = 1 |
5031 |
self.share_locks[locking.LEVEL_NODE] = 1 |
5032 |
|
5033 |
if self.op.names: |
5034 |
self.wanted = _GetWantedInstances(self, self.op.names) |
5035 |
else:
|
5036 |
self.wanted = locking.ALL_SET
|
5037 |
|
5038 |
self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields) |
5039 |
self.do_locking = self.do_node_query and self.op.use_locking |
5040 |
if self.do_locking: |
5041 |
self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted |
5042 |
self.needed_locks[locking.LEVEL_NODE] = []
|
5043 |
self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
|
5044 |
|
5045 |
def DeclareLocks(self, level): |
5046 |
if level == locking.LEVEL_NODE and self.do_locking: |
5047 |
self._LockInstancesNodes()
|
5048 |
|
5049 |
def Exec(self, feedback_fn): |
5050 |
"""Computes the list of nodes and their attributes.
|
5051 |
|
5052 |
"""
|
5053 |
# pylint: disable-msg=R0912
|
5054 |
# way too many branches here
|
5055 |
all_info = self.cfg.GetAllInstancesInfo()
|
5056 |
if self.wanted == locking.ALL_SET: |
5057 |
# caller didn't specify instance names, so ordering is not important
|
5058 |
if self.do_locking: |
5059 |
instance_names = self.acquired_locks[locking.LEVEL_INSTANCE]
|
5060 |
else:
|
5061 |
instance_names = all_info.keys() |
5062 |
instance_names = utils.NiceSort(instance_names) |
5063 |
else:
|
5064 |
# caller did specify names, so we must keep the ordering
|
5065 |
if self.do_locking: |
5066 |
tgt_set = self.acquired_locks[locking.LEVEL_INSTANCE]
|
5067 |
else:
|
5068 |
tgt_set = all_info.keys() |
5069 |
missing = set(self.wanted).difference(tgt_set) |
5070 |
if missing:
|
5071 |
raise errors.OpExecError("Some instances were removed before" |
5072 |
" retrieving their data: %s" % missing)
|
5073 |
instance_names = self.wanted
|
5074 |
|
5075 |
instance_list = [all_info[iname] for iname in instance_names] |
5076 |
|
5077 |
# begin data gathering
|
5078 |
|
5079 |
nodes = frozenset([inst.primary_node for inst in instance_list]) |
5080 |
hv_list = list(set([inst.hypervisor for inst in instance_list])) |
5081 |
|
5082 |
bad_nodes = [] |
5083 |
off_nodes = [] |
5084 |
if self.do_node_query: |
5085 |
live_data = {} |
5086 |
node_data = self.rpc.call_all_instances_info(nodes, hv_list)
|
5087 |
for name in nodes: |
5088 |
result = node_data[name] |
5089 |
if result.offline:
|
5090 |
# offline nodes will be in both lists
|
5091 |
off_nodes.append(name) |
5092 |
if result.fail_msg:
|
5093 |
bad_nodes.append(name) |
5094 |
else:
|
5095 |
if result.payload:
|
5096 |
live_data.update(result.payload) |
5097 |
# else no instance is alive
|
5098 |
else:
|
5099 |
live_data = dict([(name, {}) for name in instance_names]) |
5100 |
|
5101 |
# end data gathering
|
5102 |
|
5103 |
HVPREFIX = "hv/"
|
5104 |
BEPREFIX = "be/"
|
5105 |
output = [] |
5106 |
cluster = self.cfg.GetClusterInfo()
|
5107 |
for instance in instance_list: |
5108 |
iout = [] |
5109 |
i_hv = cluster.FillHV(instance, skip_globals=True)
|
5110 |
i_be = cluster.FillBE(instance) |
5111 |
i_nicp = [cluster.SimpleFillNIC(nic.nicparams) for nic in instance.nics] |
5112 |
for field in self.op.output_fields: |
5113 |
st_match = self._FIELDS_STATIC.Matches(field)
|
5114 |
if field in self._SIMPLE_FIELDS: |
5115 |
val = getattr(instance, field)
|
5116 |
elif field == "pnode": |
5117 |
val = instance.primary_node |
5118 |
elif field == "snodes": |
5119 |
val = list(instance.secondary_nodes)
|
5120 |
elif field == "admin_state": |
5121 |
val = instance.admin_up |
5122 |
elif field == "oper_state": |
5123 |
if instance.primary_node in bad_nodes: |
5124 |
val = None
|
5125 |
else:
|
5126 |
val = bool(live_data.get(instance.name))
|
5127 |
elif field == "status": |
5128 |
if instance.primary_node in off_nodes: |
5129 |
val = "ERROR_nodeoffline"
|
5130 |
elif instance.primary_node in bad_nodes: |
5131 |
val = "ERROR_nodedown"
|
5132 |
else:
|
5133 |
running = bool(live_data.get(instance.name))
|
5134 |
if running:
|
5135 |
if instance.admin_up:
|
5136 |
val = "running"
|
5137 |
else:
|
5138 |
val = "ERROR_up"
|
5139 |
else:
|
5140 |
if instance.admin_up:
|
5141 |
val = "ERROR_down"
|
5142 |
else:
|
5143 |
val = "ADMIN_down"
|
5144 |
elif field == "oper_ram": |
5145 |
if instance.primary_node in bad_nodes: |
5146 |
val = None
|
5147 |
elif instance.name in live_data: |
5148 |
val = live_data[instance.name].get("memory", "?") |
5149 |
else:
|
5150 |
val = "-"
|
5151 |
elif field == "vcpus": |
5152 |
val = i_be[constants.BE_VCPUS] |
5153 |
elif field == "disk_template": |
5154 |
val = instance.disk_template |
5155 |
elif field == "ip": |
5156 |
if instance.nics:
|
5157 |
val = instance.nics[0].ip
|
5158 |
else:
|
5159 |
val = None
|
5160 |
elif field == "nic_mode": |
5161 |
if instance.nics:
|
5162 |
val = i_nicp[0][constants.NIC_MODE]
|
5163 |
else:
|
5164 |
val = None
|
5165 |
elif field == "nic_link": |
5166 |
if instance.nics:
|
5167 |
val = i_nicp[0][constants.NIC_LINK]
|
5168 |
else:
|
5169 |
val = None
|
5170 |
elif field == "bridge": |
5171 |
if (instance.nics and |
5172 |
i_nicp[0][constants.NIC_MODE] == constants.NIC_MODE_BRIDGED):
|
5173 |
val = i_nicp[0][constants.NIC_LINK]
|
5174 |
else:
|
5175 |
val = None
|
5176 |
elif field == "mac": |
5177 |
if instance.nics:
|
5178 |
val = instance.nics[0].mac
|
5179 |
else:
|
5180 |
val = None
|
5181 |
elif field == "sda_size" or field == "sdb_size": |
5182 |
idx = ord(field[2]) - ord('a') |
5183 |
try:
|
5184 |
val = instance.FindDisk(idx).size |
5185 |
except errors.OpPrereqError:
|
5186 |
val = None
|
5187 |
elif field == "disk_usage": # total disk usage per node |
5188 |
disk_sizes = [{'size': disk.size} for disk in instance.disks] |
5189 |
val = _ComputeDiskSize(instance.disk_template, disk_sizes) |
5190 |
elif field == "tags": |
5191 |
val = list(instance.GetTags())
|
5192 |
elif field == "hvparams": |
5193 |
val = i_hv |
5194 |
elif (field.startswith(HVPREFIX) and |
5195 |
field[len(HVPREFIX):] in constants.HVS_PARAMETERS and |
5196 |
field[len(HVPREFIX):] not in constants.HVC_GLOBALS): |
5197 |
val = i_hv.get(field[len(HVPREFIX):], None) |
5198 |
elif field == "beparams": |
5199 |
val = i_be |
5200 |
elif (field.startswith(BEPREFIX) and |
5201 |
field[len(BEPREFIX):] in constants.BES_PARAMETERS): |
5202 |
val = i_be.get(field[len(BEPREFIX):], None) |
5203 |
elif st_match and st_match.groups(): |
5204 |
# matches a variable list
|
5205 |
st_groups = st_match.groups() |
5206 |
if st_groups and st_groups[0] == "disk": |
5207 |
if st_groups[1] == "count": |
5208 |
val = len(instance.disks)
|
5209 |
elif st_groups[1] == "sizes": |
5210 |
val = [disk.size for disk in instance.disks] |
5211 |
elif st_groups[1] == "size": |
5212 |
try:
|
5213 |
val = instance.FindDisk(st_groups[2]).size
|
5214 |
except errors.OpPrereqError:
|
5215 |
val = None
|
5216 |
else:
|
5217 |
assert False, "Unhandled disk parameter" |
5218 |
elif st_groups[0] == "nic": |
5219 |
if st_groups[1] == "count": |
5220 |
val = len(instance.nics)
|
5221 |
elif st_groups[1] == "macs": |
5222 |
val = [nic.mac for nic in instance.nics] |
5223 |
elif st_groups[1] == "ips": |
5224 |
val = [nic.ip for nic in instance.nics] |
5225 |
elif st_groups[1] == "modes": |
5226 |
val = [nicp[constants.NIC_MODE] for nicp in i_nicp] |
5227 |
elif st_groups[1] == "links": |
5228 |
val = [nicp[constants.NIC_LINK] for nicp in i_nicp] |
5229 |
elif st_groups[1] == "bridges": |
5230 |
val = [] |
5231 |
for nicp in i_nicp: |
5232 |
if nicp[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
|
5233 |
val.append(nicp[constants.NIC_LINK]) |
5234 |
else:
|
5235 |
val.append(None)
|
5236 |
else:
|
5237 |
# index-based item
|
5238 |
nic_idx = int(st_groups[2]) |
5239 |
if nic_idx >= len(instance.nics): |
5240 |
val = None
|
5241 |
else:
|
5242 |
if st_groups[1] == "mac": |
5243 |
val = instance.nics[nic_idx].mac |
5244 |
elif st_groups[1] == "ip": |
5245 |
val = instance.nics[nic_idx].ip |
5246 |
elif st_groups[1] == "mode": |
5247 |
val = i_nicp[nic_idx][constants.NIC_MODE] |
5248 |
elif st_groups[1] == "link": |
5249 |
val = i_nicp[nic_idx][constants.NIC_LINK] |
5250 |
elif st_groups[1] == "bridge": |
5251 |
nic_mode = i_nicp[nic_idx][constants.NIC_MODE] |
5252 |
if nic_mode == constants.NIC_MODE_BRIDGED:
|
5253 |
val = i_nicp[nic_idx][constants.NIC_LINK] |
5254 |
else:
|
5255 |
val = None
|
5256 |
else:
|
5257 |
assert False, "Unhandled NIC parameter" |
5258 |
else:
|
5259 |
assert False, ("Declared but unhandled variable parameter '%s'" % |
5260 |
field) |
5261 |
else:
|
5262 |
assert False, "Declared but unhandled parameter '%s'" % field |
5263 |
iout.append(val) |
5264 |
output.append(iout) |
5265 |
|
5266 |
return output
|
5267 |
|
5268 |
|
5269 |
class LUFailoverInstance(LogicalUnit): |
5270 |
"""Failover an instance.
|
5271 |
|
5272 |
"""
|
5273 |
HPATH = "instance-failover"
|
5274 |
HTYPE = constants.HTYPE_INSTANCE |
5275 |
_OP_PARAMS = [ |
5276 |
_PInstanceName, |
5277 |
("ignore_consistency", False, _TBool), |
5278 |
_PShutdownTimeout, |
5279 |
] |
5280 |
REQ_BGL = False
|
5281 |
|
5282 |
def ExpandNames(self): |
5283 |
self._ExpandAndLockInstance()
|
5284 |
self.needed_locks[locking.LEVEL_NODE] = []
|
5285 |
self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
|
5286 |
|
5287 |
def DeclareLocks(self, level): |
5288 |
if level == locking.LEVEL_NODE:
|
5289 |
self._LockInstancesNodes()
|
5290 |
|
5291 |
def BuildHooksEnv(self): |
5292 |
"""Build hooks env.
|
5293 |
|
5294 |
This runs on master, primary and secondary nodes of the instance.
|
5295 |
|
5296 |
"""
|
5297 |
instance = self.instance
|
5298 |
source_node = instance.primary_node |
5299 |
target_node = instance.secondary_nodes[0]
|
5300 |
env = { |
5301 |
"IGNORE_CONSISTENCY": self.op.ignore_consistency, |
5302 |
"SHUTDOWN_TIMEOUT": self.op.shutdown_timeout, |
5303 |
"OLD_PRIMARY": source_node,
|
5304 |
"OLD_SECONDARY": target_node,
|
5305 |
"NEW_PRIMARY": target_node,
|
5306 |
"NEW_SECONDARY": source_node,
|
5307 |
} |
5308 |
env.update(_BuildInstanceHookEnvByObject(self, instance))
|
5309 |
nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes) |
5310 |
nl_post = list(nl)
|
5311 |
nl_post.append(source_node) |
5312 |
return env, nl, nl_post
|
5313 |
|
5314 |
def CheckPrereq(self): |
5315 |
"""Check prerequisites.
|
5316 |
|
5317 |
This checks that the instance is in the cluster.
|
5318 |
|
5319 |
"""
|
5320 |
self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name) |
5321 |
assert self.instance is not None, \ |
5322 |
"Cannot retrieve locked instance %s" % self.op.instance_name |
5323 |
|
5324 |
bep = self.cfg.GetClusterInfo().FillBE(instance)
|
5325 |
if instance.disk_template not in constants.DTS_NET_MIRROR: |
5326 |
raise errors.OpPrereqError("Instance's disk layout is not" |
5327 |
" network mirrored, cannot failover.",
|
5328 |
errors.ECODE_STATE) |
5329 |
|
5330 |
secondary_nodes = instance.secondary_nodes |
5331 |
if not secondary_nodes: |
5332 |
raise errors.ProgrammerError("no secondary node but using " |
5333 |
"a mirrored disk template")
|
5334 |
|
5335 |
target_node = secondary_nodes[0]
|
5336 |
_CheckNodeOnline(self, target_node)
|
5337 |
_CheckNodeNotDrained(self, target_node)
|
5338 |
if instance.admin_up:
|
5339 |
# check memory requirements on the secondary node
|
5340 |
_CheckNodeFreeMemory(self, target_node, "failing over instance %s" % |
5341 |
instance.name, bep[constants.BE_MEMORY], |
5342 |
instance.hypervisor) |
5343 |
else:
|
5344 |
self.LogInfo("Not checking memory on the secondary node as" |
5345 |
" instance will not be started")
|
5346 |
|
5347 |
# check bridge existance
|
5348 |
_CheckInstanceBridgesExist(self, instance, node=target_node)
|
5349 |
|
5350 |
def Exec(self, feedback_fn): |
5351 |
"""Failover an instance.
|
5352 |
|
5353 |
The failover is done by shutting it down on its present node and
|
5354 |
starting it on the secondary.
|
5355 |
|
5356 |
"""
|
5357 |
instance = self.instance
|
5358 |
|
5359 |
source_node = instance.primary_node |
5360 |
target_node = instance.secondary_nodes[0]
|
5361 |
|
5362 |
if instance.admin_up:
|
5363 |
feedback_fn("* checking disk consistency between source and target")
|
5364 |
for dev in instance.disks: |
5365 |
# for drbd, these are drbd over lvm
|
5366 |
if not _CheckDiskConsistency(self, dev, target_node, False): |
5367 |
if not self.op.ignore_consistency: |
5368 |
raise errors.OpExecError("Disk %s is degraded on target node," |
5369 |
" aborting failover." % dev.iv_name)
|
5370 |
else:
|
5371 |
feedback_fn("* not checking disk consistency as instance is not running")
|
5372 |
|
5373 |
feedback_fn("* shutting down instance on source node")
|
5374 |
logging.info("Shutting down instance %s on node %s",
|
5375 |
instance.name, source_node) |
5376 |
|
5377 |
result = self.rpc.call_instance_shutdown(source_node, instance,
|
5378 |
self.op.shutdown_timeout)
|
5379 |
msg = result.fail_msg |
5380 |
if msg:
|
5381 |
if self.op.ignore_consistency: |
5382 |
self.proc.LogWarning("Could not shutdown instance %s on node %s." |
5383 |
" Proceeding anyway. Please make sure node"
|
5384 |
" %s is down. Error details: %s",
|
5385 |
instance.name, source_node, source_node, msg) |
5386 |
else:
|
5387 |
raise errors.OpExecError("Could not shutdown instance %s on" |
5388 |
" node %s: %s" %
|
5389 |
(instance.name, source_node, msg)) |
5390 |
|
5391 |
feedback_fn("* deactivating the instance's disks on source node")
|
5392 |
if not _ShutdownInstanceDisks(self, instance, ignore_primary=True): |
5393 |
raise errors.OpExecError("Can't shut down the instance's disks.") |
5394 |
|
5395 |
instance.primary_node = target_node |
5396 |
# distribute new instance config to the other nodes
|
5397 |
self.cfg.Update(instance, feedback_fn)
|
5398 |
|
5399 |
# Only start the instance if it's marked as up
|
5400 |
if instance.admin_up:
|
5401 |
feedback_fn("* activating the instance's disks on target node")
|
5402 |
logging.info("Starting instance %s on node %s",
|
5403 |
instance.name, target_node) |
5404 |
|
5405 |
disks_ok, _ = _AssembleInstanceDisks(self, instance,
|
5406 |
ignore_secondaries=True)
|
5407 |
if not disks_ok: |
5408 |
_ShutdownInstanceDisks(self, instance)
|
5409 |
raise errors.OpExecError("Can't activate the instance's disks") |
5410 |
|
5411 |
feedback_fn("* starting the instance on the target node")
|
5412 |
result = self.rpc.call_instance_start(target_node, instance, None, None) |
5413 |
msg = result.fail_msg |
5414 |
if msg:
|
5415 |
_ShutdownInstanceDisks(self, instance)
|
5416 |
raise errors.OpExecError("Could not start instance %s on node %s: %s" % |
5417 |
(instance.name, target_node, msg)) |
5418 |
|
5419 |
|
5420 |
class LUMigrateInstance(LogicalUnit): |
5421 |
"""Migrate an instance.
|
5422 |
|
5423 |
This is migration without shutting down, compared to the failover,
|
5424 |
which is done with shutdown.
|
5425 |
|
5426 |
"""
|
5427 |
HPATH = "instance-migrate"
|
5428 |
HTYPE = constants.HTYPE_INSTANCE |
5429 |
_OP_PARAMS = [ |
5430 |
_PInstanceName, |
5431 |
("live", True, _TBool), |
5432 |
("cleanup", False, _TBool), |
5433 |
] |
5434 |
|
5435 |
REQ_BGL = False
|
5436 |
|
5437 |
def ExpandNames(self): |
5438 |
self._ExpandAndLockInstance()
|
5439 |
|
5440 |
self.needed_locks[locking.LEVEL_NODE] = []
|
5441 |
self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
|
5442 |
|
5443 |
self._migrater = TLMigrateInstance(self, self.op.instance_name, |
5444 |
self.op.live, self.op.cleanup) |
5445 |
self.tasklets = [self._migrater] |
5446 |
|
5447 |
def DeclareLocks(self, level): |
5448 |
if level == locking.LEVEL_NODE:
|
5449 |
self._LockInstancesNodes()
|
5450 |
|
5451 |
def BuildHooksEnv(self): |
5452 |
"""Build hooks env.
|
5453 |
|
5454 |
This runs on master, primary and secondary nodes of the instance.
|
5455 |
|
5456 |
"""
|
5457 |
instance = self._migrater.instance
|
5458 |
source_node = instance.primary_node |
5459 |
target_node = instance.secondary_nodes[0]
|
5460 |
env = _BuildInstanceHookEnvByObject(self, instance)
|
5461 |
env["MIGRATE_LIVE"] = self.op.live |
5462 |
env["MIGRATE_CLEANUP"] = self.op.cleanup |
5463 |
env.update({ |
5464 |
"OLD_PRIMARY": source_node,
|
5465 |
"OLD_SECONDARY": target_node,
|
5466 |
"NEW_PRIMARY": target_node,
|
5467 |
"NEW_SECONDARY": source_node,
|
5468 |
}) |
5469 |
nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes) |
5470 |
nl_post = list(nl)
|
5471 |
nl_post.append(source_node) |
5472 |
return env, nl, nl_post
|
5473 |
|
5474 |
|
5475 |
class LUMoveInstance(LogicalUnit): |
5476 |
"""Move an instance by data-copying.
|
5477 |
|
5478 |
"""
|
5479 |
HPATH = "instance-move"
|
5480 |
HTYPE = constants.HTYPE_INSTANCE |
5481 |
_OP_PARAMS = [ |
5482 |
_PInstanceName, |
5483 |
("target_node", _NoDefault, _TNonEmptyString),
|
5484 |
_PShutdownTimeout, |
5485 |
] |
5486 |
REQ_BGL = False
|
5487 |
|
5488 |
def ExpandNames(self): |
5489 |
self._ExpandAndLockInstance()
|
5490 |
target_node = _ExpandNodeName(self.cfg, self.op.target_node) |
5491 |
self.op.target_node = target_node
|
5492 |
self.needed_locks[locking.LEVEL_NODE] = [target_node]
|
5493 |
self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
|
5494 |
|
5495 |
def DeclareLocks(self, level): |
5496 |
if level == locking.LEVEL_NODE:
|
5497 |
self._LockInstancesNodes(primary_only=True) |
5498 |
|
5499 |
def BuildHooksEnv(self): |
5500 |
"""Build hooks env.
|
5501 |
|
5502 |
This runs on master, primary and secondary nodes of the instance.
|
5503 |
|
5504 |
"""
|
5505 |
env = { |
5506 |
"TARGET_NODE": self.op.target_node, |
5507 |
"SHUTDOWN_TIMEOUT": self.op.shutdown_timeout, |
5508 |
} |
5509 |
env.update(_BuildInstanceHookEnvByObject(self, self.instance)) |
5510 |
nl = [self.cfg.GetMasterNode()] + [self.instance.primary_node, |
5511 |
self.op.target_node]
|
5512 |
return env, nl, nl
|
5513 |
|
5514 |
def CheckPrereq(self): |
5515 |
"""Check prerequisites.
|
5516 |
|
5517 |
This checks that the instance is in the cluster.
|
5518 |
|
5519 |
"""
|
5520 |
self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name) |
5521 |
assert self.instance is not None, \ |
5522 |
"Cannot retrieve locked instance %s" % self.op.instance_name |
5523 |
|
5524 |
node = self.cfg.GetNodeInfo(self.op.target_node) |
5525 |
assert node is not None, \ |
5526 |
"Cannot retrieve locked node %s" % self.op.target_node |
5527 |
|
5528 |
self.target_node = target_node = node.name
|
5529 |
|
5530 |
if target_node == instance.primary_node:
|
5531 |
raise errors.OpPrereqError("Instance %s is already on the node %s" % |
5532 |
(instance.name, target_node), |
5533 |
errors.ECODE_STATE) |
5534 |
|
5535 |
bep = self.cfg.GetClusterInfo().FillBE(instance)
|
5536 |
|
5537 |
for idx, dsk in enumerate(instance.disks): |
5538 |
if dsk.dev_type not in (constants.LD_LV, constants.LD_FILE): |
5539 |
raise errors.OpPrereqError("Instance disk %d has a complex layout," |
5540 |
" cannot copy" % idx, errors.ECODE_STATE)
|
5541 |
|
5542 |
_CheckNodeOnline(self, target_node)
|
5543 |
_CheckNodeNotDrained(self, target_node)
|
5544 |
|
5545 |
if instance.admin_up:
|
5546 |
# check memory requirements on the secondary node
|
5547 |
_CheckNodeFreeMemory(self, target_node, "failing over instance %s" % |
5548 |
instance.name, bep[constants.BE_MEMORY], |
5549 |
instance.hypervisor) |
5550 |
else:
|
5551 |
self.LogInfo("Not checking memory on the secondary node as" |
5552 |
" instance will not be started")
|
5553 |
|
5554 |
# check bridge existance
|
5555 |
_CheckInstanceBridgesExist(self, instance, node=target_node)
|
5556 |
|
5557 |
def Exec(self, feedback_fn): |
5558 |
"""Move an instance.
|
5559 |
|
5560 |
The move is done by shutting it down on its present node, copying
|
5561 |
the data over (slow) and starting it on the new node.
|
5562 |
|
5563 |
"""
|
5564 |
instance = self.instance
|
5565 |
|
5566 |
source_node = instance.primary_node |
5567 |
target_node = self.target_node
|
5568 |
|
5569 |
self.LogInfo("Shutting down instance %s on source node %s", |
5570 |
instance.name, source_node) |
5571 |
|
5572 |
result = self.rpc.call_instance_shutdown(source_node, instance,
|
5573 |
self.op.shutdown_timeout)
|
5574 |
msg = result.fail_msg |
5575 |
if msg:
|
5576 |
if self.op.ignore_consistency: |
5577 |
self.proc.LogWarning("Could not shutdown instance %s on node %s." |
5578 |
" Proceeding anyway. Please make sure node"
|
5579 |
" %s is down. Error details: %s",
|
5580 |
instance.name, source_node, source_node, msg) |
5581 |
else:
|
5582 |
raise errors.OpExecError("Could not shutdown instance %s on" |
5583 |
" node %s: %s" %
|
5584 |
(instance.name, source_node, msg)) |
5585 |
|
5586 |
# create the target disks
|
5587 |
try:
|
5588 |
_CreateDisks(self, instance, target_node=target_node)
|
5589 |
except errors.OpExecError:
|
5590 |
self.LogWarning("Device creation failed, reverting...") |
5591 |
try:
|
5592 |
_RemoveDisks(self, instance, target_node=target_node)
|
5593 |
finally:
|
5594 |
self.cfg.ReleaseDRBDMinors(instance.name)
|
5595 |
raise
|
5596 |
|
5597 |
cluster_name = self.cfg.GetClusterInfo().cluster_name
|
5598 |
|
5599 |
errs = [] |
5600 |
# activate, get path, copy the data over
|
5601 |
for idx, disk in enumerate(instance.disks): |
5602 |
self.LogInfo("Copying data for disk %d", idx) |
5603 |
result = self.rpc.call_blockdev_assemble(target_node, disk,
|
5604 |
instance.name, True)
|
5605 |
if result.fail_msg:
|
5606 |
self.LogWarning("Can't assemble newly created disk %d: %s", |
5607 |
idx, result.fail_msg) |
5608 |
errs.append(result.fail_msg) |
5609 |
break
|
5610 |
dev_path = result.payload |
5611 |
result = self.rpc.call_blockdev_export(source_node, disk,
|
5612 |
target_node, dev_path, |
5613 |
cluster_name) |
5614 |
if result.fail_msg:
|
5615 |
self.LogWarning("Can't copy data over for disk %d: %s", |
5616 |
idx, result.fail_msg) |
5617 |
errs.append(result.fail_msg) |
5618 |
break
|
5619 |
|
5620 |
if errs:
|
5621 |
self.LogWarning("Some disks failed to copy, aborting") |
5622 |
try:
|
5623 |
_RemoveDisks(self, instance, target_node=target_node)
|
5624 |
finally:
|
5625 |
self.cfg.ReleaseDRBDMinors(instance.name)
|
5626 |
raise errors.OpExecError("Errors during disk copy: %s" % |
5627 |
(",".join(errs),))
|
5628 |
|
5629 |
instance.primary_node = target_node |
5630 |
self.cfg.Update(instance, feedback_fn)
|
5631 |
|
5632 |
self.LogInfo("Removing the disks on the original node") |
5633 |
_RemoveDisks(self, instance, target_node=source_node)
|
5634 |
|
5635 |
# Only start the instance if it's marked as up
|
5636 |
if instance.admin_up:
|
5637 |
self.LogInfo("Starting instance %s on node %s", |
5638 |
instance.name, target_node) |
5639 |
|
5640 |
disks_ok, _ = _AssembleInstanceDisks(self, instance,
|
5641 |
ignore_secondaries=True)
|
5642 |
if not disks_ok: |
5643 |
_ShutdownInstanceDisks(self, instance)
|
5644 |
raise errors.OpExecError("Can't activate the instance's disks") |
5645 |
|
5646 |
result = self.rpc.call_instance_start(target_node, instance, None, None) |
5647 |
msg = result.fail_msg |
5648 |
if msg:
|
5649 |
_ShutdownInstanceDisks(self, instance)
|
5650 |
raise errors.OpExecError("Could not start instance %s on node %s: %s" % |
5651 |
(instance.name, target_node, msg)) |
5652 |
|
5653 |
|
5654 |
class LUMigrateNode(LogicalUnit): |
5655 |
"""Migrate all instances from a node.
|
5656 |
|
5657 |
"""
|
5658 |
HPATH = "node-migrate"
|
5659 |
HTYPE = constants.HTYPE_NODE |
5660 |
_OP_PARAMS = [ |
5661 |
_PNodeName, |
5662 |
("live", False, _TBool), |
5663 |
] |
5664 |
REQ_BGL = False
|
5665 |
|
5666 |
def ExpandNames(self): |
5667 |
self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name) |
5668 |
|
5669 |
self.needed_locks = {
|
5670 |
locking.LEVEL_NODE: [self.op.node_name],
|
5671 |
} |
5672 |
|
5673 |
self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
|
5674 |
|
5675 |
# Create tasklets for migrating instances for all instances on this node
|
5676 |
names = [] |
5677 |
tasklets = [] |
5678 |
|
5679 |
for inst in _GetNodePrimaryInstances(self.cfg, self.op.node_name): |
5680 |
logging.debug("Migrating instance %s", inst.name)
|
5681 |
names.append(inst.name) |
5682 |
|
5683 |
tasklets.append(TLMigrateInstance(self, inst.name, self.op.live, False)) |
5684 |
|
5685 |
self.tasklets = tasklets
|
5686 |
|
5687 |
# Declare instance locks
|
5688 |
self.needed_locks[locking.LEVEL_INSTANCE] = names
|
5689 |
|
5690 |
def DeclareLocks(self, level): |
5691 |
if level == locking.LEVEL_NODE:
|
5692 |
self._LockInstancesNodes()
|
5693 |
|
5694 |
def BuildHooksEnv(self): |
5695 |
"""Build hooks env.
|
5696 |
|
5697 |
This runs on the master, the primary and all the secondaries.
|
5698 |
|
5699 |
"""
|
5700 |
env = { |
5701 |
"NODE_NAME": self.op.node_name, |
5702 |
} |
5703 |
|
5704 |
nl = [self.cfg.GetMasterNode()]
|
5705 |
|
5706 |
return (env, nl, nl)
|
5707 |
|
5708 |
|
5709 |
class TLMigrateInstance(Tasklet): |
5710 |
def __init__(self, lu, instance_name, live, cleanup): |
5711 |
"""Initializes this class.
|
5712 |
|
5713 |
"""
|
5714 |
Tasklet.__init__(self, lu)
|
5715 |
|
5716 |
# Parameters
|
5717 |
self.instance_name = instance_name
|
5718 |
self.live = live
|
5719 |
self.cleanup = cleanup
|
5720 |
|
5721 |
def CheckPrereq(self): |
5722 |
"""Check prerequisites.
|
5723 |
|
5724 |
This checks that the instance is in the cluster.
|
5725 |
|
5726 |
"""
|
5727 |
instance_name = _ExpandInstanceName(self.lu.cfg, self.instance_name) |
5728 |
instance = self.cfg.GetInstanceInfo(instance_name)
|
5729 |
assert instance is not None |
5730 |
|
5731 |
if instance.disk_template != constants.DT_DRBD8:
|
5732 |
raise errors.OpPrereqError("Instance's disk layout is not" |
5733 |
" drbd8, cannot migrate.", errors.ECODE_STATE)
|
5734 |
|
5735 |
secondary_nodes = instance.secondary_nodes |
5736 |
if not secondary_nodes: |
5737 |
raise errors.ConfigurationError("No secondary node but using" |
5738 |
" drbd8 disk template")
|
5739 |
|
5740 |
i_be = self.cfg.GetClusterInfo().FillBE(instance)
|
5741 |
|
5742 |
target_node = secondary_nodes[0]
|
5743 |
# check memory requirements on the secondary node
|
5744 |
_CheckNodeFreeMemory(self.lu, target_node, "migrating instance %s" % |
5745 |
instance.name, i_be[constants.BE_MEMORY], |
5746 |
instance.hypervisor) |
5747 |
|
5748 |
# check bridge existance
|
5749 |
_CheckInstanceBridgesExist(self.lu, instance, node=target_node)
|
5750 |
|
5751 |
if not self.cleanup: |
5752 |
_CheckNodeNotDrained(self.lu, target_node)
|
5753 |
result = self.rpc.call_instance_migratable(instance.primary_node,
|
5754 |
instance) |
5755 |
result.Raise("Can't migrate, please use failover",
|
5756 |
prereq=True, ecode=errors.ECODE_STATE)
|
5757 |
|
5758 |
self.instance = instance
|
5759 |
|
5760 |
def _WaitUntilSync(self): |
5761 |
"""Poll with custom rpc for disk sync.
|
5762 |
|
5763 |
This uses our own step-based rpc call.
|
5764 |
|
5765 |
"""
|
5766 |
self.feedback_fn("* wait until resync is done") |
5767 |
all_done = False
|
5768 |
while not all_done: |
5769 |
all_done = True
|
5770 |
result = self.rpc.call_drbd_wait_sync(self.all_nodes, |
5771 |
self.nodes_ip,
|
5772 |
self.instance.disks)
|
5773 |
min_percent = 100
|
5774 |
for node, nres in result.items(): |
5775 |
nres.Raise("Cannot resync disks on node %s" % node)
|
5776 |
node_done, node_percent = nres.payload |
5777 |
all_done = all_done and node_done
|
5778 |
if node_percent is not None: |
5779 |
min_percent = min(min_percent, node_percent)
|
5780 |
if not all_done: |
5781 |
if min_percent < 100: |
5782 |
self.feedback_fn(" - progress: %.1f%%" % min_percent) |
5783 |
time.sleep(2)
|
5784 |
|
5785 |
def _EnsureSecondary(self, node): |
5786 |
"""Demote a node to secondary.
|
5787 |
|
5788 |
"""
|
5789 |
self.feedback_fn("* switching node %s to secondary mode" % node) |
5790 |
|
5791 |
for dev in self.instance.disks: |
5792 |
self.cfg.SetDiskID(dev, node)
|
5793 |
|
5794 |
result = self.rpc.call_blockdev_close(node, self.instance.name, |
5795 |
self.instance.disks)
|
5796 |
result.Raise("Cannot change disk to secondary on node %s" % node)
|
5797 |
|
5798 |
def _GoStandalone(self): |
5799 |
"""Disconnect from the network.
|
5800 |
|
5801 |
"""
|
5802 |
self.feedback_fn("* changing into standalone mode") |
5803 |
result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip, |
5804 |
self.instance.disks)
|
5805 |
for node, nres in result.items(): |
5806 |
nres.Raise("Cannot disconnect disks node %s" % node)
|
5807 |
|
5808 |
def _GoReconnect(self, multimaster): |
5809 |
"""Reconnect to the network.
|
5810 |
|
5811 |
"""
|
5812 |
if multimaster:
|
5813 |
msg = "dual-master"
|
5814 |
else:
|
5815 |
msg = "single-master"
|
5816 |
self.feedback_fn("* changing disks into %s mode" % msg) |
5817 |
result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip, |
5818 |
self.instance.disks,
|
5819 |
self.instance.name, multimaster)
|
5820 |
for node, nres in result.items(): |
5821 |
nres.Raise("Cannot change disks config on node %s" % node)
|
5822 |
|
5823 |
def _ExecCleanup(self): |
5824 |
"""Try to cleanup after a failed migration.
|
5825 |
|
5826 |
The cleanup is done by:
|
5827 |
- check that the instance is running only on one node
|
5828 |
(and update the config if needed)
|
5829 |
- change disks on its secondary node to secondary
|
5830 |
- wait until disks are fully synchronized
|
5831 |
- disconnect from the network
|
5832 |
- change disks into single-master mode
|
5833 |
- wait again until disks are fully synchronized
|
5834 |
|
5835 |
"""
|
5836 |
instance = self.instance
|
5837 |
target_node = self.target_node
|
5838 |
source_node = self.source_node
|
5839 |
|
5840 |
# check running on only one node
|
5841 |
self.feedback_fn("* checking where the instance actually runs" |
5842 |
" (if this hangs, the hypervisor might be in"
|
5843 |
" a bad state)")
|
5844 |
ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor]) |
5845 |
for node, result in ins_l.items(): |
5846 |
result.Raise("Can't contact node %s" % node)
|
5847 |
|
5848 |
runningon_source = instance.name in ins_l[source_node].payload
|
5849 |
runningon_target = instance.name in ins_l[target_node].payload
|
5850 |
|
5851 |
if runningon_source and runningon_target: |
5852 |
raise errors.OpExecError("Instance seems to be running on two nodes," |
5853 |
" or the hypervisor is confused. You will have"
|
5854 |
" to ensure manually that it runs only on one"
|
5855 |
" and restart this operation.")
|
5856 |
|
5857 |
if not (runningon_source or runningon_target): |
5858 |
raise errors.OpExecError("Instance does not seem to be running at all." |
5859 |
" In this case, it's safer to repair by"
|
5860 |
" running 'gnt-instance stop' to ensure disk"
|
5861 |
" shutdown, and then restarting it.")
|
5862 |
|
5863 |
if runningon_target:
|
5864 |
# the migration has actually succeeded, we need to update the config
|
5865 |
self.feedback_fn("* instance running on secondary node (%s)," |
5866 |
" updating config" % target_node)
|
5867 |
instance.primary_node = target_node |
5868 |
self.cfg.Update(instance, self.feedback_fn) |
5869 |
demoted_node = source_node |
5870 |
else:
|
5871 |
self.feedback_fn("* instance confirmed to be running on its" |
5872 |
" primary node (%s)" % source_node)
|
5873 |
demoted_node = target_node |
5874 |
|
5875 |
self._EnsureSecondary(demoted_node)
|
5876 |
try:
|
5877 |
self._WaitUntilSync()
|
5878 |
except errors.OpExecError:
|
5879 |
# we ignore here errors, since if the device is standalone, it
|
5880 |
# won't be able to sync
|
5881 |
pass
|
5882 |
self._GoStandalone()
|
5883 |
self._GoReconnect(False) |
5884 |
self._WaitUntilSync()
|
5885 |
|
5886 |
self.feedback_fn("* done") |
5887 |
|
5888 |
def _RevertDiskStatus(self): |
5889 |
"""Try to revert the disk status after a failed migration.
|
5890 |
|
5891 |
"""
|
5892 |
target_node = self.target_node
|
5893 |
try:
|
5894 |
self._EnsureSecondary(target_node)
|
5895 |
self._GoStandalone()
|
5896 |
self._GoReconnect(False) |
5897 |
self._WaitUntilSync()
|
5898 |
except errors.OpExecError, err:
|
5899 |
self.lu.LogWarning("Migration failed and I can't reconnect the" |
5900 |
" drives: error '%s'\n"
|
5901 |
"Please look and recover the instance status" %
|
5902 |
str(err))
|
5903 |
|
5904 |
def _AbortMigration(self): |
5905 |
"""Call the hypervisor code to abort a started migration.
|
5906 |
|
5907 |
"""
|
5908 |
instance = self.instance
|
5909 |
target_node = self.target_node
|
5910 |
migration_info = self.migration_info
|
5911 |
|
5912 |
abort_result = self.rpc.call_finalize_migration(target_node,
|
5913 |
instance, |
5914 |
migration_info, |
5915 |
False)
|
5916 |
abort_msg = abort_result.fail_msg |
5917 |
if abort_msg:
|
5918 |
logging.error("Aborting migration failed on target node %s: %s",
|
5919 |
target_node, abort_msg) |
5920 |
# Don't raise an exception here, as we stil have to try to revert the
|
5921 |
# disk status, even if this step failed.
|
5922 |
|
5923 |
def _ExecMigration(self): |
5924 |
"""Migrate an instance.
|
5925 |
|
5926 |
The migrate is done by:
|
5927 |
- change the disks into dual-master mode
|
5928 |
- wait until disks are fully synchronized again
|
5929 |
- migrate the instance
|
5930 |
- change disks on the new secondary node (the old primary) to secondary
|
5931 |
- wait until disks are fully synchronized
|
5932 |
- change disks into single-master mode
|
5933 |
|
5934 |
"""
|
5935 |
instance = self.instance
|
5936 |
target_node = self.target_node
|
5937 |
source_node = self.source_node
|
5938 |
|
5939 |
self.feedback_fn("* checking disk consistency between source and target") |
5940 |
for dev in instance.disks: |
5941 |
if not _CheckDiskConsistency(self.lu, dev, target_node, False): |
5942 |
raise errors.OpExecError("Disk %s is degraded or not fully" |
5943 |
" synchronized on target node,"
|
5944 |
" aborting migrate." % dev.iv_name)
|
5945 |
|
5946 |
# First get the migration information from the remote node
|
5947 |
result = self.rpc.call_migration_info(source_node, instance)
|
5948 |
msg = result.fail_msg |
5949 |
if msg:
|
5950 |
log_err = ("Failed fetching source migration information from %s: %s" %
|
5951 |
(source_node, msg)) |
5952 |
logging.error(log_err) |
5953 |
raise errors.OpExecError(log_err)
|
5954 |
|
5955 |
self.migration_info = migration_info = result.payload
|
5956 |
|
5957 |
# Then switch the disks to master/master mode
|
5958 |
self._EnsureSecondary(target_node)
|
5959 |
self._GoStandalone()
|
5960 |
self._GoReconnect(True) |
5961 |
self._WaitUntilSync()
|
5962 |
|
5963 |
self.feedback_fn("* preparing %s to accept the instance" % target_node) |
5964 |
result = self.rpc.call_accept_instance(target_node,
|
5965 |
instance, |
5966 |
migration_info, |
5967 |
self.nodes_ip[target_node])
|
5968 |
|
5969 |
msg = result.fail_msg |
5970 |
if msg:
|
5971 |
logging.error("Instance pre-migration failed, trying to revert"
|
5972 |
" disk status: %s", msg)
|
5973 |
self.feedback_fn("Pre-migration failed, aborting") |
5974 |
self._AbortMigration()
|
5975 |
self._RevertDiskStatus()
|
5976 |
raise errors.OpExecError("Could not pre-migrate instance %s: %s" % |
5977 |
(instance.name, msg)) |
5978 |
|
5979 |
self.feedback_fn("* migrating instance to %s" % target_node) |
5980 |
time.sleep(10)
|
5981 |
result = self.rpc.call_instance_migrate(source_node, instance,
|
5982 |
self.nodes_ip[target_node],
|
5983 |
self.live)
|
5984 |
msg = result.fail_msg |
5985 |
if msg:
|
5986 |
logging.error("Instance migration failed, trying to revert"
|
5987 |
" disk status: %s", msg)
|
5988 |
self.feedback_fn("Migration failed, aborting") |
5989 |
self._AbortMigration()
|
5990 |
self._RevertDiskStatus()
|
5991 |
raise errors.OpExecError("Could not migrate instance %s: %s" % |
5992 |
(instance.name, msg)) |
5993 |
time.sleep(10)
|
5994 |
|
5995 |
instance.primary_node = target_node |
5996 |
# distribute new instance config to the other nodes
|
5997 |
self.cfg.Update(instance, self.feedback_fn) |
5998 |
|
5999 |
result = self.rpc.call_finalize_migration(target_node,
|
6000 |
instance, |
6001 |
migration_info, |
6002 |
True)
|
6003 |
msg = result.fail_msg |
6004 |
if msg:
|
6005 |
logging.error("Instance migration succeeded, but finalization failed:"
|
6006 |
" %s", msg)
|
6007 |
raise errors.OpExecError("Could not finalize instance migration: %s" % |
6008 |
msg) |
6009 |
|
6010 |
self._EnsureSecondary(source_node)
|
6011 |
self._WaitUntilSync()
|
6012 |
self._GoStandalone()
|
6013 |
self._GoReconnect(False) |
6014 |
self._WaitUntilSync()
|
6015 |
|
6016 |
self.feedback_fn("* done") |
6017 |
|
6018 |
def Exec(self, feedback_fn): |
6019 |
"""Perform the migration.
|
6020 |
|
6021 |
"""
|
6022 |
feedback_fn("Migrating instance %s" % self.instance.name) |
6023 |
|
6024 |
self.feedback_fn = feedback_fn
|
6025 |
|
6026 |
self.source_node = self.instance.primary_node |
6027 |
self.target_node = self.instance.secondary_nodes[0] |
6028 |
self.all_nodes = [self.source_node, self.target_node] |
6029 |
self.nodes_ip = {
|
6030 |
self.source_node: self.cfg.GetNodeInfo(self.source_node).secondary_ip, |
6031 |
self.target_node: self.cfg.GetNodeInfo(self.target_node).secondary_ip, |
6032 |
} |
6033 |
|
6034 |
if self.cleanup: |
6035 |
return self._ExecCleanup() |
6036 |
else:
|
6037 |
return self._ExecMigration() |
6038 |
|
6039 |
|
6040 |
def _CreateBlockDev(lu, node, instance, device, force_create, |
6041 |
info, force_open): |
6042 |
"""Create a tree of block devices on a given node.
|
6043 |
|
6044 |
If this device type has to be created on secondaries, create it and
|
6045 |
all its children.
|
6046 |
|
6047 |
If not, just recurse to children keeping the same 'force' value.
|
6048 |
|
6049 |
@param lu: the lu on whose behalf we execute
|
6050 |
@param node: the node on which to create the device
|
6051 |
@type instance: L{objects.Instance}
|
6052 |
@param instance: the instance which owns the device
|
6053 |
@type device: L{objects.Disk}
|
6054 |
@param device: the device to create
|
6055 |
@type force_create: boolean
|
6056 |
@param force_create: whether to force creation of this device; this
|
6057 |
will be change to True whenever we find a device which has
|
6058 |
CreateOnSecondary() attribute
|
6059 |
@param info: the extra 'metadata' we should attach to the device
|
6060 |
(this will be represented as a LVM tag)
|
6061 |
@type force_open: boolean
|
6062 |
@param force_open: this parameter will be passes to the
|
6063 |
L{backend.BlockdevCreate} function where it specifies
|
6064 |
whether we run on primary or not, and it affects both
|
6065 |
the child assembly and the device own Open() execution
|
6066 |
|
6067 |
"""
|
6068 |
if device.CreateOnSecondary():
|
6069 |
force_create = True
|
6070 |
|
6071 |
if device.children:
|
6072 |
for child in device.children: |
6073 |
_CreateBlockDev(lu, node, instance, child, force_create, |
6074 |
info, force_open) |
6075 |
|
6076 |
if not force_create: |
6077 |
return
|
6078 |
|
6079 |
_CreateSingleBlockDev(lu, node, instance, device, info, force_open) |
6080 |
|
6081 |
|
6082 |
def _CreateSingleBlockDev(lu, node, instance, device, info, force_open): |
6083 |
"""Create a single block device on a given node.
|
6084 |
|
6085 |
This will not recurse over children of the device, so they must be
|
6086 |
created in advance.
|
6087 |
|
6088 |
@param lu: the lu on whose behalf we execute
|
6089 |
@param node: the node on which to create the device
|
6090 |
@type instance: L{objects.Instance}
|
6091 |
@param instance: the instance which owns the device
|
6092 |
@type device: L{objects.Disk}
|
6093 |
@param device: the device to create
|
6094 |
@param info: the extra 'metadata' we should attach to the device
|
6095 |
(this will be represented as a LVM tag)
|
6096 |
@type force_open: boolean
|
6097 |
@param force_open: this parameter will be passes to the
|
6098 |
L{backend.BlockdevCreate} function where it specifies
|
6099 |
whether we run on primary or not, and it affects both
|
6100 |
the child assembly and the device own Open() execution
|
6101 |
|
6102 |
"""
|
6103 |
lu.cfg.SetDiskID(device, node) |
6104 |
result = lu.rpc.call_blockdev_create(node, device, device.size, |
6105 |
instance.name, force_open, info) |
6106 |
result.Raise("Can't create block device %s on"
|
6107 |
" node %s for instance %s" % (device, node, instance.name))
|
6108 |
if device.physical_id is None: |
6109 |
device.physical_id = result.payload |
6110 |
|
6111 |
|
6112 |
def _GenerateUniqueNames(lu, exts): |
6113 |
"""Generate a suitable LV name.
|
6114 |
|
6115 |
This will generate a logical volume name for the given instance.
|
6116 |
|
6117 |
"""
|
6118 |
results = [] |
6119 |
for val in exts: |
6120 |
new_id = lu.cfg.GenerateUniqueID(lu.proc.GetECId()) |
6121 |
results.append("%s%s" % (new_id, val))
|
6122 |
return results
|
6123 |
|
6124 |
|
6125 |
def _GenerateDRBD8Branch(lu, primary, secondary, size, names, iv_name, |
6126 |
p_minor, s_minor): |
6127 |
"""Generate a drbd8 device complete with its children.
|
6128 |
|
6129 |
"""
|
6130 |
port = lu.cfg.AllocatePort() |
6131 |
vgname = lu.cfg.GetVGName() |
6132 |
shared_secret = lu.cfg.GenerateDRBDSecret(lu.proc.GetECId()) |
6133 |
dev_data = objects.Disk(dev_type=constants.LD_LV, size=size, |
6134 |
logical_id=(vgname, names[0]))
|
6135 |
dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
|
6136 |
logical_id=(vgname, names[1]))
|
6137 |
drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size, |
6138 |
logical_id=(primary, secondary, port, |
6139 |
p_minor, s_minor, |
6140 |
shared_secret), |
6141 |
children=[dev_data, dev_meta], |
6142 |
iv_name=iv_name) |
6143 |
return drbd_dev
|
6144 |
|
6145 |
|
6146 |
def _GenerateDiskTemplate(lu, template_name, |
6147 |
instance_name, primary_node, |
6148 |
secondary_nodes, disk_info, |
6149 |
file_storage_dir, file_driver, |
6150 |
base_index): |
6151 |
"""Generate the entire disk layout for a given template type.
|
6152 |
|
6153 |
"""
|
6154 |
#TODO: compute space requirements
|
6155 |
|
6156 |
vgname = lu.cfg.GetVGName() |
6157 |
disk_count = len(disk_info)
|
6158 |
disks = [] |
6159 |
if template_name == constants.DT_DISKLESS:
|
6160 |
pass
|
6161 |
elif template_name == constants.DT_PLAIN:
|
6162 |
if len(secondary_nodes) != 0: |
6163 |
raise errors.ProgrammerError("Wrong template configuration") |
6164 |
|
6165 |
names = _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
|
6166 |
for i in range(disk_count)]) |
6167 |
for idx, disk in enumerate(disk_info): |
6168 |
disk_index = idx + base_index |
6169 |
disk_dev = objects.Disk(dev_type=constants.LD_LV, size=disk["size"],
|
6170 |
logical_id=(vgname, names[idx]), |
6171 |
iv_name="disk/%d" % disk_index,
|
6172 |
mode=disk["mode"])
|
6173 |
disks.append(disk_dev) |
6174 |
elif template_name == constants.DT_DRBD8:
|
6175 |
if len(secondary_nodes) != 1: |
6176 |
raise errors.ProgrammerError("Wrong template configuration") |
6177 |
remote_node = secondary_nodes[0]
|
6178 |
minors = lu.cfg.AllocateDRBDMinor( |
6179 |
[primary_node, remote_node] * len(disk_info), instance_name)
|
6180 |
|
6181 |
names = [] |
6182 |
for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % (base_index + i) |
6183 |
for i in range(disk_count)]): |
6184 |
names.append(lv_prefix + "_data")
|
6185 |
names.append(lv_prefix + "_meta")
|
6186 |
for idx, disk in enumerate(disk_info): |
6187 |
disk_index = idx + base_index |
6188 |
disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node, |
6189 |
disk["size"], names[idx*2:idx*2+2], |
6190 |
"disk/%d" % disk_index,
|
6191 |
minors[idx*2], minors[idx*2+1]) |
6192 |
disk_dev.mode = disk["mode"]
|
6193 |
disks.append(disk_dev) |
6194 |
elif template_name == constants.DT_FILE:
|
6195 |
if len(secondary_nodes) != 0: |
6196 |
raise errors.ProgrammerError("Wrong template configuration") |
6197 |
|
6198 |
_RequireFileStorage() |
6199 |
|
6200 |
for idx, disk in enumerate(disk_info): |
6201 |
disk_index = idx + base_index |
6202 |
disk_dev = objects.Disk(dev_type=constants.LD_FILE, size=disk["size"],
|
6203 |
iv_name="disk/%d" % disk_index,
|
6204 |
logical_id=(file_driver, |
6205 |
"%s/disk%d" % (file_storage_dir,
|
6206 |
disk_index)), |
6207 |
mode=disk["mode"])
|
6208 |
disks.append(disk_dev) |
6209 |
else:
|
6210 |
raise errors.ProgrammerError("Invalid disk template '%s'" % template_name) |
6211 |
return disks
|
6212 |
|
6213 |
|
6214 |
def _GetInstanceInfoText(instance): |
6215 |
"""Compute that text that should be added to the disk's metadata.
|
6216 |
|
6217 |
"""
|
6218 |
return "originstname+%s" % instance.name |
6219 |
|
6220 |
|
6221 |
def _CreateDisks(lu, instance, to_skip=None, target_node=None): |
6222 |
"""Create all disks for an instance.
|
6223 |
|
6224 |
This abstracts away some work from AddInstance.
|
6225 |
|
6226 |
@type lu: L{LogicalUnit}
|
6227 |
@param lu: the logical unit on whose behalf we execute
|
6228 |
@type instance: L{objects.Instance}
|
6229 |
@param instance: the instance whose disks we should create
|
6230 |
@type to_skip: list
|
6231 |
@param to_skip: list of indices to skip
|
6232 |
@type target_node: string
|
6233 |
@param target_node: if passed, overrides the target node for creation
|
6234 |
@rtype: boolean
|
6235 |
@return: the success of the creation
|
6236 |
|
6237 |
"""
|
6238 |
info = _GetInstanceInfoText(instance) |
6239 |
if target_node is None: |
6240 |
pnode = instance.primary_node |
6241 |
all_nodes = instance.all_nodes |
6242 |
else:
|
6243 |
pnode = target_node |
6244 |
all_nodes = [pnode] |
6245 |
|
6246 |
if instance.disk_template == constants.DT_FILE:
|
6247 |
file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1]) |
6248 |
result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir) |
6249 |
|
6250 |
result.Raise("Failed to create directory '%s' on"
|
6251 |
" node %s" % (file_storage_dir, pnode))
|
6252 |
|
6253 |
# Note: this needs to be kept in sync with adding of disks in
|
6254 |
# LUSetInstanceParams
|
6255 |
for idx, device in enumerate(instance.disks): |
6256 |
if to_skip and idx in to_skip: |
6257 |
continue
|
6258 |
logging.info("Creating volume %s for instance %s",
|
6259 |
device.iv_name, instance.name) |
6260 |
#HARDCODE
|
6261 |
for node in all_nodes: |
6262 |
f_create = node == pnode |
6263 |
_CreateBlockDev(lu, node, instance, device, f_create, info, f_create) |
6264 |
|
6265 |
|
6266 |
def _RemoveDisks(lu, instance, target_node=None): |
6267 |
"""Remove all disks for an instance.
|
6268 |
|
6269 |
This abstracts away some work from `AddInstance()` and
|
6270 |
`RemoveInstance()`. Note that in case some of the devices couldn't
|
6271 |
be removed, the removal will continue with the other ones (compare
|
6272 |
with `_CreateDisks()`).
|
6273 |
|
6274 |
@type lu: L{LogicalUnit}
|
6275 |
@param lu: the logical unit on whose behalf we execute
|
6276 |
@type instance: L{objects.Instance}
|
6277 |
@param instance: the instance whose disks we should remove
|
6278 |
@type target_node: string
|
6279 |
@param target_node: used to override the node on which to remove the disks
|
6280 |
@rtype: boolean
|
6281 |
@return: the success of the removal
|
6282 |
|
6283 |
"""
|
6284 |
logging.info("Removing block devices for instance %s", instance.name)
|
6285 |
|
6286 |
all_result = True
|
6287 |
for device in instance.disks: |
6288 |
if target_node:
|
6289 |
edata = [(target_node, device)] |
6290 |
else:
|
6291 |
edata = device.ComputeNodeTree(instance.primary_node) |
6292 |
for node, disk in edata: |
6293 |
lu.cfg.SetDiskID(disk, node) |
6294 |
msg = lu.rpc.call_blockdev_remove(node, disk).fail_msg |
6295 |
if msg:
|
6296 |
lu.LogWarning("Could not remove block device %s on node %s,"
|
6297 |
" continuing anyway: %s", device.iv_name, node, msg)
|
6298 |
all_result = False
|
6299 |
|
6300 |
if instance.disk_template == constants.DT_FILE:
|
6301 |
file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1]) |
6302 |
if target_node:
|
6303 |
tgt = target_node |
6304 |
else:
|
6305 |
tgt = instance.primary_node |
6306 |
result = lu.rpc.call_file_storage_dir_remove(tgt, file_storage_dir) |
6307 |
if result.fail_msg:
|
6308 |
lu.LogWarning("Could not remove directory '%s' on node %s: %s",
|
6309 |
file_storage_dir, instance.primary_node, result.fail_msg) |
6310 |
all_result = False
|
6311 |
|
6312 |
return all_result
|
6313 |
|
6314 |
|
6315 |
def _ComputeDiskSize(disk_template, disks): |
6316 |
"""Compute disk size requirements in the volume group
|
6317 |
|
6318 |
"""
|
6319 |
# Required free disk space as a function of disk and swap space
|
6320 |
req_size_dict = { |
6321 |
constants.DT_DISKLESS: None,
|
6322 |
constants.DT_PLAIN: sum(d["size"] for d in disks), |
6323 |
# 128 MB are added for drbd metadata for each disk
|
6324 |
constants.DT_DRBD8: sum(d["size"] + 128 for d in disks), |
6325 |
constants.DT_FILE: None,
|
6326 |
} |
6327 |
|
6328 |
if disk_template not in req_size_dict: |
6329 |
raise errors.ProgrammerError("Disk template '%s' size requirement" |
6330 |
" is unknown" % disk_template)
|
6331 |
|
6332 |
return req_size_dict[disk_template]
|
6333 |
|
6334 |
|
6335 |
def _CheckHVParams(lu, nodenames, hvname, hvparams): |
6336 |
"""Hypervisor parameter validation.
|
6337 |
|
6338 |
This function abstract the hypervisor parameter validation to be
|
6339 |
used in both instance create and instance modify.
|
6340 |
|
6341 |
@type lu: L{LogicalUnit}
|
6342 |
@param lu: the logical unit for which we check
|
6343 |
@type nodenames: list
|
6344 |
@param nodenames: the list of nodes on which we should check
|
6345 |
@type hvname: string
|
6346 |
@param hvname: the name of the hypervisor we should use
|
6347 |
@type hvparams: dict
|
6348 |
@param hvparams: the parameters which we need to check
|
6349 |
@raise errors.OpPrereqError: if the parameters are not valid
|
6350 |
|
6351 |
"""
|
6352 |
hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames, |
6353 |
hvname, |
6354 |
hvparams) |
6355 |
for node in nodenames: |
6356 |
info = hvinfo[node] |
6357 |
if info.offline:
|
6358 |
continue
|
6359 |
info.Raise("Hypervisor parameter validation failed on node %s" % node)
|
6360 |
|
6361 |
|
6362 |
def _CheckOSParams(lu, required, nodenames, osname, osparams): |
6363 |
"""OS parameters validation.
|
6364 |
|
6365 |
@type lu: L{LogicalUnit}
|
6366 |
@param lu: the logical unit for which we check
|
6367 |
@type required: boolean
|
6368 |
@param required: whether the validation should fail if the OS is not
|
6369 |
found
|
6370 |
@type nodenames: list
|
6371 |
@param nodenames: the list of nodes on which we should check
|
6372 |
@type osname: string
|
6373 |
@param osname: the name of the hypervisor we should use
|
6374 |
@type osparams: dict
|
6375 |
@param osparams: the parameters which we need to check
|
6376 |
@raise errors.OpPrereqError: if the parameters are not valid
|
6377 |
|
6378 |
"""
|
6379 |
result = lu.rpc.call_os_validate(required, nodenames, osname, |
6380 |
[constants.OS_VALIDATE_PARAMETERS], |
6381 |
osparams) |
6382 |
for node, nres in result.items(): |
6383 |
# we don't check for offline cases since this should be run only
|
6384 |
# against the master node and/or an instance's nodes
|
6385 |
nres.Raise("OS Parameters validation failed on node %s" % node)
|
6386 |
if not nres.payload: |
6387 |
lu.LogInfo("OS %s not found on node %s, validation skipped",
|
6388 |
osname, node) |
6389 |
|
6390 |
|
6391 |
class LUCreateInstance(LogicalUnit): |
6392 |
"""Create an instance.
|
6393 |
|
6394 |
"""
|
6395 |
HPATH = "instance-add"
|
6396 |
HTYPE = constants.HTYPE_INSTANCE |
6397 |
_OP_PARAMS = [ |
6398 |
_PInstanceName, |
6399 |
("mode", _NoDefault, _TElemOf(constants.INSTANCE_CREATE_MODES)),
|
6400 |
("start", True, _TBool), |
6401 |
("wait_for_sync", True, _TBool), |
6402 |
("ip_check", True, _TBool), |
6403 |
("name_check", True, _TBool), |
6404 |
("disks", _NoDefault, _TListOf(_TDict)),
|
6405 |
("nics", _NoDefault, _TListOf(_TDict)),
|
6406 |
("hvparams", _EmptyDict, _TDict),
|
6407 |
("beparams", _EmptyDict, _TDict),
|
6408 |
("osparams", _EmptyDict, _TDict),
|
6409 |
("no_install", None, _TMaybeBool), |
6410 |
("os_type", None, _TMaybeString), |
6411 |
("force_variant", False, _TBool), |
6412 |
("source_handshake", None, _TOr(_TList, _TNone)), |
6413 |
("source_x509_ca", None, _TOr(_TList, _TNone)), |
6414 |
("source_instance_name", None, _TMaybeString), |
6415 |
("src_node", None, _TMaybeString), |
6416 |
("src_path", None, _TMaybeString), |
6417 |
("pnode", None, _TMaybeString), |
6418 |
("snode", None, _TMaybeString), |
6419 |
("iallocator", None, _TMaybeString), |
6420 |
("hypervisor", None, _TMaybeString), |
6421 |
("disk_template", _NoDefault, _CheckDiskTemplate),
|
6422 |
("identify_defaults", False, _TBool), |
6423 |
("file_driver", None, _TOr(_TNone, _TElemOf(constants.FILE_DRIVER))), |
6424 |
("file_storage_dir", None, _TMaybeString), |
6425 |
("dry_run", False, _TBool), |
6426 |
] |
6427 |
REQ_BGL = False
|
6428 |
|
6429 |
def CheckArguments(self): |
6430 |
"""Check arguments.
|
6431 |
|
6432 |
"""
|
6433 |
# do not require name_check to ease forward/backward compatibility
|
6434 |
# for tools
|
6435 |
if self.op.no_install and self.op.start: |
6436 |
self.LogInfo("No-installation mode selected, disabling startup") |
6437 |
self.op.start = False |
6438 |
# validate/normalize the instance name
|
6439 |
self.op.instance_name = utils.HostInfo.NormalizeName(self.op.instance_name) |
6440 |
if self.op.ip_check and not self.op.name_check: |
6441 |
# TODO: make the ip check more flexible and not depend on the name check
|
6442 |
raise errors.OpPrereqError("Cannot do ip checks without a name check", |
6443 |
errors.ECODE_INVAL) |
6444 |
|
6445 |
# check nics' parameter names
|
6446 |
for nic in self.op.nics: |
6447 |
utils.ForceDictType(nic, constants.INIC_PARAMS_TYPES) |
6448 |
|
6449 |
# check disks. parameter names and consistent adopt/no-adopt strategy
|
6450 |
has_adopt = has_no_adopt = False
|
6451 |
for disk in self.op.disks: |
6452 |
utils.ForceDictType(disk, constants.IDISK_PARAMS_TYPES) |
6453 |
if "adopt" in disk: |
6454 |
has_adopt = True
|
6455 |
else:
|
6456 |
has_no_adopt = True
|
6457 |
if has_adopt and has_no_adopt: |
6458 |
raise errors.OpPrereqError("Either all disks are adopted or none is", |
6459 |
errors.ECODE_INVAL) |
6460 |
if has_adopt:
|
6461 |
if self.op.disk_template not in constants.DTS_MAY_ADOPT: |
6462 |
raise errors.OpPrereqError("Disk adoption is not supported for the" |
6463 |
" '%s' disk template" %
|
6464 |
self.op.disk_template,
|
6465 |
errors.ECODE_INVAL) |
6466 |
if self.op.iallocator is not None: |
6467 |
raise errors.OpPrereqError("Disk adoption not allowed with an" |
6468 |
" iallocator script", errors.ECODE_INVAL)
|
6469 |
if self.op.mode == constants.INSTANCE_IMPORT: |
6470 |
raise errors.OpPrereqError("Disk adoption not allowed for" |
6471 |
" instance import", errors.ECODE_INVAL)
|
6472 |
|
6473 |
self.adopt_disks = has_adopt
|
6474 |
|
6475 |
# instance name verification
|
6476 |
if self.op.name_check: |
6477 |
self.hostname1 = utils.GetHostInfo(self.op.instance_name) |
6478 |
self.op.instance_name = self.hostname1.name |
6479 |
# used in CheckPrereq for ip ping check
|
6480 |
self.check_ip = self.hostname1.ip |
6481 |
elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT: |
6482 |
raise errors.OpPrereqError("Remote imports require names to be checked" % |
6483 |
errors.ECODE_INVAL) |
6484 |
else:
|
6485 |
self.check_ip = None |
6486 |
|
6487 |
# file storage checks
|
6488 |
if (self.op.file_driver and |
6489 |
not self.op.file_driver in constants.FILE_DRIVER): |
6490 |
raise errors.OpPrereqError("Invalid file driver name '%s'" % |
6491 |
self.op.file_driver, errors.ECODE_INVAL)
|
6492 |
|
6493 |
if self.op.file_storage_dir and os.path.isabs(self.op.file_storage_dir): |
6494 |
raise errors.OpPrereqError("File storage directory path not absolute", |
6495 |
errors.ECODE_INVAL) |
6496 |
|
6497 |
### Node/iallocator related checks
|
6498 |
if [self.op.iallocator, self.op.pnode].count(None) != 1: |
6499 |
raise errors.OpPrereqError("One and only one of iallocator and primary" |
6500 |
" node must be given",
|
6501 |
errors.ECODE_INVAL) |
6502 |
|
6503 |
self._cds = _GetClusterDomainSecret()
|
6504 |
|
6505 |
if self.op.mode == constants.INSTANCE_IMPORT: |
6506 |
# On import force_variant must be True, because if we forced it at
|
6507 |
# initial install, our only chance when importing it back is that it
|
6508 |
# works again!
|
6509 |
self.op.force_variant = True |
6510 |
|
6511 |
if self.op.no_install: |
6512 |
self.LogInfo("No-installation mode has no effect during import") |
6513 |
|
6514 |
elif self.op.mode == constants.INSTANCE_CREATE: |
6515 |
if self.op.os_type is None: |
6516 |
raise errors.OpPrereqError("No guest OS specified", |
6517 |
errors.ECODE_INVAL) |
6518 |
if self.op.disk_template is None: |
6519 |
raise errors.OpPrereqError("No disk template specified", |
6520 |
errors.ECODE_INVAL) |
6521 |
|
6522 |
elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT: |
6523 |
# Check handshake to ensure both clusters have the same domain secret
|
6524 |
src_handshake = self.op.source_handshake
|
6525 |
if not src_handshake: |
6526 |
raise errors.OpPrereqError("Missing source handshake", |
6527 |
errors.ECODE_INVAL) |
6528 |
|
6529 |
errmsg = masterd.instance.CheckRemoteExportHandshake(self._cds,
|
6530 |
src_handshake) |
6531 |
if errmsg:
|
6532 |
raise errors.OpPrereqError("Invalid handshake: %s" % errmsg, |
6533 |
errors.ECODE_INVAL) |
6534 |
|
6535 |
# Load and check source CA
|
6536 |
self.source_x509_ca_pem = self.op.source_x509_ca |
6537 |
if not self.source_x509_ca_pem: |
6538 |
raise errors.OpPrereqError("Missing source X509 CA", |
6539 |
errors.ECODE_INVAL) |
6540 |
|
6541 |
try:
|
6542 |
(cert, _) = utils.LoadSignedX509Certificate(self.source_x509_ca_pem,
|
6543 |
self._cds)
|
6544 |
except OpenSSL.crypto.Error, err:
|
6545 |
raise errors.OpPrereqError("Unable to load source X509 CA (%s)" % |
6546 |
(err, ), errors.ECODE_INVAL) |
6547 |
|
6548 |
(errcode, msg) = utils.VerifyX509Certificate(cert, None, None) |
6549 |
if errcode is not None: |
6550 |
raise errors.OpPrereqError("Invalid source X509 CA (%s)" % (msg, ), |
6551 |
errors.ECODE_INVAL) |
6552 |
|
6553 |
self.source_x509_ca = cert
|
6554 |
|
6555 |
src_instance_name = self.op.source_instance_name
|
6556 |
if not src_instance_name: |
6557 |
raise errors.OpPrereqError("Missing source instance name", |
6558 |
errors.ECODE_INVAL) |
6559 |
|
6560 |
self.source_instance_name = \
|
6561 |
utils.GetHostInfo(utils.HostInfo.NormalizeName(src_instance_name)).name |
6562 |
|
6563 |
else:
|
6564 |
raise errors.OpPrereqError("Invalid instance creation mode %r" % |
6565 |
self.op.mode, errors.ECODE_INVAL)
|
6566 |
|
6567 |
def ExpandNames(self): |
6568 |
"""ExpandNames for CreateInstance.
|
6569 |
|
6570 |
Figure out the right locks for instance creation.
|
6571 |
|
6572 |
"""
|
6573 |
self.needed_locks = {}
|
6574 |
|
6575 |
instance_name = self.op.instance_name
|
6576 |
# this is just a preventive check, but someone might still add this
|
6577 |
# instance in the meantime, and creation will fail at lock-add time
|
6578 |
if instance_name in self.cfg.GetInstanceList(): |
6579 |
raise errors.OpPrereqError("Instance '%s' is already in the cluster" % |
6580 |
instance_name, errors.ECODE_EXISTS) |
6581 |
|
6582 |
self.add_locks[locking.LEVEL_INSTANCE] = instance_name
|
6583 |
|
6584 |
if self.op.iallocator: |
6585 |
self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
|
6586 |
else:
|
6587 |
self.op.pnode = _ExpandNodeName(self.cfg, self.op.pnode) |
6588 |
nodelist = [self.op.pnode]
|
6589 |
if self.op.snode is not None: |
6590 |
self.op.snode = _ExpandNodeName(self.cfg, self.op.snode) |
6591 |
nodelist.append(self.op.snode)
|
6592 |
self.needed_locks[locking.LEVEL_NODE] = nodelist
|
6593 |
|
6594 |
# in case of import lock the source node too
|
6595 |
if self.op.mode == constants.INSTANCE_IMPORT: |
6596 |
src_node = self.op.src_node
|
6597 |
src_path = self.op.src_path
|
6598 |
|
6599 |
if src_path is None: |
6600 |
self.op.src_path = src_path = self.op.instance_name |
6601 |
|
6602 |
if src_node is None: |
6603 |
self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
|
6604 |
self.op.src_node = None |
6605 |
if os.path.isabs(src_path):
|
6606 |
raise errors.OpPrereqError("Importing an instance from an absolute" |
6607 |
" path requires a source node option.",
|
6608 |
errors.ECODE_INVAL) |
6609 |
else:
|
6610 |
self.op.src_node = src_node = _ExpandNodeName(self.cfg, src_node) |
6611 |
if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET: |
6612 |
self.needed_locks[locking.LEVEL_NODE].append(src_node)
|
6613 |
if not os.path.isabs(src_path): |
6614 |
self.op.src_path = src_path = \
|
6615 |
utils.PathJoin(constants.EXPORT_DIR, src_path) |
6616 |
|
6617 |
def _RunAllocator(self): |
6618 |
"""Run the allocator based on input opcode.
|
6619 |
|
6620 |
"""
|
6621 |
nics = [n.ToDict() for n in self.nics] |
6622 |
ial = IAllocator(self.cfg, self.rpc, |
6623 |
mode=constants.IALLOCATOR_MODE_ALLOC, |
6624 |
name=self.op.instance_name,
|
6625 |
disk_template=self.op.disk_template,
|
6626 |
tags=[], |
6627 |
os=self.op.os_type,
|
6628 |
vcpus=self.be_full[constants.BE_VCPUS],
|
6629 |
mem_size=self.be_full[constants.BE_MEMORY],
|
6630 |
disks=self.disks,
|
6631 |
nics=nics, |
6632 |
hypervisor=self.op.hypervisor,
|
6633 |
) |
6634 |
|
6635 |
ial.Run(self.op.iallocator)
|
6636 |
|
6637 |
if not ial.success: |
6638 |
raise errors.OpPrereqError("Can't compute nodes using" |
6639 |
" iallocator '%s': %s" %
|
6640 |
(self.op.iallocator, ial.info),
|
6641 |
errors.ECODE_NORES) |
6642 |
if len(ial.result) != ial.required_nodes: |
6643 |
raise errors.OpPrereqError("iallocator '%s' returned invalid number" |
6644 |
" of nodes (%s), required %s" %
|
6645 |
(self.op.iallocator, len(ial.result), |
6646 |
ial.required_nodes), errors.ECODE_FAULT) |
6647 |
self.op.pnode = ial.result[0] |
6648 |
self.LogInfo("Selected nodes for instance %s via iallocator %s: %s", |
6649 |
self.op.instance_name, self.op.iallocator, |
6650 |
utils.CommaJoin(ial.result)) |
6651 |
if ial.required_nodes == 2: |
6652 |
self.op.snode = ial.result[1] |
6653 |
|
6654 |
def BuildHooksEnv(self): |
6655 |
"""Build hooks env.
|
6656 |
|
6657 |
This runs on master, primary and secondary nodes of the instance.
|
6658 |
|
6659 |
"""
|
6660 |
env = { |
6661 |
"ADD_MODE": self.op.mode, |
6662 |
} |
6663 |
if self.op.mode == constants.INSTANCE_IMPORT: |
6664 |
env["SRC_NODE"] = self.op.src_node |
6665 |
env["SRC_PATH"] = self.op.src_path |
6666 |
env["SRC_IMAGES"] = self.src_images |
6667 |
|
6668 |
env.update(_BuildInstanceHookEnv( |
6669 |
name=self.op.instance_name,
|
6670 |
primary_node=self.op.pnode,
|
6671 |
secondary_nodes=self.secondaries,
|
6672 |
status=self.op.start,
|
6673 |
os_type=self.op.os_type,
|
6674 |
memory=self.be_full[constants.BE_MEMORY],
|
6675 |
vcpus=self.be_full[constants.BE_VCPUS],
|
6676 |
nics=_NICListToTuple(self, self.nics), |
6677 |
disk_template=self.op.disk_template,
|
6678 |
disks=[(d["size"], d["mode"]) for d in self.disks], |
6679 |
bep=self.be_full,
|
6680 |
hvp=self.hv_full,
|
6681 |
hypervisor_name=self.op.hypervisor,
|
6682 |
)) |
6683 |
|
6684 |
nl = ([self.cfg.GetMasterNode(), self.op.pnode] + |
6685 |
self.secondaries)
|
6686 |
return env, nl, nl
|
6687 |
|
6688 |
def _ReadExportInfo(self): |
6689 |
"""Reads the export information from disk.
|
6690 |
|
6691 |
It will override the opcode source node and path with the actual
|
6692 |
information, if these two were not specified before.
|
6693 |
|
6694 |
@return: the export information
|
6695 |
|
6696 |
"""
|
6697 |
assert self.op.mode == constants.INSTANCE_IMPORT |
6698 |
|
6699 |
src_node = self.op.src_node
|
6700 |
src_path = self.op.src_path
|
6701 |
|
6702 |
if src_node is None: |
6703 |
locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
|
6704 |
exp_list = self.rpc.call_export_list(locked_nodes)
|
6705 |
found = False
|
6706 |
for node in exp_list: |
6707 |
if exp_list[node].fail_msg:
|
6708 |
continue
|
6709 |
if src_path in exp_list[node].payload: |
6710 |
found = True
|
6711 |
self.op.src_node = src_node = node
|
6712 |
self.op.src_path = src_path = utils.PathJoin(constants.EXPORT_DIR,
|
6713 |
src_path) |
6714 |
break
|
6715 |
if not found: |
6716 |
raise errors.OpPrereqError("No export found for relative path %s" % |
6717 |
src_path, errors.ECODE_INVAL) |
6718 |
|
6719 |
_CheckNodeOnline(self, src_node)
|
6720 |
result = self.rpc.call_export_info(src_node, src_path)
|
6721 |
result.Raise("No export or invalid export found in dir %s" % src_path)
|
6722 |
|
6723 |
export_info = objects.SerializableConfigParser.Loads(str(result.payload))
|
6724 |
if not export_info.has_section(constants.INISECT_EXP): |
6725 |
raise errors.ProgrammerError("Corrupted export config", |
6726 |
errors.ECODE_ENVIRON) |
6727 |
|
6728 |
ei_version = export_info.get(constants.INISECT_EXP, "version")
|
6729 |
if (int(ei_version) != constants.EXPORT_VERSION): |
6730 |
raise errors.OpPrereqError("Wrong export version %s (wanted %d)" % |
6731 |
(ei_version, constants.EXPORT_VERSION), |
6732 |
errors.ECODE_ENVIRON) |
6733 |
return export_info
|
6734 |
|
6735 |
def _ReadExportParams(self, einfo): |
6736 |
"""Use export parameters as defaults.
|
6737 |
|
6738 |
In case the opcode doesn't specify (as in override) some instance
|
6739 |
parameters, then try to use them from the export information, if
|
6740 |
that declares them.
|
6741 |
|
6742 |
"""
|
6743 |
self.op.os_type = einfo.get(constants.INISECT_EXP, "os") |
6744 |
|
6745 |
if self.op.disk_template is None: |
6746 |
if einfo.has_option(constants.INISECT_INS, "disk_template"): |
6747 |
self.op.disk_template = einfo.get(constants.INISECT_INS,
|
6748 |
"disk_template")
|
6749 |
else:
|
6750 |
raise errors.OpPrereqError("No disk template specified and the export" |
6751 |
" is missing the disk_template information",
|
6752 |
errors.ECODE_INVAL) |
6753 |
|
6754 |
if not self.op.disks: |
6755 |
if einfo.has_option(constants.INISECT_INS, "disk_count"): |
6756 |
disks = [] |
6757 |
# TODO: import the disk iv_name too
|
6758 |
for idx in range(einfo.getint(constants.INISECT_INS, "disk_count")): |
6759 |
disk_sz = einfo.getint(constants.INISECT_INS, "disk%d_size" % idx)
|
6760 |
disks.append({"size": disk_sz})
|
6761 |
self.op.disks = disks
|
6762 |
else:
|
6763 |
raise errors.OpPrereqError("No disk info specified and the export" |
6764 |
" is missing the disk information",
|
6765 |
errors.ECODE_INVAL) |
6766 |
|
6767 |
if (not self.op.nics and |
6768 |
einfo.has_option(constants.INISECT_INS, "nic_count")):
|
6769 |
nics = [] |
6770 |
for idx in range(einfo.getint(constants.INISECT_INS, "nic_count")): |
6771 |
ndict = {} |
6772 |
for name in list(constants.NICS_PARAMETERS) + ["ip", "mac"]: |
6773 |
v = einfo.get(constants.INISECT_INS, "nic%d_%s" % (idx, name))
|
6774 |
ndict[name] = v |
6775 |
nics.append(ndict) |
6776 |
self.op.nics = nics
|
6777 |
|
6778 |
if (self.op.hypervisor is None and |
6779 |
einfo.has_option(constants.INISECT_INS, "hypervisor")):
|
6780 |
self.op.hypervisor = einfo.get(constants.INISECT_INS, "hypervisor") |
6781 |
if einfo.has_section(constants.INISECT_HYP):
|
6782 |
# use the export parameters but do not override the ones
|
6783 |
# specified by the user
|
6784 |
for name, value in einfo.items(constants.INISECT_HYP): |
6785 |
if name not in self.op.hvparams: |
6786 |
self.op.hvparams[name] = value
|
6787 |
|
6788 |
if einfo.has_section(constants.INISECT_BEP):
|
6789 |
# use the parameters, without overriding
|
6790 |
for name, value in einfo.items(constants.INISECT_BEP): |
6791 |
if name not in self.op.beparams: |
6792 |
self.op.beparams[name] = value
|
6793 |
else:
|
6794 |
# try to read the parameters old style, from the main section
|
6795 |
for name in constants.BES_PARAMETERS: |
6796 |
if (name not in self.op.beparams and |
6797 |
einfo.has_option(constants.INISECT_INS, name)): |
6798 |
self.op.beparams[name] = einfo.get(constants.INISECT_INS, name)
|
6799 |
|
6800 |
if einfo.has_section(constants.INISECT_OSP):
|
6801 |
# use the parameters, without overriding
|
6802 |
for name, value in einfo.items(constants.INISECT_OSP): |
6803 |
if name not in self.op.osparams: |
6804 |
self.op.osparams[name] = value
|
6805 |
|
6806 |
def _RevertToDefaults(self, cluster): |
6807 |
"""Revert the instance parameters to the default values.
|
6808 |
|
6809 |
"""
|
6810 |
# hvparams
|
6811 |
hv_defs = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type, {}) |
6812 |
for name in self.op.hvparams.keys(): |
6813 |
if name in hv_defs and hv_defs[name] == self.op.hvparams[name]: |
6814 |
del self.op.hvparams[name] |
6815 |
# beparams
|
6816 |
be_defs = cluster.SimpleFillBE({}) |
6817 |
for name in self.op.beparams.keys(): |
6818 |
if name in be_defs and be_defs[name] == self.op.beparams[name]: |
6819 |
del self.op.beparams[name] |
6820 |
# nic params
|
6821 |
nic_defs = cluster.SimpleFillNIC({}) |
6822 |
for nic in self.op.nics: |
6823 |
for name in constants.NICS_PARAMETERS: |
6824 |
if name in nic and name in nic_defs and nic[name] == nic_defs[name]: |
6825 |
del nic[name]
|
6826 |
# osparams
|
6827 |
os_defs = cluster.SimpleFillOS(self.op.os_type, {})
|
6828 |
for name in self.op.osparams.keys(): |
6829 |
if name in os_defs and os_defs[name] == self.op.osparams[name]: |
6830 |
del self.op.osparams[name] |
6831 |
|
6832 |
def CheckPrereq(self): |
6833 |
"""Check prerequisites.
|
6834 |
|
6835 |
"""
|
6836 |
if self.op.mode == constants.INSTANCE_IMPORT: |
6837 |
export_info = self._ReadExportInfo()
|
6838 |
self._ReadExportParams(export_info)
|
6839 |
|
6840 |
_CheckDiskTemplate(self.op.disk_template)
|
6841 |
|
6842 |
if (not self.cfg.GetVGName() and |
6843 |
self.op.disk_template not in constants.DTS_NOT_LVM): |
6844 |
raise errors.OpPrereqError("Cluster does not support lvm-based" |
6845 |
" instances", errors.ECODE_STATE)
|
6846 |
|
6847 |
if self.op.hypervisor is None: |
6848 |
self.op.hypervisor = self.cfg.GetHypervisorType() |
6849 |
|
6850 |
cluster = self.cfg.GetClusterInfo()
|
6851 |
enabled_hvs = cluster.enabled_hypervisors |
6852 |
if self.op.hypervisor not in enabled_hvs: |
6853 |
raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the" |
6854 |
" cluster (%s)" % (self.op.hypervisor, |
6855 |
",".join(enabled_hvs)),
|
6856 |
errors.ECODE_STATE) |
6857 |
|
6858 |
# check hypervisor parameter syntax (locally)
|
6859 |
utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
|
6860 |
filled_hvp = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type, |
6861 |
self.op.hvparams)
|
6862 |
hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
|
6863 |
hv_type.CheckParameterSyntax(filled_hvp) |
6864 |
self.hv_full = filled_hvp
|
6865 |
# check that we don't specify global parameters on an instance
|
6866 |
_CheckGlobalHvParams(self.op.hvparams)
|
6867 |
|
6868 |
# fill and remember the beparams dict
|
6869 |
utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
|
6870 |
self.be_full = cluster.SimpleFillBE(self.op.beparams) |
6871 |
|
6872 |
# build os parameters
|
6873 |
self.os_full = cluster.SimpleFillOS(self.op.os_type, self.op.osparams) |
6874 |
|
6875 |
# now that hvp/bep are in final format, let's reset to defaults,
|
6876 |
# if told to do so
|
6877 |
if self.op.identify_defaults: |
6878 |
self._RevertToDefaults(cluster)
|
6879 |
|
6880 |
# NIC buildup
|
6881 |
self.nics = []
|
6882 |
for idx, nic in enumerate(self.op.nics): |
6883 |
nic_mode_req = nic.get("mode", None) |
6884 |
nic_mode = nic_mode_req |
6885 |
if nic_mode is None: |
6886 |
nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE] |
6887 |
|
6888 |
# in routed mode, for the first nic, the default ip is 'auto'
|
6889 |
if nic_mode == constants.NIC_MODE_ROUTED and idx == 0: |
6890 |
default_ip_mode = constants.VALUE_AUTO |
6891 |
else:
|
6892 |
default_ip_mode = constants.VALUE_NONE |
6893 |
|
6894 |
# ip validity checks
|
6895 |
ip = nic.get("ip", default_ip_mode)
|
6896 |
if ip is None or ip.lower() == constants.VALUE_NONE: |
6897 |
nic_ip = None
|
6898 |
elif ip.lower() == constants.VALUE_AUTO:
|
6899 |
if not self.op.name_check: |
6900 |
raise errors.OpPrereqError("IP address set to auto but name checks" |
6901 |
" have been skipped. Aborting.",
|
6902 |
errors.ECODE_INVAL) |
6903 |
nic_ip = self.hostname1.ip
|
6904 |
else:
|
6905 |
if not utils.IsValidIP4(ip): |
6906 |
raise errors.OpPrereqError("Given IP address '%s' doesn't look" |
6907 |
" like a valid IP" % ip,
|
6908 |
errors.ECODE_INVAL) |
6909 |
nic_ip = ip |
6910 |
|
6911 |
# TODO: check the ip address for uniqueness
|
6912 |
if nic_mode == constants.NIC_MODE_ROUTED and not nic_ip: |
6913 |
raise errors.OpPrereqError("Routed nic mode requires an ip address", |
6914 |
errors.ECODE_INVAL) |
6915 |
|
6916 |
# MAC address verification
|
6917 |
mac = nic.get("mac", constants.VALUE_AUTO)
|
6918 |
if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE): |
6919 |
mac = utils.NormalizeAndValidateMac(mac) |
6920 |
|
6921 |
try:
|
6922 |
self.cfg.ReserveMAC(mac, self.proc.GetECId()) |
6923 |
except errors.ReservationError:
|
6924 |
raise errors.OpPrereqError("MAC address %s already in use" |
6925 |
" in cluster" % mac,
|
6926 |
errors.ECODE_NOTUNIQUE) |
6927 |
|
6928 |
# bridge verification
|
6929 |
bridge = nic.get("bridge", None) |
6930 |
link = nic.get("link", None) |
6931 |
if bridge and link: |
6932 |
raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'" |
6933 |
" at the same time", errors.ECODE_INVAL)
|
6934 |
elif bridge and nic_mode == constants.NIC_MODE_ROUTED: |
6935 |
raise errors.OpPrereqError("Cannot pass 'bridge' on a routed nic", |
6936 |
errors.ECODE_INVAL) |
6937 |
elif bridge:
|
6938 |
link = bridge |
6939 |
|
6940 |
nicparams = {} |
6941 |
if nic_mode_req:
|
6942 |
nicparams[constants.NIC_MODE] = nic_mode_req |
6943 |
if link:
|
6944 |
nicparams[constants.NIC_LINK] = link |
6945 |
|
6946 |
check_params = cluster.SimpleFillNIC(nicparams) |
6947 |
objects.NIC.CheckParameterSyntax(check_params) |
6948 |
self.nics.append(objects.NIC(mac=mac, ip=nic_ip, nicparams=nicparams))
|
6949 |
|
6950 |
# disk checks/pre-build
|
6951 |
self.disks = []
|
6952 |
for disk in self.op.disks: |
6953 |
mode = disk.get("mode", constants.DISK_RDWR)
|
6954 |
if mode not in constants.DISK_ACCESS_SET: |
6955 |
raise errors.OpPrereqError("Invalid disk access mode '%s'" % |
6956 |
mode, errors.ECODE_INVAL) |
6957 |
size = disk.get("size", None) |
6958 |
if size is None: |
6959 |
raise errors.OpPrereqError("Missing disk size", errors.ECODE_INVAL) |
6960 |
try:
|
6961 |
size = int(size)
|
6962 |
except (TypeError, ValueError): |
6963 |
raise errors.OpPrereqError("Invalid disk size '%s'" % size, |
6964 |
errors.ECODE_INVAL) |
6965 |
new_disk = {"size": size, "mode": mode} |
6966 |
if "adopt" in disk: |
6967 |
new_disk["adopt"] = disk["adopt"] |
6968 |
self.disks.append(new_disk)
|
6969 |
|
6970 |
if self.op.mode == constants.INSTANCE_IMPORT: |
6971 |
|
6972 |
# Check that the new instance doesn't have less disks than the export
|
6973 |
instance_disks = len(self.disks) |
6974 |
export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
|
6975 |
if instance_disks < export_disks:
|
6976 |
raise errors.OpPrereqError("Not enough disks to import." |
6977 |
" (instance: %d, export: %d)" %
|
6978 |
(instance_disks, export_disks), |
6979 |
errors.ECODE_INVAL) |
6980 |
|
6981 |
disk_images = [] |
6982 |
for idx in range(export_disks): |
6983 |
option = 'disk%d_dump' % idx
|
6984 |
if export_info.has_option(constants.INISECT_INS, option):
|
6985 |
# FIXME: are the old os-es, disk sizes, etc. useful?
|
6986 |
export_name = export_info.get(constants.INISECT_INS, option) |
6987 |
image = utils.PathJoin(self.op.src_path, export_name)
|
6988 |
disk_images.append(image) |
6989 |
else:
|
6990 |
disk_images.append(False)
|
6991 |
|
6992 |
self.src_images = disk_images
|
6993 |
|
6994 |
old_name = export_info.get(constants.INISECT_INS, 'name')
|
6995 |
try:
|
6996 |
exp_nic_count = export_info.getint(constants.INISECT_INS, 'nic_count')
|
6997 |
except (TypeError, ValueError), err: |
6998 |
raise errors.OpPrereqError("Invalid export file, nic_count is not" |
6999 |
" an integer: %s" % str(err), |
7000 |
errors.ECODE_STATE) |
7001 |
if self.op.instance_name == old_name: |
7002 |
for idx, nic in enumerate(self.nics): |
7003 |
if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx: |
7004 |
nic_mac_ini = 'nic%d_mac' % idx
|
7005 |
nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini) |
7006 |
|
7007 |
# ENDIF: self.op.mode == constants.INSTANCE_IMPORT
|
7008 |
|
7009 |
# ip ping checks (we use the same ip that was resolved in ExpandNames)
|
7010 |
if self.op.ip_check: |
7011 |
if utils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT): |
7012 |
raise errors.OpPrereqError("IP %s of instance %s already in use" % |
7013 |
(self.check_ip, self.op.instance_name), |
7014 |
errors.ECODE_NOTUNIQUE) |
7015 |
|
7016 |
#### mac address generation
|
7017 |
# By generating here the mac address both the allocator and the hooks get
|
7018 |
# the real final mac address rather than the 'auto' or 'generate' value.
|
7019 |
# There is a race condition between the generation and the instance object
|
7020 |
# creation, which means that we know the mac is valid now, but we're not
|
7021 |
# sure it will be when we actually add the instance. If things go bad
|
7022 |
# adding the instance will abort because of a duplicate mac, and the
|
7023 |
# creation job will fail.
|
7024 |
for nic in self.nics: |
7025 |
if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE): |
7026 |
nic.mac = self.cfg.GenerateMAC(self.proc.GetECId()) |
7027 |
|
7028 |
#### allocator run
|
7029 |
|
7030 |
if self.op.iallocator is not None: |
7031 |
self._RunAllocator()
|
7032 |
|
7033 |
#### node related checks
|
7034 |
|
7035 |
# check primary node
|
7036 |
self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode) |
7037 |
assert self.pnode is not None, \ |
7038 |
"Cannot retrieve locked node %s" % self.op.pnode |
7039 |
if pnode.offline:
|
7040 |
raise errors.OpPrereqError("Cannot use offline primary node '%s'" % |
7041 |
pnode.name, errors.ECODE_STATE) |
7042 |
if pnode.drained:
|
7043 |
raise errors.OpPrereqError("Cannot use drained primary node '%s'" % |
7044 |
pnode.name, errors.ECODE_STATE) |
7045 |
|
7046 |
self.secondaries = []
|
7047 |
|
7048 |
# mirror node verification
|
7049 |
if self.op.disk_template in constants.DTS_NET_MIRROR: |
7050 |
if self.op.snode is None: |
7051 |
raise errors.OpPrereqError("The networked disk templates need" |
7052 |
" a mirror node", errors.ECODE_INVAL)
|
7053 |
if self.op.snode == pnode.name: |
7054 |
raise errors.OpPrereqError("The secondary node cannot be the" |
7055 |
" primary node.", errors.ECODE_INVAL)
|
7056 |
_CheckNodeOnline(self, self.op.snode) |
7057 |
_CheckNodeNotDrained(self, self.op.snode) |
7058 |
self.secondaries.append(self.op.snode) |
7059 |
|
7060 |
nodenames = [pnode.name] + self.secondaries
|
7061 |
|
7062 |
req_size = _ComputeDiskSize(self.op.disk_template,
|
7063 |
self.disks)
|
7064 |
|
7065 |
# Check lv size requirements, if not adopting
|
7066 |
if req_size is not None and not self.adopt_disks: |
7067 |
_CheckNodesFreeDisk(self, nodenames, req_size)
|
7068 |
|
7069 |
if self.adopt_disks: # instead, we must check the adoption data |
7070 |
all_lvs = set([i["adopt"] for i in self.disks]) |
7071 |
if len(all_lvs) != len(self.disks): |
7072 |
raise errors.OpPrereqError("Duplicate volume names given for adoption", |
7073 |
errors.ECODE_INVAL) |
7074 |
for lv_name in all_lvs: |
7075 |
try:
|
7076 |
self.cfg.ReserveLV(lv_name, self.proc.GetECId()) |
7077 |
except errors.ReservationError:
|
7078 |
raise errors.OpPrereqError("LV named %s used by another instance" % |
7079 |
lv_name, errors.ECODE_NOTUNIQUE) |
7080 |
|
7081 |
node_lvs = self.rpc.call_lv_list([pnode.name],
|
7082 |
self.cfg.GetVGName())[pnode.name]
|
7083 |
node_lvs.Raise("Cannot get LV information from node %s" % pnode.name)
|
7084 |
node_lvs = node_lvs.payload |
7085 |
delta = all_lvs.difference(node_lvs.keys()) |
7086 |
if delta:
|
7087 |
raise errors.OpPrereqError("Missing logical volume(s): %s" % |
7088 |
utils.CommaJoin(delta), |
7089 |
errors.ECODE_INVAL) |
7090 |
online_lvs = [lv for lv in all_lvs if node_lvs[lv][2]] |
7091 |
if online_lvs:
|
7092 |
raise errors.OpPrereqError("Online logical volumes found, cannot" |
7093 |
" adopt: %s" % utils.CommaJoin(online_lvs),
|
7094 |
errors.ECODE_STATE) |
7095 |
# update the size of disk based on what is found
|
7096 |
for dsk in self.disks: |
7097 |
dsk["size"] = int(float(node_lvs[dsk["adopt"]][0])) |
7098 |
|
7099 |
_CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams) |
7100 |
|
7101 |
_CheckNodeHasOS(self, pnode.name, self.op.os_type, self.op.force_variant) |
7102 |
# check OS parameters (remotely)
|
7103 |
_CheckOSParams(self, True, nodenames, self.op.os_type, self.os_full) |
7104 |
|
7105 |
_CheckNicsBridgesExist(self, self.nics, self.pnode.name) |
7106 |
|
7107 |
# memory check on primary node
|
7108 |
if self.op.start: |
7109 |
_CheckNodeFreeMemory(self, self.pnode.name, |
7110 |
"creating instance %s" % self.op.instance_name, |
7111 |
self.be_full[constants.BE_MEMORY],
|
7112 |
self.op.hypervisor)
|
7113 |
|
7114 |
self.dry_run_result = list(nodenames) |
7115 |
|
7116 |
def Exec(self, feedback_fn): |
7117 |
"""Create and add the instance to the cluster.
|
7118 |
|
7119 |
"""
|
7120 |
instance = self.op.instance_name
|
7121 |
pnode_name = self.pnode.name
|
7122 |
|
7123 |
ht_kind = self.op.hypervisor
|
7124 |
if ht_kind in constants.HTS_REQ_PORT: |
7125 |
network_port = self.cfg.AllocatePort()
|
7126 |
else:
|
7127 |
network_port = None
|
7128 |
|
7129 |
if constants.ENABLE_FILE_STORAGE:
|
7130 |
# this is needed because os.path.join does not accept None arguments
|
7131 |
if self.op.file_storage_dir is None: |
7132 |
string_file_storage_dir = ""
|
7133 |
else:
|
7134 |
string_file_storage_dir = self.op.file_storage_dir
|
7135 |
|
7136 |
# build the full file storage dir path
|
7137 |
file_storage_dir = utils.PathJoin(self.cfg.GetFileStorageDir(),
|
7138 |
string_file_storage_dir, instance) |
7139 |
else:
|
7140 |
file_storage_dir = ""
|
7141 |
|
7142 |
disks = _GenerateDiskTemplate(self,
|
7143 |
self.op.disk_template,
|
7144 |
instance, pnode_name, |
7145 |
self.secondaries,
|
7146 |
self.disks,
|
7147 |
file_storage_dir, |
7148 |
self.op.file_driver,
|
7149 |
0)
|
7150 |
|
7151 |
iobj = objects.Instance(name=instance, os=self.op.os_type,
|
7152 |
primary_node=pnode_name, |
7153 |
nics=self.nics, disks=disks,
|
7154 |
disk_template=self.op.disk_template,
|
7155 |
admin_up=False,
|
7156 |
network_port=network_port, |
7157 |
beparams=self.op.beparams,
|
7158 |
hvparams=self.op.hvparams,
|
7159 |
hypervisor=self.op.hypervisor,
|
7160 |
osparams=self.op.osparams,
|
7161 |
) |
7162 |
|
7163 |
if self.adopt_disks: |
7164 |
# rename LVs to the newly-generated names; we need to construct
|
7165 |
# 'fake' LV disks with the old data, plus the new unique_id
|
7166 |
tmp_disks = [objects.Disk.FromDict(v.ToDict()) for v in disks] |
7167 |
rename_to = [] |
7168 |
for t_dsk, a_dsk in zip (tmp_disks, self.disks): |
7169 |
rename_to.append(t_dsk.logical_id) |
7170 |
t_dsk.logical_id = (t_dsk.logical_id[0], a_dsk["adopt"]) |
7171 |
self.cfg.SetDiskID(t_dsk, pnode_name)
|
7172 |
result = self.rpc.call_blockdev_rename(pnode_name,
|
7173 |
zip(tmp_disks, rename_to))
|
7174 |
result.Raise("Failed to rename adoped LVs")
|
7175 |
else:
|
7176 |
feedback_fn("* creating instance disks...")
|
7177 |
try:
|
7178 |
_CreateDisks(self, iobj)
|
7179 |
except errors.OpExecError:
|
7180 |
self.LogWarning("Device creation failed, reverting...") |
7181 |
try:
|
7182 |
_RemoveDisks(self, iobj)
|
7183 |
finally:
|
7184 |
self.cfg.ReleaseDRBDMinors(instance)
|
7185 |
raise
|
7186 |
|
7187 |
feedback_fn("adding instance %s to cluster config" % instance)
|
7188 |
|
7189 |
self.cfg.AddInstance(iobj, self.proc.GetECId()) |
7190 |
|
7191 |
# Declare that we don't want to remove the instance lock anymore, as we've
|
7192 |
# added the instance to the config
|
7193 |
del self.remove_locks[locking.LEVEL_INSTANCE] |
7194 |
# Unlock all the nodes
|
7195 |
if self.op.mode == constants.INSTANCE_IMPORT: |
7196 |
nodes_keep = [self.op.src_node]
|
7197 |
nodes_release = [node for node in self.acquired_locks[locking.LEVEL_NODE] |
7198 |
if node != self.op.src_node] |
7199 |
self.context.glm.release(locking.LEVEL_NODE, nodes_release)
|
7200 |
self.acquired_locks[locking.LEVEL_NODE] = nodes_keep
|
7201 |
else:
|
7202 |
self.context.glm.release(locking.LEVEL_NODE)
|
7203 |
del self.acquired_locks[locking.LEVEL_NODE] |
7204 |
|
7205 |
if self.op.wait_for_sync: |
7206 |
disk_abort = not _WaitForSync(self, iobj) |
7207 |
elif iobj.disk_template in constants.DTS_NET_MIRROR: |
7208 |
# make sure the disks are not degraded (still sync-ing is ok)
|
7209 |
time.sleep(15)
|
7210 |
feedback_fn("* checking mirrors status")
|
7211 |
disk_abort = not _WaitForSync(self, iobj, oneshot=True) |
7212 |
else:
|
7213 |
disk_abort = False
|
7214 |
|
7215 |
if disk_abort:
|
7216 |
_RemoveDisks(self, iobj)
|
7217 |
self.cfg.RemoveInstance(iobj.name)
|
7218 |
# Make sure the instance lock gets removed
|
7219 |
self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
|
7220 |
raise errors.OpExecError("There are some degraded disks for" |
7221 |
" this instance")
|
7222 |
|
7223 |
if iobj.disk_template != constants.DT_DISKLESS and not self.adopt_disks: |
7224 |
if self.op.mode == constants.INSTANCE_CREATE: |
7225 |
if not self.op.no_install: |
7226 |
feedback_fn("* running the instance OS create scripts...")
|
7227 |
# FIXME: pass debug option from opcode to backend
|
7228 |
result = self.rpc.call_instance_os_add(pnode_name, iobj, False, |
7229 |
self.op.debug_level)
|
7230 |
result.Raise("Could not add os for instance %s"
|
7231 |
" on node %s" % (instance, pnode_name))
|
7232 |
|
7233 |
elif self.op.mode == constants.INSTANCE_IMPORT: |
7234 |
feedback_fn("* running the instance OS import scripts...")
|
7235 |
|
7236 |
transfers = [] |
7237 |
|
7238 |
for idx, image in enumerate(self.src_images): |
7239 |
if not image: |
7240 |
continue
|
7241 |
|
7242 |
# FIXME: pass debug option from opcode to backend
|
7243 |
dt = masterd.instance.DiskTransfer("disk/%s" % idx,
|
7244 |
constants.IEIO_FILE, (image, ), |
7245 |
constants.IEIO_SCRIPT, |
7246 |
(iobj.disks[idx], idx), |
7247 |
None)
|
7248 |
transfers.append(dt) |
7249 |
|
7250 |
import_result = \ |
7251 |
masterd.instance.TransferInstanceData(self, feedback_fn,
|
7252 |
self.op.src_node, pnode_name,
|
7253 |
self.pnode.secondary_ip,
|
7254 |
iobj, transfers) |
7255 |
if not compat.all(import_result): |
7256 |
self.LogWarning("Some disks for instance %s on node %s were not" |
7257 |
" imported successfully" % (instance, pnode_name))
|
7258 |
|
7259 |
elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT: |
7260 |
feedback_fn("* preparing remote import...")
|
7261 |
connect_timeout = constants.RIE_CONNECT_TIMEOUT |
7262 |
timeouts = masterd.instance.ImportExportTimeouts(connect_timeout) |
7263 |
|
7264 |
disk_results = masterd.instance.RemoteImport(self, feedback_fn, iobj,
|
7265 |
self.source_x509_ca,
|
7266 |
self._cds, timeouts)
|
7267 |
if not compat.all(disk_results): |
7268 |
# TODO: Should the instance still be started, even if some disks
|
7269 |
# failed to import (valid for local imports, too)?
|
7270 |
self.LogWarning("Some disks for instance %s on node %s were not" |
7271 |
" imported successfully" % (instance, pnode_name))
|
7272 |
|
7273 |
# Run rename script on newly imported instance
|
7274 |
assert iobj.name == instance
|
7275 |
feedback_fn("Running rename script for %s" % instance)
|
7276 |
result = self.rpc.call_instance_run_rename(pnode_name, iobj,
|
7277 |
self.source_instance_name,
|
7278 |
self.op.debug_level)
|
7279 |
if result.fail_msg:
|
7280 |
self.LogWarning("Failed to run rename script for %s on node" |
7281 |
" %s: %s" % (instance, pnode_name, result.fail_msg))
|
7282 |
|
7283 |
else:
|
7284 |
# also checked in the prereq part
|
7285 |
raise errors.ProgrammerError("Unknown OS initialization mode '%s'" |
7286 |
% self.op.mode)
|
7287 |
|
7288 |
if self.op.start: |
7289 |
iobj.admin_up = True
|
7290 |
self.cfg.Update(iobj, feedback_fn)
|
7291 |
logging.info("Starting instance %s on node %s", instance, pnode_name)
|
7292 |
feedback_fn("* starting instance...")
|
7293 |
result = self.rpc.call_instance_start(pnode_name, iobj, None, None) |
7294 |
result.Raise("Could not start instance")
|
7295 |
|
7296 |
return list(iobj.all_nodes) |
7297 |
|
7298 |
|
7299 |
class LUConnectConsole(NoHooksLU): |
7300 |
"""Connect to an instance's console.
|
7301 |
|
7302 |
This is somewhat special in that it returns the command line that
|
7303 |
you need to run on the master node in order to connect to the
|
7304 |
console.
|
7305 |
|
7306 |
"""
|
7307 |
_OP_PARAMS = [ |
7308 |
_PInstanceName |
7309 |
] |
7310 |
REQ_BGL = False
|
7311 |
|
7312 |
def ExpandNames(self): |
7313 |
self._ExpandAndLockInstance()
|
7314 |
|
7315 |
def CheckPrereq(self): |
7316 |
"""Check prerequisites.
|
7317 |
|
7318 |
This checks that the instance is in the cluster.
|
7319 |
|
7320 |
"""
|
7321 |
self.instance = self.cfg.GetInstanceInfo(self.op.instance_name) |
7322 |
assert self.instance is not None, \ |
7323 |
"Cannot retrieve locked instance %s" % self.op.instance_name |
7324 |
_CheckNodeOnline(self, self.instance.primary_node) |
7325 |
|
7326 |
def Exec(self, feedback_fn): |
7327 |
"""Connect to the console of an instance
|
7328 |
|
7329 |
"""
|
7330 |
instance = self.instance
|
7331 |
node = instance.primary_node |
7332 |
|
7333 |
node_insts = self.rpc.call_instance_list([node],
|
7334 |
[instance.hypervisor])[node] |
7335 |
node_insts.Raise("Can't get node information from %s" % node)
|
7336 |
|
7337 |
if instance.name not in node_insts.payload: |
7338 |
raise errors.OpExecError("Instance %s is not running." % instance.name) |
7339 |
|
7340 |
logging.debug("Connecting to console of %s on %s", instance.name, node)
|
7341 |
|
7342 |
hyper = hypervisor.GetHypervisor(instance.hypervisor) |
7343 |
cluster = self.cfg.GetClusterInfo()
|
7344 |
# beparams and hvparams are passed separately, to avoid editing the
|
7345 |
# instance and then saving the defaults in the instance itself.
|
7346 |
hvparams = cluster.FillHV(instance) |
7347 |
beparams = cluster.FillBE(instance) |
7348 |
console_cmd = hyper.GetShellCommandForConsole(instance, hvparams, beparams) |
7349 |
|
7350 |
# build ssh cmdline
|
7351 |
return self.ssh.BuildCmd(node, "root", console_cmd, batch=True, tty=True) |
7352 |
|
7353 |
|
7354 |
class LUReplaceDisks(LogicalUnit): |
7355 |
"""Replace the disks of an instance.
|
7356 |
|
7357 |
"""
|
7358 |
HPATH = "mirrors-replace"
|
7359 |
HTYPE = constants.HTYPE_INSTANCE |
7360 |
_OP_PARAMS = [ |
7361 |
_PInstanceName, |
7362 |
("mode", _NoDefault, _TElemOf(constants.REPLACE_MODES)),
|
7363 |
("disks", _EmptyList, _TListOf(_TPositiveInt)),
|
7364 |
("remote_node", None, _TMaybeString), |
7365 |
("iallocator", None, _TMaybeString), |
7366 |
("early_release", False, _TBool), |
7367 |
] |
7368 |
REQ_BGL = False
|
7369 |
|
7370 |
def CheckArguments(self): |
7371 |
TLReplaceDisks.CheckArguments(self.op.mode, self.op.remote_node, |
7372 |
self.op.iallocator)
|
7373 |
|
7374 |
def ExpandNames(self): |
7375 |
self._ExpandAndLockInstance()
|
7376 |
|
7377 |
if self.op.iallocator is not None: |
7378 |
self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
|
7379 |
|
7380 |
elif self.op.remote_node is not None: |
7381 |
remote_node = _ExpandNodeName(self.cfg, self.op.remote_node) |
7382 |
self.op.remote_node = remote_node
|
7383 |
|
7384 |
# Warning: do not remove the locking of the new secondary here
|
7385 |
# unless DRBD8.AddChildren is changed to work in parallel;
|
7386 |
# currently it doesn't since parallel invocations of
|
7387 |
# FindUnusedMinor will conflict
|
7388 |
self.needed_locks[locking.LEVEL_NODE] = [remote_node]
|
7389 |
self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
|
7390 |
|
7391 |
else:
|
7392 |
self.needed_locks[locking.LEVEL_NODE] = []
|
7393 |
self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
|
7394 |
|
7395 |
self.replacer = TLReplaceDisks(self, self.op.instance_name, self.op.mode, |
7396 |
self.op.iallocator, self.op.remote_node, |
7397 |
self.op.disks, False, self.op.early_release) |
7398 |
|
7399 |
self.tasklets = [self.replacer] |
7400 |
|
7401 |
def DeclareLocks(self, level): |
7402 |
# If we're not already locking all nodes in the set we have to declare the
|
7403 |
# instance's primary/secondary nodes.
|
7404 |
if (level == locking.LEVEL_NODE and |
7405 |
self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET): |
7406 |
self._LockInstancesNodes()
|
7407 |
|
7408 |
def BuildHooksEnv(self): |
7409 |
"""Build hooks env.
|
7410 |
|
7411 |
This runs on the master, the primary and all the secondaries.
|
7412 |
|
7413 |
"""
|
7414 |
instance = self.replacer.instance
|
7415 |
env = { |
7416 |
"MODE": self.op.mode, |
7417 |
"NEW_SECONDARY": self.op.remote_node, |
7418 |
"OLD_SECONDARY": instance.secondary_nodes[0], |
7419 |
} |
7420 |
env.update(_BuildInstanceHookEnvByObject(self, instance))
|
7421 |
nl = [ |
7422 |
self.cfg.GetMasterNode(),
|
7423 |
instance.primary_node, |
7424 |
] |
7425 |
if self.op.remote_node is not None: |
7426 |
nl.append(self.op.remote_node)
|
7427 |
return env, nl, nl
|
7428 |
|
7429 |
|
7430 |
class TLReplaceDisks(Tasklet): |
7431 |
"""Replaces disks for an instance.
|
7432 |
|
7433 |
Note: Locking is not within the scope of this class.
|
7434 |
|
7435 |
"""
|
7436 |
def __init__(self, lu, instance_name, mode, iallocator_name, remote_node, |
7437 |
disks, delay_iallocator, early_release): |
7438 |
"""Initializes this class.
|
7439 |
|
7440 |
"""
|
7441 |
Tasklet.__init__(self, lu)
|
7442 |
|
7443 |
# Parameters
|
7444 |
self.instance_name = instance_name
|
7445 |
self.mode = mode
|
7446 |
self.iallocator_name = iallocator_name
|
7447 |
self.remote_node = remote_node
|
7448 |
self.disks = disks
|
7449 |
self.delay_iallocator = delay_iallocator
|
7450 |
self.early_release = early_release
|
7451 |
|
7452 |
# Runtime data
|
7453 |
self.instance = None |
7454 |
self.new_node = None |
7455 |
self.target_node = None |
7456 |
self.other_node = None |
7457 |
self.remote_node_info = None |
7458 |
self.node_secondary_ip = None |
7459 |
|
7460 |
@staticmethod
|
7461 |
def CheckArguments(mode, remote_node, iallocator): |
7462 |
"""Helper function for users of this class.
|
7463 |
|
7464 |
"""
|
7465 |
# check for valid parameter combination
|
7466 |
if mode == constants.REPLACE_DISK_CHG:
|
7467 |
if remote_node is None and iallocator is None: |
7468 |
raise errors.OpPrereqError("When changing the secondary either an" |
7469 |
" iallocator script must be used or the"
|
7470 |
" new node given", errors.ECODE_INVAL)
|
7471 |
|
7472 |
if remote_node is not None and iallocator is not None: |
7473 |
raise errors.OpPrereqError("Give either the iallocator or the new" |
7474 |
" secondary, not both", errors.ECODE_INVAL)
|
7475 |
|
7476 |
elif remote_node is not None or iallocator is not None: |
7477 |
# Not replacing the secondary
|
7478 |
raise errors.OpPrereqError("The iallocator and new node options can" |
7479 |
" only be used when changing the"
|
7480 |
" secondary node", errors.ECODE_INVAL)
|
7481 |
|
7482 |
@staticmethod
|
7483 |
def _RunAllocator(lu, iallocator_name, instance_name, relocate_from): |
7484 |
"""Compute a new secondary node using an IAllocator.
|
7485 |
|
7486 |
"""
|
7487 |
ial = IAllocator(lu.cfg, lu.rpc, |
7488 |
mode=constants.IALLOCATOR_MODE_RELOC, |
7489 |
name=instance_name, |
7490 |
relocate_from=relocate_from) |
7491 |
|
7492 |
ial.Run(iallocator_name) |
7493 |
|
7494 |
if not ial.success: |
7495 |
raise errors.OpPrereqError("Can't compute nodes using iallocator '%s':" |
7496 |
" %s" % (iallocator_name, ial.info),
|
7497 |
errors.ECODE_NORES) |
7498 |
|
7499 |
if len(ial.result) != ial.required_nodes: |
7500 |
raise errors.OpPrereqError("iallocator '%s' returned invalid number" |
7501 |
" of nodes (%s), required %s" %
|
7502 |
(iallocator_name, |
7503 |
len(ial.result), ial.required_nodes),
|
7504 |
errors.ECODE_FAULT) |
7505 |
|
7506 |
remote_node_name = ial.result[0]
|
7507 |
|
7508 |
lu.LogInfo("Selected new secondary for instance '%s': %s",
|
7509 |
instance_name, remote_node_name) |
7510 |
|
7511 |
return remote_node_name
|
7512 |
|
7513 |
def _FindFaultyDisks(self, node_name): |
7514 |
return _FindFaultyInstanceDisks(self.cfg, self.rpc, self.instance, |
7515 |
node_name, True)
|
7516 |
|
7517 |
def CheckPrereq(self): |
7518 |
"""Check prerequisites.
|
7519 |
|
7520 |
This checks that the instance is in the cluster.
|
7521 |
|
7522 |
"""
|
7523 |
self.instance = instance = self.cfg.GetInstanceInfo(self.instance_name) |
7524 |
assert instance is not None, \ |
7525 |
"Cannot retrieve locked instance %s" % self.instance_name |
7526 |
|
7527 |
if instance.disk_template != constants.DT_DRBD8:
|
7528 |
raise errors.OpPrereqError("Can only run replace disks for DRBD8-based" |
7529 |
" instances", errors.ECODE_INVAL)
|
7530 |
|
7531 |
if len(instance.secondary_nodes) != 1: |
7532 |
raise errors.OpPrereqError("The instance has a strange layout," |
7533 |
" expected one secondary but found %d" %
|
7534 |
len(instance.secondary_nodes),
|
7535 |
errors.ECODE_FAULT) |
7536 |
|
7537 |
if not self.delay_iallocator: |
7538 |
self._CheckPrereq2()
|
7539 |
|
7540 |
def _CheckPrereq2(self): |
7541 |
"""Check prerequisites, second part.
|
7542 |
|
7543 |
This function should always be part of CheckPrereq. It was separated and is
|
7544 |
now called from Exec because during node evacuation iallocator was only
|
7545 |
called with an unmodified cluster model, not taking planned changes into
|
7546 |
account.
|
7547 |
|
7548 |
"""
|
7549 |
instance = self.instance
|
7550 |
secondary_node = instance.secondary_nodes[0]
|
7551 |
|
7552 |
if self.iallocator_name is None: |
7553 |
remote_node = self.remote_node
|
7554 |
else:
|
7555 |
remote_node = self._RunAllocator(self.lu, self.iallocator_name, |
7556 |
instance.name, instance.secondary_nodes) |
7557 |
|
7558 |
if remote_node is not None: |
7559 |
self.remote_node_info = self.cfg.GetNodeInfo(remote_node) |
7560 |
assert self.remote_node_info is not None, \ |
7561 |
"Cannot retrieve locked node %s" % remote_node
|
7562 |
else:
|
7563 |
self.remote_node_info = None |
7564 |
|
7565 |
if remote_node == self.instance.primary_node: |
7566 |
raise errors.OpPrereqError("The specified node is the primary node of" |
7567 |
" the instance.", errors.ECODE_INVAL)
|
7568 |
|
7569 |
if remote_node == secondary_node:
|
7570 |
raise errors.OpPrereqError("The specified node is already the" |
7571 |
" secondary node of the instance.",
|
7572 |
errors.ECODE_INVAL) |
7573 |
|
7574 |
if self.disks and self.mode in (constants.REPLACE_DISK_AUTO, |
7575 |
constants.REPLACE_DISK_CHG): |
7576 |
raise errors.OpPrereqError("Cannot specify disks to be replaced", |
7577 |
errors.ECODE_INVAL) |
7578 |
|
7579 |
if self.mode == constants.REPLACE_DISK_AUTO: |
7580 |
faulty_primary = self._FindFaultyDisks(instance.primary_node)
|
7581 |
faulty_secondary = self._FindFaultyDisks(secondary_node)
|
7582 |
|
7583 |
if faulty_primary and faulty_secondary: |
7584 |
raise errors.OpPrereqError("Instance %s has faulty disks on more than" |
7585 |
" one node and can not be repaired"
|
7586 |
" automatically" % self.instance_name, |
7587 |
errors.ECODE_STATE) |
7588 |
|
7589 |
if faulty_primary:
|
7590 |
self.disks = faulty_primary
|
7591 |
self.target_node = instance.primary_node
|
7592 |
self.other_node = secondary_node
|
7593 |
check_nodes = [self.target_node, self.other_node] |
7594 |
elif faulty_secondary:
|
7595 |
self.disks = faulty_secondary
|
7596 |
self.target_node = secondary_node
|
7597 |
self.other_node = instance.primary_node
|
7598 |
check_nodes = [self.target_node, self.other_node] |
7599 |
else:
|
7600 |
self.disks = []
|
7601 |
check_nodes = [] |
7602 |
|
7603 |
else:
|
7604 |
# Non-automatic modes
|
7605 |
if self.mode == constants.REPLACE_DISK_PRI: |
7606 |
self.target_node = instance.primary_node
|
7607 |
self.other_node = secondary_node
|
7608 |
check_nodes = [self.target_node, self.other_node] |
7609 |
|
7610 |
elif self.mode == constants.REPLACE_DISK_SEC: |
7611 |
self.target_node = secondary_node
|
7612 |
self.other_node = instance.primary_node
|
7613 |
check_nodes = [self.target_node, self.other_node] |
7614 |
|
7615 |
elif self.mode == constants.REPLACE_DISK_CHG: |
7616 |
self.new_node = remote_node
|
7617 |
self.other_node = instance.primary_node
|
7618 |
self.target_node = secondary_node
|
7619 |
check_nodes = [self.new_node, self.other_node] |
7620 |
|
7621 |
_CheckNodeNotDrained(self.lu, remote_node)
|
7622 |
|
7623 |
old_node_info = self.cfg.GetNodeInfo(secondary_node)
|
7624 |
assert old_node_info is not None |
7625 |
if old_node_info.offline and not self.early_release: |
7626 |
# doesn't make sense to delay the release
|
7627 |
self.early_release = True |
7628 |
self.lu.LogInfo("Old secondary %s is offline, automatically enabling" |
7629 |
" early-release mode", secondary_node)
|
7630 |
|
7631 |
else:
|
7632 |
raise errors.ProgrammerError("Unhandled disk replace mode (%s)" % |
7633 |
self.mode)
|
7634 |
|
7635 |
# If not specified all disks should be replaced
|
7636 |
if not self.disks: |
7637 |
self.disks = range(len(self.instance.disks)) |
7638 |
|
7639 |
for node in check_nodes: |
7640 |
_CheckNodeOnline(self.lu, node)
|
7641 |
|
7642 |
# Check whether disks are valid
|
7643 |
for disk_idx in self.disks: |
7644 |
instance.FindDisk(disk_idx) |
7645 |
|
7646 |
# Get secondary node IP addresses
|
7647 |
node_2nd_ip = {} |
7648 |
|
7649 |
for node_name in [self.target_node, self.other_node, self.new_node]: |
7650 |
if node_name is not None: |
7651 |
node_2nd_ip[node_name] = self.cfg.GetNodeInfo(node_name).secondary_ip
|
7652 |
|
7653 |
self.node_secondary_ip = node_2nd_ip
|
7654 |
|
7655 |
def Exec(self, feedback_fn): |
7656 |
"""Execute disk replacement.
|
7657 |
|
7658 |
This dispatches the disk replacement to the appropriate handler.
|
7659 |
|
7660 |
"""
|
7661 |
if self.delay_iallocator: |
7662 |
self._CheckPrereq2()
|
7663 |
|
7664 |
if not self.disks: |
7665 |
feedback_fn("No disks need replacement")
|
7666 |
return
|
7667 |
|
7668 |
feedback_fn("Replacing disk(s) %s for %s" %
|
7669 |
(utils.CommaJoin(self.disks), self.instance.name)) |
7670 |
|
7671 |
activate_disks = (not self.instance.admin_up) |
7672 |
|
7673 |
# Activate the instance disks if we're replacing them on a down instance
|
7674 |
if activate_disks:
|
7675 |
_StartInstanceDisks(self.lu, self.instance, True) |
7676 |
|
7677 |
try:
|
7678 |
# Should we replace the secondary node?
|
7679 |
if self.new_node is not None: |
7680 |
fn = self._ExecDrbd8Secondary
|
7681 |
else:
|
7682 |
fn = self._ExecDrbd8DiskOnly
|
7683 |
|
7684 |
return fn(feedback_fn)
|
7685 |
|
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 |
def _CheckVolumeGroup(self, nodes): |
7693 |
self.lu.LogInfo("Checking volume groups") |
7694 |
|
7695 |
vgname = self.cfg.GetVGName()
|
7696 |
|
7697 |
# Make sure volume group exists on all involved nodes
|
7698 |
results = self.rpc.call_vg_list(nodes)
|
7699 |
if not results: |
7700 |
raise errors.OpExecError("Can't list volume groups on the nodes") |
7701 |
|
7702 |
for node in nodes: |
7703 |
res = results[node] |
7704 |
res.Raise("Error checking node %s" % node)
|
7705 |
if vgname not in res.payload: |
7706 |
raise errors.OpExecError("Volume group '%s' not found on node %s" % |
7707 |
(vgname, node)) |
7708 |
|
7709 |
def _CheckDisksExistence(self, nodes): |
7710 |
# Check disk existence
|
7711 |
for idx, dev in enumerate(self.instance.disks): |
7712 |
if idx not in self.disks: |
7713 |
continue
|
7714 |
|
7715 |
for node in nodes: |
7716 |
self.lu.LogInfo("Checking disk/%d on %s" % (idx, node)) |
7717 |
self.cfg.SetDiskID(dev, node)
|
7718 |
|
7719 |
result = self.rpc.call_blockdev_find(node, dev)
|
7720 |
|
7721 |
msg = result.fail_msg |
7722 |
if msg or not result.payload: |
7723 |
if not msg: |
7724 |
msg = "disk not found"
|
7725 |
raise errors.OpExecError("Can't find disk/%d on node %s: %s" % |
7726 |
(idx, node, msg)) |
7727 |
|
7728 |
def _CheckDisksConsistency(self, node_name, on_primary, ldisk): |
7729 |
for idx, dev in enumerate(self.instance.disks): |
7730 |
if idx not in self.disks: |
7731 |
continue
|
7732 |
|
7733 |
self.lu.LogInfo("Checking disk/%d consistency on node %s" % |
7734 |
(idx, node_name)) |
7735 |
|
7736 |
if not _CheckDiskConsistency(self.lu, dev, node_name, on_primary, |
7737 |
ldisk=ldisk): |
7738 |
raise errors.OpExecError("Node %s has degraded storage, unsafe to" |
7739 |
" replace disks for instance %s" %
|
7740 |
(node_name, self.instance.name))
|
7741 |
|
7742 |
def _CreateNewStorage(self, node_name): |
7743 |
vgname = self.cfg.GetVGName()
|
7744 |
iv_names = {} |
7745 |
|
7746 |
for idx, dev in enumerate(self.instance.disks): |
7747 |
if idx not in self.disks: |
7748 |
continue
|
7749 |
|
7750 |
self.lu.LogInfo("Adding storage on %s for disk/%d" % (node_name, idx)) |
7751 |
|
7752 |
self.cfg.SetDiskID(dev, node_name)
|
7753 |
|
7754 |
lv_names = [".disk%d_%s" % (idx, suffix) for suffix in ["data", "meta"]] |
7755 |
names = _GenerateUniqueNames(self.lu, lv_names)
|
7756 |
|
7757 |
lv_data = objects.Disk(dev_type=constants.LD_LV, size=dev.size, |
7758 |
logical_id=(vgname, names[0]))
|
7759 |
lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
|
7760 |
logical_id=(vgname, names[1]))
|
7761 |
|
7762 |
new_lvs = [lv_data, lv_meta] |
7763 |
old_lvs = dev.children |
7764 |
iv_names[dev.iv_name] = (dev, old_lvs, new_lvs) |
7765 |
|
7766 |
# we pass force_create=True to force the LVM creation
|
7767 |
for new_lv in new_lvs: |
7768 |
_CreateBlockDev(self.lu, node_name, self.instance, new_lv, True, |
7769 |
_GetInstanceInfoText(self.instance), False) |
7770 |
|
7771 |
return iv_names
|
7772 |
|
7773 |
def _CheckDevices(self, node_name, iv_names): |
7774 |
for name, (dev, _, _) in iv_names.iteritems(): |
7775 |
self.cfg.SetDiskID(dev, node_name)
|
7776 |
|
7777 |
result = self.rpc.call_blockdev_find(node_name, dev)
|
7778 |
|
7779 |
msg = result.fail_msg |
7780 |
if msg or not result.payload: |
7781 |
if not msg: |
7782 |
msg = "disk not found"
|
7783 |
raise errors.OpExecError("Can't find DRBD device %s: %s" % |
7784 |
(name, msg)) |
7785 |
|
7786 |
if result.payload.is_degraded:
|
7787 |
raise errors.OpExecError("DRBD device %s is degraded!" % name) |
7788 |
|
7789 |
def _RemoveOldStorage(self, node_name, iv_names): |
7790 |
for name, (_, old_lvs, _) in iv_names.iteritems(): |
7791 |
self.lu.LogInfo("Remove logical volumes for %s" % name) |
7792 |
|
7793 |
for lv in old_lvs: |
7794 |
self.cfg.SetDiskID(lv, node_name)
|
7795 |
|
7796 |
msg = self.rpc.call_blockdev_remove(node_name, lv).fail_msg
|
7797 |
if msg:
|
7798 |
self.lu.LogWarning("Can't remove old LV: %s" % msg, |
7799 |
hint="remove unused LVs manually")
|
7800 |
|
7801 |
def _ReleaseNodeLock(self, node_name): |
7802 |
"""Releases the lock for a given node."""
|
7803 |
self.lu.context.glm.release(locking.LEVEL_NODE, node_name)
|
7804 |
|
7805 |
def _ExecDrbd8DiskOnly(self, feedback_fn): |
7806 |
"""Replace a disk on the primary or secondary for DRBD 8.
|
7807 |
|
7808 |
The algorithm for replace is quite complicated:
|
7809 |
|
7810 |
1. for each disk to be replaced:
|
7811 |
|
7812 |
1. create new LVs on the target node with unique names
|
7813 |
1. detach old LVs from the drbd device
|
7814 |
1. rename old LVs to name_replaced.<time_t>
|
7815 |
1. rename new LVs to old LVs
|
7816 |
1. attach the new LVs (with the old names now) to the drbd device
|
7817 |
|
7818 |
1. wait for sync across all devices
|
7819 |
|
7820 |
1. for each modified disk:
|
7821 |
|
7822 |
1. remove old LVs (which have the name name_replaces.<time_t>)
|
7823 |
|
7824 |
Failures are not very well handled.
|
7825 |
|
7826 |
"""
|
7827 |
steps_total = 6
|
7828 |
|
7829 |
# Step: check device activation
|
7830 |
self.lu.LogStep(1, steps_total, "Check device existence") |
7831 |
self._CheckDisksExistence([self.other_node, self.target_node]) |
7832 |
self._CheckVolumeGroup([self.target_node, self.other_node]) |
7833 |
|
7834 |
# Step: check other node consistency
|
7835 |
self.lu.LogStep(2, steps_total, "Check peer consistency") |
7836 |
self._CheckDisksConsistency(self.other_node, |
7837 |
self.other_node == self.instance.primary_node, |
7838 |
False)
|
7839 |
|
7840 |
# Step: create new storage
|
7841 |
self.lu.LogStep(3, steps_total, "Allocate new storage") |
7842 |
iv_names = self._CreateNewStorage(self.target_node) |
7843 |
|
7844 |
# Step: for each lv, detach+rename*2+attach
|
7845 |
self.lu.LogStep(4, steps_total, "Changing drbd configuration") |
7846 |
for dev, old_lvs, new_lvs in iv_names.itervalues(): |
7847 |
self.lu.LogInfo("Detaching %s drbd from local storage" % dev.iv_name) |
7848 |
|
7849 |
result = self.rpc.call_blockdev_removechildren(self.target_node, dev, |
7850 |
old_lvs) |
7851 |
result.Raise("Can't detach drbd from local storage on node"
|
7852 |
" %s for device %s" % (self.target_node, dev.iv_name)) |
7853 |
#dev.children = []
|
7854 |
#cfg.Update(instance)
|
7855 |
|
7856 |
# ok, we created the new LVs, so now we know we have the needed
|
7857 |
# storage; as such, we proceed on the target node to rename
|
7858 |
# old_lv to _old, and new_lv to old_lv; note that we rename LVs
|
7859 |
# using the assumption that logical_id == physical_id (which in
|
7860 |
# turn is the unique_id on that node)
|
7861 |
|
7862 |
# FIXME(iustin): use a better name for the replaced LVs
|
7863 |
temp_suffix = int(time.time())
|
7864 |
ren_fn = lambda d, suff: (d.physical_id[0], |
7865 |
d.physical_id[1] + "_replaced-%s" % suff) |
7866 |
|
7867 |
# Build the rename list based on what LVs exist on the node
|
7868 |
rename_old_to_new = [] |
7869 |
for to_ren in old_lvs: |
7870 |
result = self.rpc.call_blockdev_find(self.target_node, to_ren) |
7871 |
if not result.fail_msg and result.payload: |
7872 |
# device exists
|
7873 |
rename_old_to_new.append((to_ren, ren_fn(to_ren, temp_suffix))) |
7874 |
|
7875 |
self.lu.LogInfo("Renaming the old LVs on the target node") |
7876 |
result = self.rpc.call_blockdev_rename(self.target_node, |
7877 |
rename_old_to_new) |
7878 |
result.Raise("Can't rename old LVs on node %s" % self.target_node) |
7879 |
|
7880 |
# Now we rename the new LVs to the old LVs
|
7881 |
self.lu.LogInfo("Renaming the new LVs on the target node") |
7882 |
rename_new_to_old = [(new, old.physical_id) |
7883 |
for old, new in zip(old_lvs, new_lvs)] |
7884 |
result = self.rpc.call_blockdev_rename(self.target_node, |
7885 |
rename_new_to_old) |
7886 |
result.Raise("Can't rename new LVs on node %s" % self.target_node) |
7887 |
|
7888 |
for old, new in zip(old_lvs, new_lvs): |
7889 |
new.logical_id = old.logical_id |
7890 |
self.cfg.SetDiskID(new, self.target_node) |
7891 |
|
7892 |
for disk in old_lvs: |
7893 |
disk.logical_id = ren_fn(disk, temp_suffix) |
7894 |
self.cfg.SetDiskID(disk, self.target_node) |
7895 |
|
7896 |
# Now that the new lvs have the old name, we can add them to the device
|
7897 |
self.lu.LogInfo("Adding new mirror component on %s" % self.target_node) |
7898 |
result = self.rpc.call_blockdev_addchildren(self.target_node, dev, |
7899 |
new_lvs) |
7900 |
msg = result.fail_msg |
7901 |
if msg:
|
7902 |
for new_lv in new_lvs: |
7903 |
msg2 = self.rpc.call_blockdev_remove(self.target_node, |
7904 |
new_lv).fail_msg |
7905 |
if msg2:
|
7906 |
self.lu.LogWarning("Can't rollback device %s: %s", dev, msg2, |
7907 |
hint=("cleanup manually the unused logical"
|
7908 |
"volumes"))
|
7909 |
raise errors.OpExecError("Can't add local storage to drbd: %s" % msg) |
7910 |
|
7911 |
dev.children = new_lvs |
7912 |
|
7913 |
self.cfg.Update(self.instance, feedback_fn) |
7914 |
|
7915 |
cstep = 5
|
7916 |
if self.early_release: |
7917 |
self.lu.LogStep(cstep, steps_total, "Removing old storage") |
7918 |
cstep += 1
|
7919 |
self._RemoveOldStorage(self.target_node, iv_names) |
7920 |
# WARNING: we release both node locks here, do not do other RPCs
|
7921 |
# than WaitForSync to the primary node
|
7922 |
self._ReleaseNodeLock([self.target_node, self.other_node]) |
7923 |
|
7924 |
# Wait for sync
|
7925 |
# This can fail as the old devices are degraded and _WaitForSync
|
7926 |
# does a combined result over all disks, so we don't check its return value
|
7927 |
self.lu.LogStep(cstep, steps_total, "Sync devices") |
7928 |
cstep += 1
|
7929 |
_WaitForSync(self.lu, self.instance) |
7930 |
|
7931 |
# Check all devices manually
|
7932 |
self._CheckDevices(self.instance.primary_node, iv_names) |
7933 |
|
7934 |
# Step: remove old storage
|
7935 |
if not self.early_release: |
7936 |
self.lu.LogStep(cstep, steps_total, "Removing old storage") |
7937 |
cstep += 1
|
7938 |
self._RemoveOldStorage(self.target_node, iv_names) |
7939 |
|
7940 |
def _ExecDrbd8Secondary(self, feedback_fn): |
7941 |
"""Replace the secondary node for DRBD 8.
|
7942 |
|
7943 |
The algorithm for replace is quite complicated:
|
7944 |
- for all disks of the instance:
|
7945 |
- create new LVs on the new node with same names
|
7946 |
- shutdown the drbd device on the old secondary
|
7947 |
- disconnect the drbd network on the primary
|
7948 |
- create the drbd device on the new secondary
|
7949 |
- network attach the drbd on the primary, using an artifice:
|
7950 |
the drbd code for Attach() will connect to the network if it
|
7951 |
finds a device which is connected to the good local disks but
|
7952 |
not network enabled
|
7953 |
- wait for sync across all devices
|
7954 |
- remove all disks from the old secondary
|
7955 |
|
7956 |
Failures are not very well handled.
|
7957 |
|
7958 |
"""
|
7959 |
steps_total = 6
|
7960 |
|
7961 |
# Step: check device activation
|
7962 |
self.lu.LogStep(1, steps_total, "Check device existence") |
7963 |
self._CheckDisksExistence([self.instance.primary_node]) |
7964 |
self._CheckVolumeGroup([self.instance.primary_node]) |
7965 |
|
7966 |
# Step: check other node consistency
|
7967 |
self.lu.LogStep(2, steps_total, "Check peer consistency") |
7968 |
self._CheckDisksConsistency(self.instance.primary_node, True, True) |
7969 |
|
7970 |
# Step: create new storage
|
7971 |
self.lu.LogStep(3, steps_total, "Allocate new storage") |
7972 |
for idx, dev in enumerate(self.instance.disks): |
7973 |
self.lu.LogInfo("Adding new local storage on %s for disk/%d" % |
7974 |
(self.new_node, idx))
|
7975 |
# we pass force_create=True to force LVM creation
|
7976 |
for new_lv in dev.children: |
7977 |
_CreateBlockDev(self.lu, self.new_node, self.instance, new_lv, True, |
7978 |
_GetInstanceInfoText(self.instance), False) |
7979 |
|
7980 |
# Step 4: dbrd minors and drbd setups changes
|
7981 |
# after this, we must manually remove the drbd minors on both the
|
7982 |
# error and the success paths
|
7983 |
self.lu.LogStep(4, steps_total, "Changing drbd configuration") |
7984 |
minors = self.cfg.AllocateDRBDMinor([self.new_node |
7985 |
for dev in self.instance.disks], |
7986 |
self.instance.name)
|
7987 |
logging.debug("Allocated minors %r", minors)
|
7988 |
|
7989 |
iv_names = {} |
7990 |
for idx, (dev, new_minor) in enumerate(zip(self.instance.disks, minors)): |
7991 |
self.lu.LogInfo("activating a new drbd on %s for disk/%d" % |
7992 |
(self.new_node, idx))
|
7993 |
# create new devices on new_node; note that we create two IDs:
|
7994 |
# one without port, so the drbd will be activated without
|
7995 |
# networking information on the new node at this stage, and one
|
7996 |
# with network, for the latter activation in step 4
|
7997 |
(o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id |
7998 |
if self.instance.primary_node == o_node1: |
7999 |
p_minor = o_minor1 |
8000 |
else:
|
8001 |
assert self.instance.primary_node == o_node2, "Three-node instance?" |
8002 |
p_minor = o_minor2 |
8003 |
|
8004 |
new_alone_id = (self.instance.primary_node, self.new_node, None, |
8005 |
p_minor, new_minor, o_secret) |
8006 |
new_net_id = (self.instance.primary_node, self.new_node, o_port, |
8007 |
p_minor, new_minor, o_secret) |
8008 |
|
8009 |
iv_names[idx] = (dev, dev.children, new_net_id) |
8010 |
logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
|
8011 |
new_net_id) |
8012 |
new_drbd = objects.Disk(dev_type=constants.LD_DRBD8, |
8013 |
logical_id=new_alone_id, |
8014 |
children=dev.children, |
8015 |
size=dev.size) |
8016 |
try:
|
8017 |
_CreateSingleBlockDev(self.lu, self.new_node, self.instance, new_drbd, |
8018 |
_GetInstanceInfoText(self.instance), False) |
8019 |
except errors.GenericError:
|
8020 |
self.cfg.ReleaseDRBDMinors(self.instance.name) |
8021 |
raise
|
8022 |
|
8023 |
# We have new devices, shutdown the drbd on the old secondary
|
8024 |
for idx, dev in enumerate(self.instance.disks): |
8025 |
self.lu.LogInfo("Shutting down drbd for disk/%d on old node" % idx) |
8026 |
self.cfg.SetDiskID(dev, self.target_node) |
8027 |
msg = self.rpc.call_blockdev_shutdown(self.target_node, dev).fail_msg |
8028 |
if msg:
|
8029 |
self.lu.LogWarning("Failed to shutdown drbd for disk/%d on old" |
8030 |
"node: %s" % (idx, msg),
|
8031 |
hint=("Please cleanup this device manually as"
|
8032 |
" soon as possible"))
|
8033 |
|
8034 |
self.lu.LogInfo("Detaching primary drbds from the network (=> standalone)") |
8035 |
result = self.rpc.call_drbd_disconnect_net([self.instance.primary_node], |
8036 |
self.node_secondary_ip,
|
8037 |
self.instance.disks)\
|
8038 |
[self.instance.primary_node]
|
8039 |
|
8040 |
msg = result.fail_msg |
8041 |
if msg:
|
8042 |
# detaches didn't succeed (unlikely)
|
8043 |
self.cfg.ReleaseDRBDMinors(self.instance.name) |
8044 |
raise errors.OpExecError("Can't detach the disks from the network on" |
8045 |
" old node: %s" % (msg,))
|
8046 |
|
8047 |
# if we managed to detach at least one, we update all the disks of
|
8048 |
# the instance to point to the new secondary
|
8049 |
self.lu.LogInfo("Updating instance configuration") |
8050 |
for dev, _, new_logical_id in iv_names.itervalues(): |
8051 |
dev.logical_id = new_logical_id |
8052 |
self.cfg.SetDiskID(dev, self.instance.primary_node) |
8053 |
|
8054 |
self.cfg.Update(self.instance, feedback_fn) |
8055 |
|
8056 |
# and now perform the drbd attach
|
8057 |
self.lu.LogInfo("Attaching primary drbds to new secondary" |
8058 |
" (standalone => connected)")
|
8059 |
result = self.rpc.call_drbd_attach_net([self.instance.primary_node, |
8060 |
self.new_node],
|
8061 |
self.node_secondary_ip,
|
8062 |
self.instance.disks,
|
8063 |
self.instance.name,
|
8064 |
False)
|
8065 |
for to_node, to_result in result.items(): |
8066 |
msg = to_result.fail_msg |
8067 |
if msg:
|
8068 |
self.lu.LogWarning("Can't attach drbd disks on node %s: %s", |
8069 |
to_node, msg, |
8070 |
hint=("please do a gnt-instance info to see the"
|
8071 |
" status of disks"))
|
8072 |
cstep = 5
|
8073 |
if self.early_release: |
8074 |
self.lu.LogStep(cstep, steps_total, "Removing old storage") |
8075 |
cstep += 1
|
8076 |
self._RemoveOldStorage(self.target_node, iv_names) |
8077 |
# WARNING: we release all node locks here, do not do other RPCs
|
8078 |
# than WaitForSync to the primary node
|
8079 |
self._ReleaseNodeLock([self.instance.primary_node, |
8080 |
self.target_node,
|
8081 |
self.new_node])
|
8082 |
|
8083 |
# Wait for sync
|
8084 |
# This can fail as the old devices are degraded and _WaitForSync
|
8085 |
# does a combined result over all disks, so we don't check its return value
|
8086 |
self.lu.LogStep(cstep, steps_total, "Sync devices") |
8087 |
cstep += 1
|
8088 |
_WaitForSync(self.lu, self.instance) |
8089 |
|
8090 |
# Check all devices manually
|
8091 |
self._CheckDevices(self.instance.primary_node, iv_names) |
8092 |
|
8093 |
# Step: remove old storage
|
8094 |
if not self.early_release: |
8095 |
self.lu.LogStep(cstep, steps_total, "Removing old storage") |
8096 |
self._RemoveOldStorage(self.target_node, iv_names) |
8097 |
|
8098 |
|
8099 |
class LURepairNodeStorage(NoHooksLU): |
8100 |
"""Repairs the volume group on a node.
|
8101 |
|
8102 |
"""
|
8103 |
_OP_PARAMS = [ |
8104 |
_PNodeName, |
8105 |
("storage_type", _NoDefault, _CheckStorageType),
|
8106 |
("name", _NoDefault, _TNonEmptyString),
|
8107 |
("ignore_consistency", False, _TBool), |
8108 |
] |
8109 |
REQ_BGL = False
|
8110 |
|
8111 |
def CheckArguments(self): |
8112 |
self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name) |
8113 |
|
8114 |
storage_type = self.op.storage_type
|
8115 |
|
8116 |
if (constants.SO_FIX_CONSISTENCY not in |
8117 |
constants.VALID_STORAGE_OPERATIONS.get(storage_type, [])): |
8118 |
raise errors.OpPrereqError("Storage units of type '%s' can not be" |
8119 |
" repaired" % storage_type,
|
8120 |
errors.ECODE_INVAL) |
8121 |
|
8122 |
def ExpandNames(self): |
8123 |
self.needed_locks = {
|
8124 |
locking.LEVEL_NODE: [self.op.node_name],
|
8125 |
} |
8126 |
|
8127 |
def _CheckFaultyDisks(self, instance, node_name): |
8128 |
"""Ensure faulty disks abort the opcode or at least warn."""
|
8129 |
try:
|
8130 |
if _FindFaultyInstanceDisks(self.cfg, self.rpc, instance, |
8131 |
node_name, True):
|
8132 |
raise errors.OpPrereqError("Instance '%s' has faulty disks on" |
8133 |
" node '%s'" % (instance.name, node_name),
|
8134 |
errors.ECODE_STATE) |
8135 |
except errors.OpPrereqError, err:
|
8136 |
if self.op.ignore_consistency: |
8137 |
self.proc.LogWarning(str(err.args[0])) |
8138 |
else:
|
8139 |
raise
|
8140 |
|
8141 |
def CheckPrereq(self): |
8142 |
"""Check prerequisites.
|
8143 |
|
8144 |
"""
|
8145 |
# Check whether any instance on this node has faulty disks
|
8146 |
for inst in _GetNodeInstances(self.cfg, self.op.node_name): |
8147 |
if not inst.admin_up: |
8148 |
continue
|
8149 |
check_nodes = set(inst.all_nodes)
|
8150 |
check_nodes.discard(self.op.node_name)
|
8151 |
for inst_node_name in check_nodes: |
8152 |
self._CheckFaultyDisks(inst, inst_node_name)
|
8153 |
|
8154 |
def Exec(self, feedback_fn): |
8155 |
feedback_fn("Repairing storage unit '%s' on %s ..." %
|
8156 |
(self.op.name, self.op.node_name)) |
8157 |
|
8158 |
st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type) |
8159 |
result = self.rpc.call_storage_execute(self.op.node_name, |
8160 |
self.op.storage_type, st_args,
|
8161 |
self.op.name,
|
8162 |
constants.SO_FIX_CONSISTENCY) |
8163 |
result.Raise("Failed to repair storage unit '%s' on %s" %
|
8164 |
(self.op.name, self.op.node_name)) |
8165 |
|
8166 |
|
8167 |
class LUNodeEvacuationStrategy(NoHooksLU): |
8168 |
"""Computes the node evacuation strategy.
|
8169 |
|
8170 |
"""
|
8171 |
_OP_PARAMS = [ |
8172 |
("nodes", _NoDefault, _TListOf(_TNonEmptyString)),
|
8173 |
("remote_node", None, _TMaybeString), |
8174 |
("iallocator", None, _TMaybeString), |
8175 |
] |
8176 |
REQ_BGL = False
|
8177 |
|
8178 |
def CheckArguments(self): |
8179 |
if self.op.remote_node is not None and self.op.iallocator is not None: |
8180 |
raise errors.OpPrereqError("Give either the iallocator or the new" |
8181 |
" secondary, not both", errors.ECODE_INVAL)
|
8182 |
|
8183 |
def ExpandNames(self): |
8184 |
self.op.nodes = _GetWantedNodes(self, self.op.nodes) |
8185 |
self.needed_locks = locks = {}
|
8186 |
if self.op.remote_node is None: |
8187 |
locks[locking.LEVEL_NODE] = locking.ALL_SET |
8188 |
else:
|
8189 |
self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node) |
8190 |
locks[locking.LEVEL_NODE] = self.op.nodes + [self.op.remote_node] |
8191 |
|
8192 |
def Exec(self, feedback_fn): |
8193 |
if self.op.remote_node is not None: |
8194 |
instances = [] |
8195 |
for node in self.op.nodes: |
8196 |
instances.extend(_GetNodeSecondaryInstances(self.cfg, node))
|
8197 |
result = [] |
8198 |
for i in instances: |
8199 |
if i.primary_node == self.op.remote_node: |
8200 |
raise errors.OpPrereqError("Node %s is the primary node of" |
8201 |
" instance %s, cannot use it as"
|
8202 |
" secondary" %
|
8203 |
(self.op.remote_node, i.name),
|
8204 |
errors.ECODE_INVAL) |
8205 |
result.append([i.name, self.op.remote_node])
|
8206 |
else:
|
8207 |
ial = IAllocator(self.cfg, self.rpc, |
8208 |
mode=constants.IALLOCATOR_MODE_MEVAC, |
8209 |
evac_nodes=self.op.nodes)
|
8210 |
ial.Run(self.op.iallocator, validate=True) |
8211 |
if not ial.success: |
8212 |
raise errors.OpExecError("No valid evacuation solution: %s" % ial.info, |
8213 |
errors.ECODE_NORES) |
8214 |
result = ial.result |
8215 |
return result
|
8216 |
|
8217 |
|
8218 |
class LUGrowDisk(LogicalUnit): |
8219 |
"""Grow a disk of an instance.
|
8220 |
|
8221 |
"""
|
8222 |
HPATH = "disk-grow"
|
8223 |
HTYPE = constants.HTYPE_INSTANCE |
8224 |
_OP_PARAMS = [ |
8225 |
_PInstanceName, |
8226 |
("disk", _NoDefault, _TInt),
|
8227 |
("amount", _NoDefault, _TInt),
|
8228 |
("wait_for_sync", True, _TBool), |
8229 |
] |
8230 |
REQ_BGL = False
|
8231 |
|
8232 |
def ExpandNames(self): |
8233 |
self._ExpandAndLockInstance()
|
8234 |
self.needed_locks[locking.LEVEL_NODE] = []
|
8235 |
self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
|
8236 |
|
8237 |
def DeclareLocks(self, level): |
8238 |
if level == locking.LEVEL_NODE:
|
8239 |
self._LockInstancesNodes()
|
8240 |
|
8241 |
def BuildHooksEnv(self): |
8242 |
"""Build hooks env.
|
8243 |
|
8244 |
This runs on the master, the primary and all the secondaries.
|
8245 |
|
8246 |
"""
|
8247 |
env = { |
8248 |
"DISK": self.op.disk, |
8249 |
"AMOUNT": self.op.amount, |
8250 |
} |
8251 |
env.update(_BuildInstanceHookEnvByObject(self, self.instance)) |
8252 |
nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes) |
8253 |
return env, nl, nl
|
8254 |
|
8255 |
def CheckPrereq(self): |
8256 |
"""Check prerequisites.
|
8257 |
|
8258 |
This checks that the instance is in the cluster.
|
8259 |
|
8260 |
"""
|
8261 |
instance = self.cfg.GetInstanceInfo(self.op.instance_name) |
8262 |
assert instance is not None, \ |
8263 |
"Cannot retrieve locked instance %s" % self.op.instance_name |
8264 |
nodenames = list(instance.all_nodes)
|
8265 |
for node in nodenames: |
8266 |
_CheckNodeOnline(self, node)
|
8267 |
|
8268 |
self.instance = instance
|
8269 |
|
8270 |
if instance.disk_template not in constants.DTS_GROWABLE: |
8271 |
raise errors.OpPrereqError("Instance's disk layout does not support" |
8272 |
" growing.", errors.ECODE_INVAL)
|
8273 |
|
8274 |
self.disk = instance.FindDisk(self.op.disk) |
8275 |
|
8276 |
if instance.disk_template != constants.DT_FILE:
|
8277 |
# TODO: check the free disk space for file, when that feature will be
|
8278 |
# supported
|
8279 |
_CheckNodesFreeDisk(self, nodenames, self.op.amount) |
8280 |
|
8281 |
def Exec(self, feedback_fn): |
8282 |
"""Execute disk grow.
|
8283 |
|
8284 |
"""
|
8285 |
instance = self.instance
|
8286 |
disk = self.disk
|
8287 |
|
8288 |
disks_ok, _ = _AssembleInstanceDisks(self, self.instance, disks=[disk]) |
8289 |
if not disks_ok: |
8290 |
raise errors.OpExecError("Cannot activate block device to grow") |
8291 |
|
8292 |
for node in instance.all_nodes: |
8293 |
self.cfg.SetDiskID(disk, node)
|
8294 |
result = self.rpc.call_blockdev_grow(node, disk, self.op.amount) |
8295 |
result.Raise("Grow request failed to node %s" % node)
|
8296 |
|
8297 |
# TODO: Rewrite code to work properly
|
8298 |
# DRBD goes into sync mode for a short amount of time after executing the
|
8299 |
# "resize" command. DRBD 8.x below version 8.0.13 contains a bug whereby
|
8300 |
# calling "resize" in sync mode fails. Sleeping for a short amount of
|
8301 |
# time is a work-around.
|
8302 |
time.sleep(5)
|
8303 |
|
8304 |
disk.RecordGrow(self.op.amount)
|
8305 |
self.cfg.Update(instance, feedback_fn)
|
8306 |
if self.op.wait_for_sync: |
8307 |
disk_abort = not _WaitForSync(self, instance, disks=[disk]) |
8308 |
if disk_abort:
|
8309 |
self.proc.LogWarning("Warning: disk sync-ing has not returned a good" |
8310 |
" status.\nPlease check the instance.")
|
8311 |
if not instance.admin_up: |
8312 |
_SafeShutdownInstanceDisks(self, instance, disks=[disk])
|
8313 |
elif not instance.admin_up: |
8314 |
self.proc.LogWarning("Not shutting down the disk even if the instance is" |
8315 |
" not supposed to be running because no wait for"
|
8316 |
" sync mode was requested.")
|
8317 |
|
8318 |
|
8319 |
class LUQueryInstanceData(NoHooksLU): |
8320 |
"""Query runtime instance data.
|
8321 |
|
8322 |
"""
|
8323 |
_OP_PARAMS = [ |
8324 |
("instances", _EmptyList, _TListOf(_TNonEmptyString)),
|
8325 |
("static", False, _TBool), |
8326 |
] |
8327 |
REQ_BGL = False
|
8328 |
|
8329 |
def ExpandNames(self): |
8330 |
self.needed_locks = {}
|
8331 |
self.share_locks = dict.fromkeys(locking.LEVELS, 1) |
8332 |
|
8333 |
if self.op.instances: |
8334 |
self.wanted_names = []
|
8335 |
for name in self.op.instances: |
8336 |
full_name = _ExpandInstanceName(self.cfg, name)
|
8337 |
self.wanted_names.append(full_name)
|
8338 |
self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names |
8339 |
else:
|
8340 |
self.wanted_names = None |
8341 |
self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
|
8342 |
|
8343 |
self.needed_locks[locking.LEVEL_NODE] = []
|
8344 |
self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
|
8345 |
|
8346 |
def DeclareLocks(self, level): |
8347 |
if level == locking.LEVEL_NODE:
|
8348 |
self._LockInstancesNodes()
|
8349 |
|
8350 |
def CheckPrereq(self): |
8351 |
"""Check prerequisites.
|
8352 |
|
8353 |
This only checks the optional instance list against the existing names.
|
8354 |
|
8355 |
"""
|
8356 |
if self.wanted_names is None: |
8357 |
self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE] |
8358 |
|
8359 |
self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name |
8360 |
in self.wanted_names] |
8361 |
|
8362 |
def _ComputeBlockdevStatus(self, node, instance_name, dev): |
8363 |
"""Returns the status of a block device
|
8364 |
|
8365 |
"""
|
8366 |
if self.op.static or not node: |
8367 |
return None |
8368 |
|
8369 |
self.cfg.SetDiskID(dev, node)
|
8370 |
|
8371 |
result = self.rpc.call_blockdev_find(node, dev)
|
8372 |
if result.offline:
|
8373 |
return None |
8374 |
|
8375 |
result.Raise("Can't compute disk status for %s" % instance_name)
|
8376 |
|
8377 |
status = result.payload |
8378 |
if status is None: |
8379 |
return None |
8380 |
|
8381 |
return (status.dev_path, status.major, status.minor,
|
8382 |
status.sync_percent, status.estimated_time, |
8383 |
status.is_degraded, status.ldisk_status) |
8384 |
|
8385 |
def _ComputeDiskStatus(self, instance, snode, dev): |
8386 |
"""Compute block device status.
|
8387 |
|
8388 |
"""
|
8389 |
if dev.dev_type in constants.LDS_DRBD: |
8390 |
# we change the snode then (otherwise we use the one passed in)
|
8391 |
if dev.logical_id[0] == instance.primary_node: |
8392 |
snode = dev.logical_id[1]
|
8393 |
else:
|
8394 |
snode = dev.logical_id[0]
|
8395 |
|
8396 |
dev_pstatus = self._ComputeBlockdevStatus(instance.primary_node,
|
8397 |
instance.name, dev) |
8398 |
dev_sstatus = self._ComputeBlockdevStatus(snode, instance.name, dev)
|
8399 |
|
8400 |
if dev.children:
|
8401 |
dev_children = [self._ComputeDiskStatus(instance, snode, child)
|
8402 |
for child in dev.children] |
8403 |
else:
|
8404 |
dev_children = [] |
8405 |
|
8406 |
data = { |
8407 |
"iv_name": dev.iv_name,
|
8408 |
"dev_type": dev.dev_type,
|
8409 |
"logical_id": dev.logical_id,
|
8410 |
"physical_id": dev.physical_id,
|
8411 |
"pstatus": dev_pstatus,
|
8412 |
"sstatus": dev_sstatus,
|
8413 |
"children": dev_children,
|
8414 |
"mode": dev.mode,
|
8415 |
"size": dev.size,
|
8416 |
} |
8417 |
|
8418 |
return data
|
8419 |
|
8420 |
def Exec(self, feedback_fn): |
8421 |
"""Gather and return data"""
|
8422 |
result = {} |
8423 |
|
8424 |
cluster = self.cfg.GetClusterInfo()
|
8425 |
|
8426 |
for instance in self.wanted_instances: |
8427 |
if not self.op.static: |
8428 |
remote_info = self.rpc.call_instance_info(instance.primary_node,
|
8429 |
instance.name, |
8430 |
instance.hypervisor) |
8431 |
remote_info.Raise("Error checking node %s" % instance.primary_node)
|
8432 |
remote_info = remote_info.payload |
8433 |
if remote_info and "state" in remote_info: |
8434 |
remote_state = "up"
|
8435 |
else:
|
8436 |
remote_state = "down"
|
8437 |
else:
|
8438 |
remote_state = None
|
8439 |
if instance.admin_up:
|
8440 |
config_state = "up"
|
8441 |
else:
|
8442 |
config_state = "down"
|
8443 |
|
8444 |
disks = [self._ComputeDiskStatus(instance, None, device) |
8445 |
for device in instance.disks] |
8446 |
|
8447 |
idict = { |
8448 |
"name": instance.name,
|
8449 |
"config_state": config_state,
|
8450 |
"run_state": remote_state,
|
8451 |
"pnode": instance.primary_node,
|
8452 |
"snodes": instance.secondary_nodes,
|
8453 |
"os": instance.os,
|
8454 |
# this happens to be the same format used for hooks
|
8455 |
"nics": _NICListToTuple(self, instance.nics), |
8456 |
"disk_template": instance.disk_template,
|
8457 |
"disks": disks,
|
8458 |
"hypervisor": instance.hypervisor,
|
8459 |
"network_port": instance.network_port,
|
8460 |
"hv_instance": instance.hvparams,
|
8461 |
"hv_actual": cluster.FillHV(instance, skip_globals=True), |
8462 |
"be_instance": instance.beparams,
|
8463 |
"be_actual": cluster.FillBE(instance),
|
8464 |
"os_instance": instance.osparams,
|
8465 |
"os_actual": cluster.SimpleFillOS(instance.os, instance.osparams),
|
8466 |
"serial_no": instance.serial_no,
|
8467 |
"mtime": instance.mtime,
|
8468 |
"ctime": instance.ctime,
|
8469 |
"uuid": instance.uuid,
|
8470 |
} |
8471 |
|
8472 |
result[instance.name] = idict |
8473 |
|
8474 |
return result
|
8475 |
|
8476 |
|
8477 |
class LUSetInstanceParams(LogicalUnit): |
8478 |
"""Modifies an instances's parameters.
|
8479 |
|
8480 |
"""
|
8481 |
HPATH = "instance-modify"
|
8482 |
HTYPE = constants.HTYPE_INSTANCE |
8483 |
_OP_PARAMS = [ |
8484 |
_PInstanceName, |
8485 |
("nics", _EmptyList, _TList),
|
8486 |
("disks", _EmptyList, _TList),
|
8487 |
("beparams", _EmptyDict, _TDict),
|
8488 |
("hvparams", _EmptyDict, _TDict),
|
8489 |
("disk_template", None, _TMaybeString), |
8490 |
("remote_node", None, _TMaybeString), |
8491 |
("os_name", None, _TMaybeString), |
8492 |
("force_variant", False, _TBool), |
8493 |
("osparams", None, _TOr(_TDict, _TNone)), |
8494 |
_PForce, |
8495 |
] |
8496 |
REQ_BGL = False
|
8497 |
|
8498 |
def CheckArguments(self): |
8499 |
if not (self.op.nics or self.op.disks or self.op.disk_template or |
8500 |
self.op.hvparams or self.op.beparams or self.op.os_name): |
8501 |
raise errors.OpPrereqError("No changes submitted", errors.ECODE_INVAL) |
8502 |
|
8503 |
if self.op.hvparams: |
8504 |
_CheckGlobalHvParams(self.op.hvparams)
|
8505 |
|
8506 |
# Disk validation
|
8507 |
disk_addremove = 0
|
8508 |
for disk_op, disk_dict in self.op.disks: |
8509 |
utils.ForceDictType(disk_dict, constants.IDISK_PARAMS_TYPES) |
8510 |
if disk_op == constants.DDM_REMOVE:
|
8511 |
disk_addremove += 1
|
8512 |
continue
|
8513 |
elif disk_op == constants.DDM_ADD:
|
8514 |
disk_addremove += 1
|
8515 |
else:
|
8516 |
if not isinstance(disk_op, int): |
8517 |
raise errors.OpPrereqError("Invalid disk index", errors.ECODE_INVAL) |
8518 |
if not isinstance(disk_dict, dict): |
8519 |
msg = "Invalid disk value: expected dict, got '%s'" % disk_dict
|
8520 |
raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
|
8521 |
|
8522 |
if disk_op == constants.DDM_ADD:
|
8523 |
mode = disk_dict.setdefault('mode', constants.DISK_RDWR)
|
8524 |
if mode not in constants.DISK_ACCESS_SET: |
8525 |
raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode, |
8526 |
errors.ECODE_INVAL) |
8527 |
size = disk_dict.get('size', None) |
8528 |
if size is None: |
8529 |
raise errors.OpPrereqError("Required disk parameter size missing", |
8530 |
errors.ECODE_INVAL) |
8531 |
try:
|
8532 |
size = int(size)
|
8533 |
except (TypeError, ValueError), err: |
8534 |
raise errors.OpPrereqError("Invalid disk size parameter: %s" % |
8535 |
str(err), errors.ECODE_INVAL)
|
8536 |
disk_dict['size'] = size
|
8537 |
else:
|
8538 |
# modification of disk
|
8539 |
if 'size' in disk_dict: |
8540 |
raise errors.OpPrereqError("Disk size change not possible, use" |
8541 |
" grow-disk", errors.ECODE_INVAL)
|
8542 |
|
8543 |
if disk_addremove > 1: |
8544 |
raise errors.OpPrereqError("Only one disk add or remove operation" |
8545 |
" supported at a time", errors.ECODE_INVAL)
|
8546 |
|
8547 |
if self.op.disks and self.op.disk_template is not None: |
8548 |
raise errors.OpPrereqError("Disk template conversion and other disk" |
8549 |
" changes not supported at the same time",
|
8550 |
errors.ECODE_INVAL) |
8551 |
|
8552 |
if self.op.disk_template: |
8553 |
_CheckDiskTemplate(self.op.disk_template)
|
8554 |
if (self.op.disk_template in constants.DTS_NET_MIRROR and |
8555 |
self.op.remote_node is None): |
8556 |
raise errors.OpPrereqError("Changing the disk template to a mirrored" |
8557 |
" one requires specifying a secondary node",
|
8558 |
errors.ECODE_INVAL) |
8559 |
|
8560 |
# NIC validation
|
8561 |
nic_addremove = 0
|
8562 |
for nic_op, nic_dict in self.op.nics: |
8563 |
utils.ForceDictType(nic_dict, constants.INIC_PARAMS_TYPES) |
8564 |
if nic_op == constants.DDM_REMOVE:
|
8565 |
nic_addremove += 1
|
8566 |
continue
|
8567 |
elif nic_op == constants.DDM_ADD:
|
8568 |
nic_addremove += 1
|
8569 |
else:
|
8570 |
if not isinstance(nic_op, int): |
8571 |
raise errors.OpPrereqError("Invalid nic index", errors.ECODE_INVAL) |
8572 |
if not isinstance(nic_dict, dict): |
8573 |
msg = "Invalid nic value: expected dict, got '%s'" % nic_dict
|
8574 |
raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
|
8575 |
|
8576 |
# nic_dict should be a dict
|
8577 |
nic_ip = nic_dict.get('ip', None) |
8578 |
if nic_ip is not None: |
8579 |
if nic_ip.lower() == constants.VALUE_NONE:
|
8580 |
nic_dict['ip'] = None |
8581 |
else:
|
8582 |
if not utils.IsValidIP4(nic_ip): |
8583 |
raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip, |
8584 |
errors.ECODE_INVAL) |
8585 |
|
8586 |
nic_bridge = nic_dict.get('bridge', None) |
8587 |
nic_link = nic_dict.get('link', None) |
8588 |
if nic_bridge and nic_link: |
8589 |
raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'" |
8590 |
" at the same time", errors.ECODE_INVAL)
|
8591 |
elif nic_bridge and nic_bridge.lower() == constants.VALUE_NONE: |
8592 |
nic_dict['bridge'] = None |
8593 |
elif nic_link and nic_link.lower() == constants.VALUE_NONE: |
8594 |
nic_dict['link'] = None |
8595 |
|
8596 |
if nic_op == constants.DDM_ADD:
|
8597 |
nic_mac = nic_dict.get('mac', None) |
8598 |
if nic_mac is None: |
8599 |
nic_dict['mac'] = constants.VALUE_AUTO
|
8600 |
|
8601 |
if 'mac' in nic_dict: |
8602 |
nic_mac = nic_dict['mac']
|
8603 |
if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE): |
8604 |
nic_mac = utils.NormalizeAndValidateMac(nic_mac) |
8605 |
|
8606 |
if nic_op != constants.DDM_ADD and nic_mac == constants.VALUE_AUTO: |
8607 |
raise errors.OpPrereqError("'auto' is not a valid MAC address when" |
8608 |
" modifying an existing nic",
|
8609 |
errors.ECODE_INVAL) |
8610 |
|
8611 |
if nic_addremove > 1: |
8612 |
raise errors.OpPrereqError("Only one NIC add or remove operation" |
8613 |
" supported at a time", errors.ECODE_INVAL)
|
8614 |
|
8615 |
def ExpandNames(self): |
8616 |
self._ExpandAndLockInstance()
|
8617 |
self.needed_locks[locking.LEVEL_NODE] = []
|
8618 |
self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
|
8619 |
|
8620 |
def DeclareLocks(self, level): |
8621 |
if level == locking.LEVEL_NODE:
|
8622 |
self._LockInstancesNodes()
|
8623 |
if self.op.disk_template and self.op.remote_node: |
8624 |
self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node) |
8625 |
self.needed_locks[locking.LEVEL_NODE].append(self.op.remote_node) |
8626 |
|
8627 |
def BuildHooksEnv(self): |
8628 |
"""Build hooks env.
|
8629 |
|
8630 |
This runs on the master, primary and secondaries.
|
8631 |
|
8632 |
"""
|
8633 |
args = dict()
|
8634 |
if constants.BE_MEMORY in self.be_new: |
8635 |
args['memory'] = self.be_new[constants.BE_MEMORY] |
8636 |
if constants.BE_VCPUS in self.be_new: |
8637 |
args['vcpus'] = self.be_new[constants.BE_VCPUS] |
8638 |
# TODO: export disk changes. Note: _BuildInstanceHookEnv* don't export disk
|
8639 |
# information at all.
|
8640 |
if self.op.nics: |
8641 |
args['nics'] = []
|
8642 |
nic_override = dict(self.op.nics) |
8643 |
for idx, nic in enumerate(self.instance.nics): |
8644 |
if idx in nic_override: |
8645 |
this_nic_override = nic_override[idx] |
8646 |
else:
|
8647 |
this_nic_override = {} |
8648 |
if 'ip' in this_nic_override: |
8649 |
ip = this_nic_override['ip']
|
8650 |
else:
|
8651 |
ip = nic.ip |
8652 |
if 'mac' in this_nic_override: |
8653 |
mac = this_nic_override['mac']
|
8654 |
else:
|
8655 |
mac = nic.mac |
8656 |
if idx in self.nic_pnew: |
8657 |
nicparams = self.nic_pnew[idx]
|
8658 |
else:
|
8659 |
nicparams = self.cluster.SimpleFillNIC(nic.nicparams)
|
8660 |
mode = nicparams[constants.NIC_MODE] |
8661 |
link = nicparams[constants.NIC_LINK] |
8662 |
args['nics'].append((ip, mac, mode, link))
|
8663 |
if constants.DDM_ADD in nic_override: |
8664 |
ip = nic_override[constants.DDM_ADD].get('ip', None) |
8665 |
mac = nic_override[constants.DDM_ADD]['mac']
|
8666 |
nicparams = self.nic_pnew[constants.DDM_ADD]
|
8667 |
mode = nicparams[constants.NIC_MODE] |
8668 |
link = nicparams[constants.NIC_LINK] |
8669 |
args['nics'].append((ip, mac, mode, link))
|
8670 |
elif constants.DDM_REMOVE in nic_override: |
8671 |
del args['nics'][-1] |
8672 |
|
8673 |
env = _BuildInstanceHookEnvByObject(self, self.instance, override=args) |
8674 |
if self.op.disk_template: |
8675 |
env["NEW_DISK_TEMPLATE"] = self.op.disk_template |
8676 |
nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes) |
8677 |
return env, nl, nl
|
8678 |
|
8679 |
def CheckPrereq(self): |
8680 |
"""Check prerequisites.
|
8681 |
|
8682 |
This only checks the instance list against the existing names.
|
8683 |
|
8684 |
"""
|
8685 |
# checking the new params on the primary/secondary nodes
|
8686 |
|
8687 |
instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name) |
8688 |
cluster = self.cluster = self.cfg.GetClusterInfo() |
8689 |
assert self.instance is not None, \ |
8690 |
"Cannot retrieve locked instance %s" % self.op.instance_name |
8691 |
pnode = instance.primary_node |
8692 |
nodelist = list(instance.all_nodes)
|
8693 |
|
8694 |
# OS change
|
8695 |
if self.op.os_name and not self.op.force: |
8696 |
_CheckNodeHasOS(self, instance.primary_node, self.op.os_name, |
8697 |
self.op.force_variant)
|
8698 |
instance_os = self.op.os_name
|
8699 |
else:
|
8700 |
instance_os = instance.os |
8701 |
|
8702 |
if self.op.disk_template: |
8703 |
if instance.disk_template == self.op.disk_template: |
8704 |
raise errors.OpPrereqError("Instance already has disk template %s" % |
8705 |
instance.disk_template, errors.ECODE_INVAL) |
8706 |
|
8707 |
if (instance.disk_template,
|
8708 |
self.op.disk_template) not in self._DISK_CONVERSIONS: |
8709 |
raise errors.OpPrereqError("Unsupported disk template conversion from" |
8710 |
" %s to %s" % (instance.disk_template,
|
8711 |
self.op.disk_template),
|
8712 |
errors.ECODE_INVAL) |
8713 |
_CheckInstanceDown(self, instance, "cannot change disk template") |
8714 |
if self.op.disk_template in constants.DTS_NET_MIRROR: |
8715 |
_CheckNodeOnline(self, self.op.remote_node) |
8716 |
_CheckNodeNotDrained(self, self.op.remote_node) |
8717 |
disks = [{"size": d.size} for d in instance.disks] |
8718 |
required = _ComputeDiskSize(self.op.disk_template, disks)
|
8719 |
_CheckNodesFreeDisk(self, [self.op.remote_node], required) |
8720 |
|
8721 |
# hvparams processing
|
8722 |
if self.op.hvparams: |
8723 |
hv_type = instance.hypervisor |
8724 |
i_hvdict = _GetUpdatedParams(instance.hvparams, self.op.hvparams)
|
8725 |
utils.ForceDictType(i_hvdict, constants.HVS_PARAMETER_TYPES) |
8726 |
hv_new = cluster.SimpleFillHV(hv_type, instance.os, i_hvdict) |
8727 |
|
8728 |
# local check
|
8729 |
hypervisor.GetHypervisor(hv_type).CheckParameterSyntax(hv_new) |
8730 |
_CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
|
8731 |
self.hv_new = hv_new # the new actual values |
8732 |
self.hv_inst = i_hvdict # the new dict (without defaults) |
8733 |
else:
|
8734 |
self.hv_new = self.hv_inst = {} |
8735 |
|
8736 |
# beparams processing
|
8737 |
if self.op.beparams: |
8738 |
i_bedict = _GetUpdatedParams(instance.beparams, self.op.beparams,
|
8739 |
use_none=True)
|
8740 |
utils.ForceDictType(i_bedict, constants.BES_PARAMETER_TYPES) |
8741 |
be_new = cluster.SimpleFillBE(i_bedict) |
8742 |
self.be_new = be_new # the new actual values |
8743 |
self.be_inst = i_bedict # the new dict (without defaults) |
8744 |
else:
|
8745 |
self.be_new = self.be_inst = {} |
8746 |
|
8747 |
# osparams processing
|
8748 |
if self.op.osparams: |
8749 |
i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
|
8750 |
_CheckOSParams(self, True, nodelist, instance_os, i_osdict) |
8751 |
self.os_new = cluster.SimpleFillOS(instance_os, i_osdict)
|
8752 |
self.os_inst = i_osdict # the new dict (without defaults) |
8753 |
else:
|
8754 |
self.os_new = self.os_inst = {} |
8755 |
|
8756 |
self.warn = []
|
8757 |
|
8758 |
if constants.BE_MEMORY in self.op.beparams and not self.op.force: |
8759 |
mem_check_list = [pnode] |
8760 |
if be_new[constants.BE_AUTO_BALANCE]:
|
8761 |
# either we changed auto_balance to yes or it was from before
|
8762 |
mem_check_list.extend(instance.secondary_nodes) |
8763 |
instance_info = self.rpc.call_instance_info(pnode, instance.name,
|
8764 |
instance.hypervisor) |
8765 |
nodeinfo = self.rpc.call_node_info(mem_check_list, self.cfg.GetVGName(), |
8766 |
instance.hypervisor) |
8767 |
pninfo = nodeinfo[pnode] |
8768 |
msg = pninfo.fail_msg |
8769 |
if msg:
|
8770 |
# Assume the primary node is unreachable and go ahead
|
8771 |
self.warn.append("Can't get info from primary node %s: %s" % |
8772 |
(pnode, msg)) |
8773 |
elif not isinstance(pninfo.payload.get('memory_free', None), int): |
8774 |
self.warn.append("Node data from primary node %s doesn't contain" |
8775 |
" free memory information" % pnode)
|
8776 |
elif instance_info.fail_msg:
|
8777 |
self.warn.append("Can't get instance runtime information: %s" % |
8778 |
instance_info.fail_msg) |
8779 |
else:
|
8780 |
if instance_info.payload:
|
8781 |
current_mem = int(instance_info.payload['memory']) |
8782 |
else:
|
8783 |
# Assume instance not running
|
8784 |
# (there is a slight race condition here, but it's not very probable,
|
8785 |
# and we have no other way to check)
|
8786 |
current_mem = 0
|
8787 |
miss_mem = (be_new[constants.BE_MEMORY] - current_mem - |
8788 |
pninfo.payload['memory_free'])
|
8789 |
if miss_mem > 0: |
8790 |
raise errors.OpPrereqError("This change will prevent the instance" |
8791 |
" from starting, due to %d MB of memory"
|
8792 |
" missing on its primary node" % miss_mem,
|
8793 |
errors.ECODE_NORES) |
8794 |
|
8795 |
if be_new[constants.BE_AUTO_BALANCE]:
|
8796 |
for node, nres in nodeinfo.items(): |
8797 |
if node not in instance.secondary_nodes: |
8798 |
continue
|
8799 |
msg = nres.fail_msg |
8800 |
if msg:
|
8801 |
self.warn.append("Can't get info from secondary node %s: %s" % |
8802 |
(node, msg)) |
8803 |
elif not isinstance(nres.payload.get('memory_free', None), int): |
8804 |
self.warn.append("Secondary node %s didn't return free" |
8805 |
" memory information" % node)
|
8806 |
elif be_new[constants.BE_MEMORY] > nres.payload['memory_free']: |
8807 |
self.warn.append("Not enough memory to failover instance to" |
8808 |
" secondary node %s" % node)
|
8809 |
|
8810 |
# NIC processing
|
8811 |
self.nic_pnew = {}
|
8812 |
self.nic_pinst = {}
|
8813 |
for nic_op, nic_dict in self.op.nics: |
8814 |
if nic_op == constants.DDM_REMOVE:
|
8815 |
if not instance.nics: |
8816 |
raise errors.OpPrereqError("Instance has no NICs, cannot remove", |
8817 |
errors.ECODE_INVAL) |
8818 |
continue
|
8819 |
if nic_op != constants.DDM_ADD:
|
8820 |
# an existing nic
|
8821 |
if not instance.nics: |
8822 |
raise errors.OpPrereqError("Invalid NIC index %s, instance has" |
8823 |
" no NICs" % nic_op,
|
8824 |
errors.ECODE_INVAL) |
8825 |
if nic_op < 0 or nic_op >= len(instance.nics): |
8826 |
raise errors.OpPrereqError("Invalid NIC index %s, valid values" |
8827 |
" are 0 to %d" %
|
8828 |
(nic_op, len(instance.nics) - 1), |
8829 |
errors.ECODE_INVAL) |
8830 |
old_nic_params = instance.nics[nic_op].nicparams |
8831 |
old_nic_ip = instance.nics[nic_op].ip |
8832 |
else:
|
8833 |
old_nic_params = {} |
8834 |
old_nic_ip = None
|
8835 |
|
8836 |
update_params_dict = dict([(key, nic_dict[key])
|
8837 |
for key in constants.NICS_PARAMETERS |
8838 |
if key in nic_dict]) |
8839 |
|
8840 |
if 'bridge' in nic_dict: |
8841 |
update_params_dict[constants.NIC_LINK] = nic_dict['bridge']
|
8842 |
|
8843 |
new_nic_params = _GetUpdatedParams(old_nic_params, |
8844 |
update_params_dict) |
8845 |
utils.ForceDictType(new_nic_params, constants.NICS_PARAMETER_TYPES) |
8846 |
new_filled_nic_params = cluster.SimpleFillNIC(new_nic_params) |
8847 |
objects.NIC.CheckParameterSyntax(new_filled_nic_params) |
8848 |
self.nic_pinst[nic_op] = new_nic_params
|
8849 |
self.nic_pnew[nic_op] = new_filled_nic_params
|
8850 |
new_nic_mode = new_filled_nic_params[constants.NIC_MODE] |
8851 |
|
8852 |
if new_nic_mode == constants.NIC_MODE_BRIDGED:
|
8853 |
nic_bridge = new_filled_nic_params[constants.NIC_LINK] |
8854 |
msg = self.rpc.call_bridges_exist(pnode, [nic_bridge]).fail_msg
|
8855 |
if msg:
|
8856 |
msg = "Error checking bridges on node %s: %s" % (pnode, msg)
|
8857 |
if self.op.force: |
8858 |
self.warn.append(msg)
|
8859 |
else:
|
8860 |
raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
|
8861 |
if new_nic_mode == constants.NIC_MODE_ROUTED:
|
8862 |
if 'ip' in nic_dict: |
8863 |
nic_ip = nic_dict['ip']
|
8864 |
else:
|
8865 |
nic_ip = old_nic_ip |
8866 |
if nic_ip is None: |
8867 |
raise errors.OpPrereqError('Cannot set the nic ip to None' |
8868 |
' on a routed nic', errors.ECODE_INVAL)
|
8869 |
if 'mac' in nic_dict: |
8870 |
nic_mac = nic_dict['mac']
|
8871 |
if nic_mac is None: |
8872 |
raise errors.OpPrereqError('Cannot set the nic mac to None', |
8873 |
errors.ECODE_INVAL) |
8874 |
elif nic_mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE): |
8875 |
# otherwise generate the mac
|
8876 |
nic_dict['mac'] = self.cfg.GenerateMAC(self.proc.GetECId()) |
8877 |
else:
|
8878 |
# or validate/reserve the current one
|
8879 |
try:
|
8880 |
self.cfg.ReserveMAC(nic_mac, self.proc.GetECId()) |
8881 |
except errors.ReservationError:
|
8882 |
raise errors.OpPrereqError("MAC address %s already in use" |
8883 |
" in cluster" % nic_mac,
|
8884 |
errors.ECODE_NOTUNIQUE) |
8885 |
|
8886 |
# DISK processing
|
8887 |
if self.op.disks and instance.disk_template == constants.DT_DISKLESS: |
8888 |
raise errors.OpPrereqError("Disk operations not supported for" |
8889 |
" diskless instances",
|
8890 |
errors.ECODE_INVAL) |
8891 |
for disk_op, _ in self.op.disks: |
8892 |
if disk_op == constants.DDM_REMOVE:
|
8893 |
if len(instance.disks) == 1: |
8894 |
raise errors.OpPrereqError("Cannot remove the last disk of" |
8895 |
" an instance", errors.ECODE_INVAL)
|
8896 |
_CheckInstanceDown(self, instance, "cannot remove disks") |
8897 |
|
8898 |
if (disk_op == constants.DDM_ADD and |
8899 |
len(instance.nics) >= constants.MAX_DISKS):
|
8900 |
raise errors.OpPrereqError("Instance has too many disks (%d), cannot" |
8901 |
" add more" % constants.MAX_DISKS,
|
8902 |
errors.ECODE_STATE) |
8903 |
if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE): |
8904 |
# an existing disk
|
8905 |
if disk_op < 0 or disk_op >= len(instance.disks): |
8906 |
raise errors.OpPrereqError("Invalid disk index %s, valid values" |
8907 |
" are 0 to %d" %
|
8908 |
(disk_op, len(instance.disks)),
|
8909 |
errors.ECODE_INVAL) |
8910 |
|
8911 |
return
|
8912 |
|
8913 |
def _ConvertPlainToDrbd(self, feedback_fn): |
8914 |
"""Converts an instance from plain to drbd.
|
8915 |
|
8916 |
"""
|
8917 |
feedback_fn("Converting template to drbd")
|
8918 |
instance = self.instance
|
8919 |
pnode = instance.primary_node |
8920 |
snode = self.op.remote_node
|
8921 |
|
8922 |
# create a fake disk info for _GenerateDiskTemplate
|
8923 |
disk_info = [{"size": d.size, "mode": d.mode} for d in instance.disks] |
8924 |
new_disks = _GenerateDiskTemplate(self, self.op.disk_template, |
8925 |
instance.name, pnode, [snode], |
8926 |
disk_info, None, None, 0) |
8927 |
info = _GetInstanceInfoText(instance) |
8928 |
feedback_fn("Creating aditional volumes...")
|
8929 |
# first, create the missing data and meta devices
|
8930 |
for disk in new_disks: |
8931 |
# unfortunately this is... not too nice
|
8932 |
_CreateSingleBlockDev(self, pnode, instance, disk.children[1], |
8933 |
info, True)
|
8934 |
for child in disk.children: |
8935 |
_CreateSingleBlockDev(self, snode, instance, child, info, True) |
8936 |
# at this stage, all new LVs have been created, we can rename the
|
8937 |
# old ones
|
8938 |
feedback_fn("Renaming original volumes...")
|
8939 |
rename_list = [(o, n.children[0].logical_id)
|
8940 |
for (o, n) in zip(instance.disks, new_disks)] |
8941 |
result = self.rpc.call_blockdev_rename(pnode, rename_list)
|
8942 |
result.Raise("Failed to rename original LVs")
|
8943 |
|
8944 |
feedback_fn("Initializing DRBD devices...")
|
8945 |
# all child devices are in place, we can now create the DRBD devices
|
8946 |
for disk in new_disks: |
8947 |
for node in [pnode, snode]: |
8948 |
f_create = node == pnode |
8949 |
_CreateSingleBlockDev(self, node, instance, disk, info, f_create)
|
8950 |
|
8951 |
# at this point, the instance has been modified
|
8952 |
instance.disk_template = constants.DT_DRBD8 |
8953 |
instance.disks = new_disks |
8954 |
self.cfg.Update(instance, feedback_fn)
|
8955 |
|
8956 |
# disks are created, waiting for sync
|
8957 |
disk_abort = not _WaitForSync(self, instance) |
8958 |
if disk_abort:
|
8959 |
raise errors.OpExecError("There are some degraded disks for" |
8960 |
" this instance, please cleanup manually")
|
8961 |
|
8962 |
def _ConvertDrbdToPlain(self, feedback_fn): |
8963 |
"""Converts an instance from drbd to plain.
|
8964 |
|
8965 |
"""
|
8966 |
instance = self.instance
|
8967 |
assert len(instance.secondary_nodes) == 1 |
8968 |
pnode = instance.primary_node |
8969 |
snode = instance.secondary_nodes[0]
|
8970 |
feedback_fn("Converting template to plain")
|
8971 |
|
8972 |
old_disks = instance.disks |
8973 |
new_disks = [d.children[0] for d in old_disks] |
8974 |
|
8975 |
# copy over size and mode
|
8976 |
for parent, child in zip(old_disks, new_disks): |
8977 |
child.size = parent.size |
8978 |
child.mode = parent.mode |
8979 |
|
8980 |
# update instance structure
|
8981 |
instance.disks = new_disks |
8982 |
instance.disk_template = constants.DT_PLAIN |
8983 |
self.cfg.Update(instance, feedback_fn)
|
8984 |
|
8985 |
feedback_fn("Removing volumes on the secondary node...")
|
8986 |
for disk in old_disks: |
8987 |
self.cfg.SetDiskID(disk, snode)
|
8988 |
msg = self.rpc.call_blockdev_remove(snode, disk).fail_msg
|
8989 |
if msg:
|
8990 |
self.LogWarning("Could not remove block device %s on node %s," |
8991 |
" continuing anyway: %s", disk.iv_name, snode, msg)
|
8992 |
|
8993 |
feedback_fn("Removing unneeded volumes on the primary node...")
|
8994 |
for idx, disk in enumerate(old_disks): |
8995 |
meta = disk.children[1]
|
8996 |
self.cfg.SetDiskID(meta, pnode)
|
8997 |
msg = self.rpc.call_blockdev_remove(pnode, meta).fail_msg
|
8998 |
if msg:
|
8999 |
self.LogWarning("Could not remove metadata for disk %d on node %s," |
9000 |
" continuing anyway: %s", idx, pnode, msg)
|
9001 |
|
9002 |
|
9003 |
def Exec(self, feedback_fn): |
9004 |
"""Modifies an instance.
|
9005 |
|
9006 |
All parameters take effect only at the next restart of the instance.
|
9007 |
|
9008 |
"""
|
9009 |
# Process here the warnings from CheckPrereq, as we don't have a
|
9010 |
# feedback_fn there.
|
9011 |
for warn in self.warn: |
9012 |
feedback_fn("WARNING: %s" % warn)
|
9013 |
|
9014 |
result = [] |
9015 |
instance = self.instance
|
9016 |
# disk changes
|
9017 |
for disk_op, disk_dict in self.op.disks: |
9018 |
if disk_op == constants.DDM_REMOVE:
|
9019 |
# remove the last disk
|
9020 |
device = instance.disks.pop() |
9021 |
device_idx = len(instance.disks)
|
9022 |
for node, disk in device.ComputeNodeTree(instance.primary_node): |
9023 |
self.cfg.SetDiskID(disk, node)
|
9024 |
msg = self.rpc.call_blockdev_remove(node, disk).fail_msg
|
9025 |
if msg:
|
9026 |
self.LogWarning("Could not remove disk/%d on node %s: %s," |
9027 |
" continuing anyway", device_idx, node, msg)
|
9028 |
result.append(("disk/%d" % device_idx, "remove")) |
9029 |
elif disk_op == constants.DDM_ADD:
|
9030 |
# add a new disk
|
9031 |
if instance.disk_template == constants.DT_FILE:
|
9032 |
file_driver, file_path = instance.disks[0].logical_id
|
9033 |
file_path = os.path.dirname(file_path) |
9034 |
else:
|
9035 |
file_driver = file_path = None
|
9036 |
disk_idx_base = len(instance.disks)
|
9037 |
new_disk = _GenerateDiskTemplate(self,
|
9038 |
instance.disk_template, |
9039 |
instance.name, instance.primary_node, |
9040 |
instance.secondary_nodes, |
9041 |
[disk_dict], |
9042 |
file_path, |
9043 |
file_driver, |
9044 |
disk_idx_base)[0]
|
9045 |
instance.disks.append(new_disk) |
9046 |
info = _GetInstanceInfoText(instance) |
9047 |
|
9048 |
logging.info("Creating volume %s for instance %s",
|
9049 |
new_disk.iv_name, instance.name) |
9050 |
# Note: this needs to be kept in sync with _CreateDisks
|
9051 |
#HARDCODE
|
9052 |
for node in instance.all_nodes: |
9053 |
f_create = node == instance.primary_node |
9054 |
try:
|
9055 |
_CreateBlockDev(self, node, instance, new_disk,
|
9056 |
f_create, info, f_create) |
9057 |
except errors.OpExecError, err:
|
9058 |
self.LogWarning("Failed to create volume %s (%s) on" |
9059 |
" node %s: %s",
|
9060 |
new_disk.iv_name, new_disk, node, err) |
9061 |
result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" % |
9062 |
(new_disk.size, new_disk.mode))) |
9063 |
else:
|
9064 |
# change a given disk
|
9065 |
instance.disks[disk_op].mode = disk_dict['mode']
|
9066 |
result.append(("disk.mode/%d" % disk_op, disk_dict['mode'])) |
9067 |
|
9068 |
if self.op.disk_template: |
9069 |
r_shut = _ShutdownInstanceDisks(self, instance)
|
9070 |
if not r_shut: |
9071 |
raise errors.OpExecError("Cannot shutdow instance disks, unable to" |
9072 |
" proceed with disk template conversion")
|
9073 |
mode = (instance.disk_template, self.op.disk_template)
|
9074 |
try:
|
9075 |
self._DISK_CONVERSIONS[mode](self, feedback_fn) |
9076 |
except:
|
9077 |
self.cfg.ReleaseDRBDMinors(instance.name)
|
9078 |
raise
|
9079 |
result.append(("disk_template", self.op.disk_template)) |
9080 |
|
9081 |
# NIC changes
|
9082 |
for nic_op, nic_dict in self.op.nics: |
9083 |
if nic_op == constants.DDM_REMOVE:
|
9084 |
# remove the last nic
|
9085 |
del instance.nics[-1] |
9086 |
result.append(("nic.%d" % len(instance.nics), "remove")) |
9087 |
elif nic_op == constants.DDM_ADD:
|
9088 |
# mac and bridge should be set, by now
|
9089 |
mac = nic_dict['mac']
|
9090 |
ip = nic_dict.get('ip', None) |
9091 |
nicparams = self.nic_pinst[constants.DDM_ADD]
|
9092 |
new_nic = objects.NIC(mac=mac, ip=ip, nicparams=nicparams) |
9093 |
instance.nics.append(new_nic) |
9094 |
result.append(("nic.%d" % (len(instance.nics) - 1), |
9095 |
"add:mac=%s,ip=%s,mode=%s,link=%s" %
|
9096 |
(new_nic.mac, new_nic.ip, |
9097 |
self.nic_pnew[constants.DDM_ADD][constants.NIC_MODE],
|
9098 |
self.nic_pnew[constants.DDM_ADD][constants.NIC_LINK]
|
9099 |
))) |
9100 |
else:
|
9101 |
for key in 'mac', 'ip': |
9102 |
if key in nic_dict: |
9103 |
setattr(instance.nics[nic_op], key, nic_dict[key])
|
9104 |
if nic_op in self.nic_pinst: |
9105 |
instance.nics[nic_op].nicparams = self.nic_pinst[nic_op]
|
9106 |
for key, val in nic_dict.iteritems(): |
9107 |
result.append(("nic.%s/%d" % (key, nic_op), val))
|
9108 |
|
9109 |
# hvparams changes
|
9110 |
if self.op.hvparams: |
9111 |
instance.hvparams = self.hv_inst
|
9112 |
for key, val in self.op.hvparams.iteritems(): |
9113 |
result.append(("hv/%s" % key, val))
|
9114 |
|
9115 |
# beparams changes
|
9116 |
if self.op.beparams: |
9117 |
instance.beparams = self.be_inst
|
9118 |
for key, val in self.op.beparams.iteritems(): |
9119 |
result.append(("be/%s" % key, val))
|
9120 |
|
9121 |
# OS change
|
9122 |
if self.op.os_name: |
9123 |
instance.os = self.op.os_name
|
9124 |
|
9125 |
# osparams changes
|
9126 |
if self.op.osparams: |
9127 |
instance.osparams = self.os_inst
|
9128 |
for key, val in self.op.osparams.iteritems(): |
9129 |
result.append(("os/%s" % key, val))
|
9130 |
|
9131 |
self.cfg.Update(instance, feedback_fn)
|
9132 |
|
9133 |
return result
|
9134 |
|
9135 |
_DISK_CONVERSIONS = { |
9136 |
(constants.DT_PLAIN, constants.DT_DRBD8): _ConvertPlainToDrbd, |
9137 |
(constants.DT_DRBD8, constants.DT_PLAIN): _ConvertDrbdToPlain, |
9138 |
} |
9139 |
|
9140 |
|
9141 |
class LUQueryExports(NoHooksLU): |
9142 |
"""Query the exports list
|
9143 |
|
9144 |
"""
|
9145 |
_OP_PARAMS = [ |
9146 |
("nodes", _EmptyList, _TListOf(_TNonEmptyString)),
|
9147 |
("use_locking", False, _TBool), |
9148 |
] |
9149 |
REQ_BGL = False
|
9150 |
|
9151 |
def ExpandNames(self): |
9152 |
self.needed_locks = {}
|
9153 |
self.share_locks[locking.LEVEL_NODE] = 1 |
9154 |
if not self.op.nodes: |
9155 |
self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
|
9156 |
else:
|
9157 |
self.needed_locks[locking.LEVEL_NODE] = \
|
9158 |
_GetWantedNodes(self, self.op.nodes) |
9159 |
|
9160 |
def Exec(self, feedback_fn): |
9161 |
"""Compute the list of all the exported system images.
|
9162 |
|
9163 |
@rtype: dict
|
9164 |
@return: a dictionary with the structure node->(export-list)
|
9165 |
where export-list is a list of the instances exported on
|
9166 |
that node.
|
9167 |
|
9168 |
"""
|
9169 |
self.nodes = self.acquired_locks[locking.LEVEL_NODE] |
9170 |
rpcresult = self.rpc.call_export_list(self.nodes) |
9171 |
result = {} |
9172 |
for node in rpcresult: |
9173 |
if rpcresult[node].fail_msg:
|
9174 |
result[node] = False
|
9175 |
else:
|
9176 |
result[node] = rpcresult[node].payload |
9177 |
|
9178 |
return result
|
9179 |
|
9180 |
|
9181 |
class LUPrepareExport(NoHooksLU): |
9182 |
"""Prepares an instance for an export and returns useful information.
|
9183 |
|
9184 |
"""
|
9185 |
_OP_PARAMS = [ |
9186 |
_PInstanceName, |
9187 |
("mode", _NoDefault, _TElemOf(constants.EXPORT_MODES)),
|
9188 |
] |
9189 |
REQ_BGL = False
|
9190 |
|
9191 |
def ExpandNames(self): |
9192 |
self._ExpandAndLockInstance()
|
9193 |
|
9194 |
def CheckPrereq(self): |
9195 |
"""Check prerequisites.
|
9196 |
|
9197 |
"""
|
9198 |
instance_name = self.op.instance_name
|
9199 |
|
9200 |
self.instance = self.cfg.GetInstanceInfo(instance_name) |
9201 |
assert self.instance is not None, \ |
9202 |
"Cannot retrieve locked instance %s" % self.op.instance_name |
9203 |
_CheckNodeOnline(self, self.instance.primary_node) |
9204 |
|
9205 |
self._cds = _GetClusterDomainSecret()
|
9206 |
|
9207 |
def Exec(self, feedback_fn): |
9208 |
"""Prepares an instance for an export.
|
9209 |
|
9210 |
"""
|
9211 |
instance = self.instance
|
9212 |
|
9213 |
if self.op.mode == constants.EXPORT_MODE_REMOTE: |
9214 |
salt = utils.GenerateSecret(8)
|
9215 |
|
9216 |
feedback_fn("Generating X509 certificate on %s" % instance.primary_node)
|
9217 |
result = self.rpc.call_x509_cert_create(instance.primary_node,
|
9218 |
constants.RIE_CERT_VALIDITY) |
9219 |
result.Raise("Can't create X509 key and certificate on %s" % result.node)
|
9220 |
|
9221 |
(name, cert_pem) = result.payload |
9222 |
|
9223 |
cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM, |
9224 |
cert_pem) |
9225 |
|
9226 |
return {
|
9227 |
"handshake": masterd.instance.ComputeRemoteExportHandshake(self._cds), |
9228 |
"x509_key_name": (name, utils.Sha1Hmac(self._cds, name, salt=salt), |
9229 |
salt), |
9230 |
"x509_ca": utils.SignX509Certificate(cert, self._cds, salt), |
9231 |
} |
9232 |
|
9233 |
return None |
9234 |
|
9235 |
|
9236 |
class LUExportInstance(LogicalUnit): |
9237 |
"""Export an instance to an image in the cluster.
|
9238 |
|
9239 |
"""
|
9240 |
HPATH = "instance-export"
|
9241 |
HTYPE = constants.HTYPE_INSTANCE |
9242 |
_OP_PARAMS = [ |
9243 |
_PInstanceName, |
9244 |
("target_node", _NoDefault, _TOr(_TNonEmptyString, _TList)),
|
9245 |
("shutdown", True, _TBool), |
9246 |
_PShutdownTimeout, |
9247 |
("remove_instance", False, _TBool), |
9248 |
("ignore_remove_failures", False, _TBool), |
9249 |
("mode", constants.EXPORT_MODE_LOCAL, _TElemOf(constants.EXPORT_MODES)),
|
9250 |
("x509_key_name", None, _TOr(_TList, _TNone)), |
9251 |
("destination_x509_ca", None, _TMaybeString), |
9252 |
] |
9253 |
REQ_BGL = False
|
9254 |
|
9255 |
def CheckArguments(self): |
9256 |
"""Check the arguments.
|
9257 |
|
9258 |
"""
|
9259 |
self.x509_key_name = self.op.x509_key_name |
9260 |
self.dest_x509_ca_pem = self.op.destination_x509_ca |
9261 |
|
9262 |
if self.op.remove_instance and not self.op.shutdown: |
9263 |
raise errors.OpPrereqError("Can not remove instance without shutting it" |
9264 |
" down before")
|
9265 |
|
9266 |
if self.op.mode == constants.EXPORT_MODE_REMOTE: |
9267 |
if not self.x509_key_name: |
9268 |
raise errors.OpPrereqError("Missing X509 key name for encryption", |
9269 |
errors.ECODE_INVAL) |
9270 |
|
9271 |
if not self.dest_x509_ca_pem: |
9272 |
raise errors.OpPrereqError("Missing destination X509 CA", |
9273 |
errors.ECODE_INVAL) |
9274 |
|
9275 |
def ExpandNames(self): |
9276 |
self._ExpandAndLockInstance()
|
9277 |
|
9278 |
# Lock all nodes for local exports
|
9279 |
if self.op.mode == constants.EXPORT_MODE_LOCAL: |
9280 |
# FIXME: lock only instance primary and destination node
|
9281 |
#
|
9282 |
# Sad but true, for now we have do lock all nodes, as we don't know where
|
9283 |
# the previous export might be, and in this LU we search for it and
|
9284 |
# remove it from its current node. In the future we could fix this by:
|
9285 |
# - making a tasklet to search (share-lock all), then create the
|
9286 |
# new one, then one to remove, after
|
9287 |
# - removing the removal operation altogether
|
9288 |
self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
|
9289 |
|
9290 |
def DeclareLocks(self, level): |
9291 |
"""Last minute lock declaration."""
|
9292 |
# All nodes are locked anyway, so nothing to do here.
|
9293 |
|
9294 |
def BuildHooksEnv(self): |
9295 |
"""Build hooks env.
|
9296 |
|
9297 |
This will run on the master, primary node and target node.
|
9298 |
|
9299 |
"""
|
9300 |
env = { |
9301 |
"EXPORT_MODE": self.op.mode, |
9302 |
"EXPORT_NODE": self.op.target_node, |
9303 |
"EXPORT_DO_SHUTDOWN": self.op.shutdown, |
9304 |
"SHUTDOWN_TIMEOUT": self.op.shutdown_timeout, |
9305 |
# TODO: Generic function for boolean env variables
|
9306 |
"REMOVE_INSTANCE": str(bool(self.op.remove_instance)), |
9307 |
} |
9308 |
|
9309 |
env.update(_BuildInstanceHookEnvByObject(self, self.instance)) |
9310 |
|
9311 |
nl = [self.cfg.GetMasterNode(), self.instance.primary_node] |
9312 |
|
9313 |
if self.op.mode == constants.EXPORT_MODE_LOCAL: |
9314 |
nl.append(self.op.target_node)
|
9315 |
|
9316 |
return env, nl, nl
|
9317 |
|
9318 |
def CheckPrereq(self): |
9319 |
"""Check prerequisites.
|
9320 |
|
9321 |
This checks that the instance and node names are valid.
|
9322 |
|
9323 |
"""
|
9324 |
instance_name = self.op.instance_name
|
9325 |
|
9326 |
self.instance = self.cfg.GetInstanceInfo(instance_name) |
9327 |
assert self.instance is not None, \ |
9328 |
"Cannot retrieve locked instance %s" % self.op.instance_name |
9329 |
_CheckNodeOnline(self, self.instance.primary_node) |
9330 |
|
9331 |
if self.op.mode == constants.EXPORT_MODE_LOCAL: |
9332 |
self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node) |
9333 |
self.dst_node = self.cfg.GetNodeInfo(self.op.target_node) |
9334 |
assert self.dst_node is not None |
9335 |
|
9336 |
_CheckNodeOnline(self, self.dst_node.name) |
9337 |
_CheckNodeNotDrained(self, self.dst_node.name) |
9338 |
|
9339 |
self._cds = None |
9340 |
self.dest_disk_info = None |
9341 |
self.dest_x509_ca = None |
9342 |
|
9343 |
elif self.op.mode == constants.EXPORT_MODE_REMOTE: |
9344 |
self.dst_node = None |
9345 |
|
9346 |
if len(self.op.target_node) != len(self.instance.disks): |
9347 |
raise errors.OpPrereqError(("Received destination information for %s" |
9348 |
" disks, but instance %s has %s disks") %
|
9349 |
(len(self.op.target_node), instance_name, |
9350 |
len(self.instance.disks)), |
9351 |
errors.ECODE_INVAL) |
9352 |
|
9353 |
cds = _GetClusterDomainSecret() |
9354 |
|
9355 |
# Check X509 key name
|
9356 |
try:
|
9357 |
(key_name, hmac_digest, hmac_salt) = self.x509_key_name
|
9358 |
except (TypeError, ValueError), err: |
9359 |
raise errors.OpPrereqError("Invalid data for X509 key name: %s" % err) |
9360 |
|
9361 |
if not utils.VerifySha1Hmac(cds, key_name, hmac_digest, salt=hmac_salt): |
9362 |
raise errors.OpPrereqError("HMAC for X509 key name is wrong", |
9363 |
errors.ECODE_INVAL) |
9364 |
|
9365 |
# Load and verify CA
|
9366 |
try:
|
9367 |
(cert, _) = utils.LoadSignedX509Certificate(self.dest_x509_ca_pem, cds)
|
9368 |
except OpenSSL.crypto.Error, err:
|
9369 |
raise errors.OpPrereqError("Unable to load destination X509 CA (%s)" % |
9370 |
(err, ), errors.ECODE_INVAL) |
9371 |
|
9372 |
(errcode, msg) = utils.VerifyX509Certificate(cert, None, None) |
9373 |
if errcode is not None: |
9374 |
raise errors.OpPrereqError("Invalid destination X509 CA (%s)" % |
9375 |
(msg, ), errors.ECODE_INVAL) |
9376 |
|
9377 |
self.dest_x509_ca = cert
|
9378 |
|
9379 |
# Verify target information
|
9380 |
disk_info = [] |
9381 |
for idx, disk_data in enumerate(self.op.target_node): |
9382 |
try:
|
9383 |
(host, port, magic) = \ |
9384 |
masterd.instance.CheckRemoteExportDiskInfo(cds, idx, disk_data) |
9385 |
except errors.GenericError, err:
|
9386 |
raise errors.OpPrereqError("Target info for disk %s: %s" % |
9387 |
(idx, err), errors.ECODE_INVAL) |
9388 |
|
9389 |
disk_info.append((host, port, magic)) |
9390 |
|
9391 |
assert len(disk_info) == len(self.op.target_node) |
9392 |
self.dest_disk_info = disk_info
|
9393 |
|
9394 |
else:
|
9395 |
raise errors.ProgrammerError("Unhandled export mode %r" % |
9396 |
self.op.mode)
|
9397 |
|
9398 |
# instance disk type verification
|
9399 |
# TODO: Implement export support for file-based disks
|
9400 |
for disk in self.instance.disks: |
9401 |
if disk.dev_type == constants.LD_FILE:
|
9402 |
raise errors.OpPrereqError("Export not supported for instances with" |
9403 |
" file-based disks", errors.ECODE_INVAL)
|
9404 |
|
9405 |
def _CleanupExports(self, feedback_fn): |
9406 |
"""Removes exports of current instance from all other nodes.
|
9407 |
|
9408 |
If an instance in a cluster with nodes A..D was exported to node C, its
|
9409 |
exports will be removed from the nodes A, B and D.
|
9410 |
|
9411 |
"""
|
9412 |
assert self.op.mode != constants.EXPORT_MODE_REMOTE |
9413 |
|
9414 |
nodelist = self.cfg.GetNodeList()
|
9415 |
nodelist.remove(self.dst_node.name)
|
9416 |
|
9417 |
# on one-node clusters nodelist will be empty after the removal
|
9418 |
# if we proceed the backup would be removed because OpQueryExports
|
9419 |
# substitutes an empty list with the full cluster node list.
|
9420 |
iname = self.instance.name
|
9421 |
if nodelist:
|
9422 |
feedback_fn("Removing old exports for instance %s" % iname)
|
9423 |
exportlist = self.rpc.call_export_list(nodelist)
|
9424 |
for node in exportlist: |
9425 |
if exportlist[node].fail_msg:
|
9426 |
continue
|
9427 |
if iname in exportlist[node].payload: |
9428 |
msg = self.rpc.call_export_remove(node, iname).fail_msg
|
9429 |
if msg:
|
9430 |
self.LogWarning("Could not remove older export for instance %s" |
9431 |
" on node %s: %s", iname, node, msg)
|
9432 |
|
9433 |
def Exec(self, feedback_fn): |
9434 |
"""Export an instance to an image in the cluster.
|
9435 |
|
9436 |
"""
|
9437 |
assert self.op.mode in constants.EXPORT_MODES |
9438 |
|
9439 |
instance = self.instance
|
9440 |
src_node = instance.primary_node |
9441 |
|
9442 |
if self.op.shutdown: |
9443 |
# shutdown the instance, but not the disks
|
9444 |
feedback_fn("Shutting down instance %s" % instance.name)
|
9445 |
result = self.rpc.call_instance_shutdown(src_node, instance,
|
9446 |
self.op.shutdown_timeout)
|
9447 |
# TODO: Maybe ignore failures if ignore_remove_failures is set
|
9448 |
result.Raise("Could not shutdown instance %s on"
|
9449 |
" node %s" % (instance.name, src_node))
|
9450 |
|
9451 |
# set the disks ID correctly since call_instance_start needs the
|
9452 |
# correct drbd minor to create the symlinks
|
9453 |
for disk in instance.disks: |
9454 |
self.cfg.SetDiskID(disk, src_node)
|
9455 |
|
9456 |
activate_disks = (not instance.admin_up)
|
9457 |
|
9458 |
if activate_disks:
|
9459 |
# Activate the instance disks if we'exporting a stopped instance
|
9460 |
feedback_fn("Activating disks for %s" % instance.name)
|
9461 |
_StartInstanceDisks(self, instance, None) |
9462 |
|
9463 |
try:
|
9464 |
helper = masterd.instance.ExportInstanceHelper(self, feedback_fn,
|
9465 |
instance) |
9466 |
|
9467 |
helper.CreateSnapshots() |
9468 |
try:
|
9469 |
if (self.op.shutdown and instance.admin_up and |
9470 |
not self.op.remove_instance): |
9471 |
assert not activate_disks |
9472 |
feedback_fn("Starting instance %s" % instance.name)
|
9473 |
result = self.rpc.call_instance_start(src_node, instance, None, None) |
9474 |
msg = result.fail_msg |
9475 |
if msg:
|
9476 |
feedback_fn("Failed to start instance: %s" % msg)
|
9477 |
_ShutdownInstanceDisks(self, instance)
|
9478 |
raise errors.OpExecError("Could not start instance: %s" % msg) |
9479 |
|
9480 |
if self.op.mode == constants.EXPORT_MODE_LOCAL: |
9481 |
(fin_resu, dresults) = helper.LocalExport(self.dst_node)
|
9482 |
elif self.op.mode == constants.EXPORT_MODE_REMOTE: |
9483 |
connect_timeout = constants.RIE_CONNECT_TIMEOUT |
9484 |
timeouts = masterd.instance.ImportExportTimeouts(connect_timeout) |
9485 |
|
9486 |
(key_name, _, _) = self.x509_key_name
|
9487 |
|
9488 |
dest_ca_pem = \ |
9489 |
OpenSSL.crypto.dump_certificate(OpenSSL.crypto.FILETYPE_PEM, |
9490 |
self.dest_x509_ca)
|
9491 |
|
9492 |
(fin_resu, dresults) = helper.RemoteExport(self.dest_disk_info,
|
9493 |
key_name, dest_ca_pem, |
9494 |
timeouts) |
9495 |
finally:
|
9496 |
helper.Cleanup() |
9497 |
|
9498 |
# Check for backwards compatibility
|
9499 |
assert len(dresults) == len(instance.disks) |
9500 |
assert compat.all(isinstance(i, bool) for i in dresults), \ |
9501 |
"Not all results are boolean: %r" % dresults
|
9502 |
|
9503 |
finally:
|
9504 |
if activate_disks:
|
9505 |
feedback_fn("Deactivating disks for %s" % instance.name)
|
9506 |
_ShutdownInstanceDisks(self, instance)
|
9507 |
|
9508 |
# Remove instance if requested
|
9509 |
if self.op.remove_instance: |
9510 |
if not (compat.all(dresults) and fin_resu): |
9511 |
feedback_fn("Not removing instance %s as parts of the export failed" %
|
9512 |
instance.name) |
9513 |
else:
|
9514 |
feedback_fn("Removing instance %s" % instance.name)
|
9515 |
_RemoveInstance(self, feedback_fn, instance,
|
9516 |
self.op.ignore_remove_failures)
|
9517 |
|
9518 |
if self.op.mode == constants.EXPORT_MODE_LOCAL: |
9519 |
self._CleanupExports(feedback_fn)
|
9520 |
|
9521 |
return fin_resu, dresults
|
9522 |
|
9523 |
|
9524 |
class LURemoveExport(NoHooksLU): |
9525 |
"""Remove exports related to the named instance.
|
9526 |
|
9527 |
"""
|
9528 |
_OP_PARAMS = [ |
9529 |
_PInstanceName, |
9530 |
] |
9531 |
REQ_BGL = False
|
9532 |
|
9533 |
def ExpandNames(self): |
9534 |
self.needed_locks = {}
|
9535 |
# We need all nodes to be locked in order for RemoveExport to work, but we
|
9536 |
# don't need to lock the instance itself, as nothing will happen to it (and
|
9537 |
# we can remove exports also for a removed instance)
|
9538 |
self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
|
9539 |
|
9540 |
def Exec(self, feedback_fn): |
9541 |
"""Remove any export.
|
9542 |
|
9543 |
"""
|
9544 |
instance_name = self.cfg.ExpandInstanceName(self.op.instance_name) |
9545 |
# If the instance was not found we'll try with the name that was passed in.
|
9546 |
# This will only work if it was an FQDN, though.
|
9547 |
fqdn_warn = False
|
9548 |
if not instance_name: |
9549 |
fqdn_warn = True
|
9550 |
instance_name = self.op.instance_name
|
9551 |
|
9552 |
locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
|
9553 |
exportlist = self.rpc.call_export_list(locked_nodes)
|
9554 |
found = False
|
9555 |
for node in exportlist: |
9556 |
msg = exportlist[node].fail_msg |
9557 |
if msg:
|
9558 |
self.LogWarning("Failed to query node %s (continuing): %s", node, msg) |
9559 |
continue
|
9560 |
if instance_name in exportlist[node].payload: |
9561 |
found = True
|
9562 |
result = self.rpc.call_export_remove(node, instance_name)
|
9563 |
msg = result.fail_msg |
9564 |
if msg:
|
9565 |
logging.error("Could not remove export for instance %s"
|
9566 |
" on node %s: %s", instance_name, node, msg)
|
9567 |
|
9568 |
if fqdn_warn and not found: |
9569 |
feedback_fn("Export not found. If trying to remove an export belonging"
|
9570 |
" to a deleted instance please use its Fully Qualified"
|
9571 |
" Domain Name.")
|
9572 |
|
9573 |
|
9574 |
class TagsLU(NoHooksLU): # pylint: disable-msg=W0223 |
9575 |
"""Generic tags LU.
|
9576 |
|
9577 |
This is an abstract class which is the parent of all the other tags LUs.
|
9578 |
|
9579 |
"""
|
9580 |
|
9581 |
def ExpandNames(self): |
9582 |
self.needed_locks = {}
|
9583 |
if self.op.kind == constants.TAG_NODE: |
9584 |
self.op.name = _ExpandNodeName(self.cfg, self.op.name) |
9585 |
self.needed_locks[locking.LEVEL_NODE] = self.op.name |
9586 |
elif self.op.kind == constants.TAG_INSTANCE: |
9587 |
self.op.name = _ExpandInstanceName(self.cfg, self.op.name) |
9588 |
self.needed_locks[locking.LEVEL_INSTANCE] = self.op.name |
9589 |
|
9590 |
def CheckPrereq(self): |
9591 |
"""Check prerequisites.
|
9592 |
|
9593 |
"""
|
9594 |
if self.op.kind == constants.TAG_CLUSTER: |
9595 |
self.target = self.cfg.GetClusterInfo() |
9596 |
elif self.op.kind == constants.TAG_NODE: |
9597 |
self.target = self.cfg.GetNodeInfo(self.op.name) |
9598 |
elif self.op.kind == constants.TAG_INSTANCE: |
9599 |
self.target = self.cfg.GetInstanceInfo(self.op.name) |
9600 |
else:
|
9601 |
raise errors.OpPrereqError("Wrong tag type requested (%s)" % |
9602 |
str(self.op.kind), errors.ECODE_INVAL) |
9603 |
|
9604 |
|
9605 |
class LUGetTags(TagsLU): |
9606 |
"""Returns the tags of a given object.
|
9607 |
|
9608 |
"""
|
9609 |
_OP_PARAMS = [ |
9610 |
("kind", _NoDefault, _TElemOf(constants.VALID_TAG_TYPES)),
|
9611 |
("name", _NoDefault, _TNonEmptyString),
|
9612 |
] |
9613 |
REQ_BGL = False
|
9614 |
|
9615 |
def Exec(self, feedback_fn): |
9616 |
"""Returns the tag list.
|
9617 |
|
9618 |
"""
|
9619 |
return list(self.target.GetTags()) |
9620 |
|
9621 |
|
9622 |
class LUSearchTags(NoHooksLU): |
9623 |
"""Searches the tags for a given pattern.
|
9624 |
|
9625 |
"""
|
9626 |
_OP_PARAMS = [ |
9627 |
("pattern", _NoDefault, _TNonEmptyString),
|
9628 |
] |
9629 |
REQ_BGL = False
|
9630 |
|
9631 |
def ExpandNames(self): |
9632 |
self.needed_locks = {}
|
9633 |
|
9634 |
def CheckPrereq(self): |
9635 |
"""Check prerequisites.
|
9636 |
|
9637 |
This checks the pattern passed for validity by compiling it.
|
9638 |
|
9639 |
"""
|
9640 |
try:
|
9641 |
self.re = re.compile(self.op.pattern) |
9642 |
except re.error, err:
|
9643 |
raise errors.OpPrereqError("Invalid search pattern '%s': %s" % |
9644 |
(self.op.pattern, err), errors.ECODE_INVAL)
|
9645 |
|
9646 |
def Exec(self, feedback_fn): |
9647 |
"""Returns the tag list.
|
9648 |
|
9649 |
"""
|
9650 |
cfg = self.cfg
|
9651 |
tgts = [("/cluster", cfg.GetClusterInfo())]
|
9652 |
ilist = cfg.GetAllInstancesInfo().values() |
9653 |
tgts.extend([("/instances/%s" % i.name, i) for i in ilist]) |
9654 |
nlist = cfg.GetAllNodesInfo().values() |
9655 |
tgts.extend([("/nodes/%s" % n.name, n) for n in nlist]) |
9656 |
results = [] |
9657 |
for path, target in tgts: |
9658 |
for tag in target.GetTags(): |
9659 |
if self.re.search(tag): |
9660 |
results.append((path, tag)) |
9661 |
return results
|
9662 |
|
9663 |
|
9664 |
class LUAddTags(TagsLU): |
9665 |
"""Sets a tag on a given object.
|
9666 |
|
9667 |
"""
|
9668 |
_OP_PARAMS = [ |
9669 |
("kind", _NoDefault, _TElemOf(constants.VALID_TAG_TYPES)),
|
9670 |
("name", _NoDefault, _TNonEmptyString),
|
9671 |
("tags", _NoDefault, _TListOf(_TNonEmptyString)),
|
9672 |
] |
9673 |
REQ_BGL = False
|
9674 |
|
9675 |
def CheckPrereq(self): |
9676 |
"""Check prerequisites.
|
9677 |
|
9678 |
This checks the type and length of the tag name and value.
|
9679 |
|
9680 |
"""
|
9681 |
TagsLU.CheckPrereq(self)
|
9682 |
for tag in self.op.tags: |
9683 |
objects.TaggableObject.ValidateTag(tag) |
9684 |
|
9685 |
def Exec(self, feedback_fn): |
9686 |
"""Sets the tag.
|
9687 |
|
9688 |
"""
|
9689 |
try:
|
9690 |
for tag in self.op.tags: |
9691 |
self.target.AddTag(tag)
|
9692 |
except errors.TagError, err:
|
9693 |
raise errors.OpExecError("Error while setting tag: %s" % str(err)) |
9694 |
self.cfg.Update(self.target, feedback_fn) |
9695 |
|
9696 |
|
9697 |
class LUDelTags(TagsLU): |
9698 |
"""Delete a list of tags from a given object.
|
9699 |
|
9700 |
"""
|
9701 |
_OP_PARAMS = [ |
9702 |
("kind", _NoDefault, _TElemOf(constants.VALID_TAG_TYPES)),
|
9703 |
("name", _NoDefault, _TNonEmptyString),
|
9704 |
("tags", _NoDefault, _TListOf(_TNonEmptyString)),
|
9705 |
] |
9706 |
REQ_BGL = False
|
9707 |
|
9708 |
def CheckPrereq(self): |
9709 |
"""Check prerequisites.
|
9710 |
|
9711 |
This checks that we have the given tag.
|
9712 |
|
9713 |
"""
|
9714 |
TagsLU.CheckPrereq(self)
|
9715 |
for tag in self.op.tags: |
9716 |
objects.TaggableObject.ValidateTag(tag) |
9717 |
del_tags = frozenset(self.op.tags) |
9718 |
cur_tags = self.target.GetTags()
|
9719 |
if not del_tags <= cur_tags: |
9720 |
diff_tags = del_tags - cur_tags |
9721 |
diff_names = ["'%s'" % tag for tag in diff_tags] |
9722 |
diff_names.sort() |
9723 |
raise errors.OpPrereqError("Tag(s) %s not found" % |
9724 |
(",".join(diff_names)), errors.ECODE_NOENT)
|
9725 |
|
9726 |
def Exec(self, feedback_fn): |
9727 |
"""Remove the tag from the object.
|
9728 |
|
9729 |
"""
|
9730 |
for tag in self.op.tags: |
9731 |
self.target.RemoveTag(tag)
|
9732 |
self.cfg.Update(self.target, feedback_fn) |
9733 |
|
9734 |
|
9735 |
class LUTestDelay(NoHooksLU): |
9736 |
"""Sleep for a specified amount of time.
|
9737 |
|
9738 |
This LU sleeps on the master and/or nodes for a specified amount of
|
9739 |
time.
|
9740 |
|
9741 |
"""
|
9742 |
_OP_PARAMS = [ |
9743 |
("duration", _NoDefault, _TFloat),
|
9744 |
("on_master", True, _TBool), |
9745 |
("on_nodes", _EmptyList, _TListOf(_TNonEmptyString)),
|
9746 |
("repeat", 0, _TPositiveInt) |
9747 |
] |
9748 |
REQ_BGL = False
|
9749 |
|
9750 |
def ExpandNames(self): |
9751 |
"""Expand names and set required locks.
|
9752 |
|
9753 |
This expands the node list, if any.
|
9754 |
|
9755 |
"""
|
9756 |
self.needed_locks = {}
|
9757 |
if self.op.on_nodes: |
9758 |
# _GetWantedNodes can be used here, but is not always appropriate to use
|
9759 |
# this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
|
9760 |
# more information.
|
9761 |
self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes) |
9762 |
self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes |
9763 |
|
9764 |
def _TestDelay(self): |
9765 |
"""Do the actual sleep.
|
9766 |
|
9767 |
"""
|
9768 |
if self.op.on_master: |
9769 |
if not utils.TestDelay(self.op.duration): |
9770 |
raise errors.OpExecError("Error during master delay test") |
9771 |
if self.op.on_nodes: |
9772 |
result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration) |
9773 |
for node, node_result in result.items(): |
9774 |
node_result.Raise("Failure during rpc call to node %s" % node)
|
9775 |
|
9776 |
def Exec(self, feedback_fn): |
9777 |
"""Execute the test delay opcode, with the wanted repetitions.
|
9778 |
|
9779 |
"""
|
9780 |
if self.op.repeat == 0: |
9781 |
self._TestDelay()
|
9782 |
else:
|
9783 |
top_value = self.op.repeat - 1 |
9784 |
for i in range(self.op.repeat): |
9785 |
self.LogInfo("Test delay iteration %d/%d" % (i, top_value)) |
9786 |
self._TestDelay()
|
9787 |
|
9788 |
|
9789 |
class IAllocator(object): |
9790 |
"""IAllocator framework.
|
9791 |
|
9792 |
An IAllocator instance has three sets of attributes:
|
9793 |
- cfg that is needed to query the cluster
|
9794 |
- input data (all members of the _KEYS class attribute are required)
|
9795 |
- four buffer attributes (in|out_data|text), that represent the
|
9796 |
input (to the external script) in text and data structure format,
|
9797 |
and the output from it, again in two formats
|
9798 |
- the result variables from the script (success, info, nodes) for
|
9799 |
easy usage
|
9800 |
|
9801 |
"""
|
9802 |
# pylint: disable-msg=R0902
|
9803 |
# lots of instance attributes
|
9804 |
_ALLO_KEYS = [ |
9805 |
"name", "mem_size", "disks", "disk_template", |
9806 |
"os", "tags", "nics", "vcpus", "hypervisor", |
9807 |
] |
9808 |
_RELO_KEYS = [ |
9809 |
"name", "relocate_from", |
9810 |
] |
9811 |
_EVAC_KEYS = [ |
9812 |
"evac_nodes",
|
9813 |
] |
9814 |
|
9815 |
def __init__(self, cfg, rpc, mode, **kwargs): |
9816 |
self.cfg = cfg
|
9817 |
self.rpc = rpc
|
9818 |
# init buffer variables
|
9819 |
self.in_text = self.out_text = self.in_data = self.out_data = None |
9820 |
# init all input fields so that pylint is happy
|
9821 |
self.mode = mode
|
9822 |
self.mem_size = self.disks = self.disk_template = None |
9823 |
self.os = self.tags = self.nics = self.vcpus = None |
9824 |
self.hypervisor = None |
9825 |
self.relocate_from = None |
9826 |
self.name = None |
9827 |
self.evac_nodes = None |
9828 |
# computed fields
|
9829 |
self.required_nodes = None |
9830 |
# init result fields
|
9831 |
self.success = self.info = self.result = None |
9832 |
if self.mode == constants.IALLOCATOR_MODE_ALLOC: |
9833 |
keyset = self._ALLO_KEYS
|
9834 |
fn = self._AddNewInstance
|
9835 |
elif self.mode == constants.IALLOCATOR_MODE_RELOC: |
9836 |
keyset = self._RELO_KEYS
|
9837 |
fn = self._AddRelocateInstance
|
9838 |
elif self.mode == constants.IALLOCATOR_MODE_MEVAC: |
9839 |
keyset = self._EVAC_KEYS
|
9840 |
fn = self._AddEvacuateNodes
|
9841 |
else:
|
9842 |
raise errors.ProgrammerError("Unknown mode '%s' passed to the" |
9843 |
" IAllocator" % self.mode) |
9844 |
for key in kwargs: |
9845 |
if key not in keyset: |
9846 |
raise errors.ProgrammerError("Invalid input parameter '%s' to" |
9847 |
" IAllocator" % key)
|
9848 |
setattr(self, key, kwargs[key]) |
9849 |
|
9850 |
for key in keyset: |
9851 |
if key not in kwargs: |
9852 |
raise errors.ProgrammerError("Missing input parameter '%s' to" |
9853 |
" IAllocator" % key)
|
9854 |
self._BuildInputData(fn)
|
9855 |
|
9856 |
def _ComputeClusterData(self): |
9857 |
"""Compute the generic allocator input data.
|
9858 |
|
9859 |
This is the data that is independent of the actual operation.
|
9860 |
|
9861 |
"""
|
9862 |
cfg = self.cfg
|
9863 |
cluster_info = cfg.GetClusterInfo() |
9864 |
# cluster data
|
9865 |
data = { |
9866 |
"version": constants.IALLOCATOR_VERSION,
|
9867 |
"cluster_name": cfg.GetClusterName(),
|
9868 |
"cluster_tags": list(cluster_info.GetTags()), |
9869 |
"enabled_hypervisors": list(cluster_info.enabled_hypervisors), |
9870 |
# we don't have job IDs
|
9871 |
} |
9872 |
iinfo = cfg.GetAllInstancesInfo().values() |
9873 |
i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo] |
9874 |
|
9875 |
# node data
|
9876 |
node_results = {} |
9877 |
node_list = cfg.GetNodeList() |
9878 |
|
9879 |
if self.mode == constants.IALLOCATOR_MODE_ALLOC: |
9880 |
hypervisor_name = self.hypervisor
|
9881 |
elif self.mode == constants.IALLOCATOR_MODE_RELOC: |
9882 |
hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
|
9883 |
elif self.mode == constants.IALLOCATOR_MODE_MEVAC: |
9884 |
hypervisor_name = cluster_info.enabled_hypervisors[0]
|
9885 |
|
9886 |
node_data = self.rpc.call_node_info(node_list, cfg.GetVGName(),
|
9887 |
hypervisor_name) |
9888 |
node_iinfo = \ |
9889 |
self.rpc.call_all_instances_info(node_list,
|
9890 |
cluster_info.enabled_hypervisors) |
9891 |
for nname, nresult in node_data.items(): |
9892 |
# first fill in static (config-based) values
|
9893 |
ninfo = cfg.GetNodeInfo(nname) |
9894 |
pnr = { |
9895 |
"tags": list(ninfo.GetTags()), |
9896 |
"primary_ip": ninfo.primary_ip,
|
9897 |
"secondary_ip": ninfo.secondary_ip,
|
9898 |
"offline": ninfo.offline,
|
9899 |
"drained": ninfo.drained,
|
9900 |
"master_candidate": ninfo.master_candidate,
|
9901 |
} |
9902 |
|
9903 |
if not (ninfo.offline or ninfo.drained): |
9904 |
nresult.Raise("Can't get data for node %s" % nname)
|
9905 |
node_iinfo[nname].Raise("Can't get node instance info from node %s" %
|
9906 |
nname) |
9907 |
remote_info = nresult.payload |
9908 |
|
9909 |
for attr in ['memory_total', 'memory_free', 'memory_dom0', |
9910 |
'vg_size', 'vg_free', 'cpu_total']: |
9911 |
if attr not in remote_info: |
9912 |
raise errors.OpExecError("Node '%s' didn't return attribute" |
9913 |
" '%s'" % (nname, attr))
|
9914 |
if not isinstance(remote_info[attr], int): |
9915 |
raise errors.OpExecError("Node '%s' returned invalid value" |
9916 |
" for '%s': %s" %
|
9917 |
(nname, attr, remote_info[attr])) |
9918 |
# compute memory used by primary instances
|
9919 |
i_p_mem = i_p_up_mem = 0
|
9920 |
for iinfo, beinfo in i_list: |
9921 |
if iinfo.primary_node == nname:
|
9922 |
i_p_mem += beinfo[constants.BE_MEMORY] |
9923 |
if iinfo.name not in node_iinfo[nname].payload: |
9924 |
i_used_mem = 0
|
9925 |
else:
|
9926 |
i_used_mem = int(node_iinfo[nname].payload[iinfo.name]['memory']) |
9927 |
i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem |
9928 |
remote_info['memory_free'] -= max(0, i_mem_diff) |
9929 |
|
9930 |
if iinfo.admin_up:
|
9931 |
i_p_up_mem += beinfo[constants.BE_MEMORY] |
9932 |
|
9933 |
# compute memory used by instances
|
9934 |
pnr_dyn = { |
9935 |
"total_memory": remote_info['memory_total'], |
9936 |
"reserved_memory": remote_info['memory_dom0'], |
9937 |
"free_memory": remote_info['memory_free'], |
9938 |
"total_disk": remote_info['vg_size'], |
9939 |
"free_disk": remote_info['vg_free'], |
9940 |
"total_cpus": remote_info['cpu_total'], |
9941 |
"i_pri_memory": i_p_mem,
|
9942 |
"i_pri_up_memory": i_p_up_mem,
|
9943 |
} |
9944 |
pnr.update(pnr_dyn) |
9945 |
|
9946 |
node_results[nname] = pnr |
9947 |
data["nodes"] = node_results
|
9948 |
|
9949 |
# instance data
|
9950 |
instance_data = {} |
9951 |
for iinfo, beinfo in i_list: |
9952 |
nic_data = [] |
9953 |
for nic in iinfo.nics: |
9954 |
filled_params = cluster_info.SimpleFillNIC(nic.nicparams) |
9955 |
nic_dict = {"mac": nic.mac,
|
9956 |
"ip": nic.ip,
|
9957 |
"mode": filled_params[constants.NIC_MODE],
|
9958 |
"link": filled_params[constants.NIC_LINK],
|
9959 |
} |
9960 |
if filled_params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
|
9961 |
nic_dict["bridge"] = filled_params[constants.NIC_LINK]
|
9962 |
nic_data.append(nic_dict) |
9963 |
pir = { |
9964 |
"tags": list(iinfo.GetTags()), |
9965 |
"admin_up": iinfo.admin_up,
|
9966 |
"vcpus": beinfo[constants.BE_VCPUS],
|
9967 |
"memory": beinfo[constants.BE_MEMORY],
|
9968 |
"os": iinfo.os,
|
9969 |
"nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes), |
9970 |
"nics": nic_data,
|
9971 |
"disks": [{"size": dsk.size, "mode": dsk.mode} for dsk in iinfo.disks], |
9972 |
"disk_template": iinfo.disk_template,
|
9973 |
"hypervisor": iinfo.hypervisor,
|
9974 |
} |
9975 |
pir["disk_space_total"] = _ComputeDiskSize(iinfo.disk_template,
|
9976 |
pir["disks"])
|
9977 |
instance_data[iinfo.name] = pir |
9978 |
|
9979 |
data["instances"] = instance_data
|
9980 |
|
9981 |
self.in_data = data
|
9982 |
|
9983 |
def _AddNewInstance(self): |
9984 |
"""Add new instance data to allocator structure.
|
9985 |
|
9986 |
This in combination with _AllocatorGetClusterData will create the
|
9987 |
correct structure needed as input for the allocator.
|
9988 |
|
9989 |
The checks for the completeness of the opcode must have already been
|
9990 |
done.
|
9991 |
|
9992 |
"""
|
9993 |
disk_space = _ComputeDiskSize(self.disk_template, self.disks) |
9994 |
|
9995 |
if self.disk_template in constants.DTS_NET_MIRROR: |
9996 |
self.required_nodes = 2 |
9997 |
else:
|
9998 |
self.required_nodes = 1 |
9999 |
request = { |
10000 |
"name": self.name, |
10001 |
"disk_template": self.disk_template, |
10002 |
"tags": self.tags, |
10003 |
"os": self.os, |
10004 |
"vcpus": self.vcpus, |
10005 |
"memory": self.mem_size, |
10006 |
"disks": self.disks, |
10007 |
"disk_space_total": disk_space,
|
10008 |
"nics": self.nics, |
10009 |
"required_nodes": self.required_nodes, |
10010 |
} |
10011 |
return request
|
10012 |
|
10013 |
def _AddRelocateInstance(self): |
10014 |
"""Add relocate instance data to allocator structure.
|
10015 |
|
10016 |
This in combination with _IAllocatorGetClusterData will create the
|
10017 |
correct structure needed as input for the allocator.
|
10018 |
|
10019 |
The checks for the completeness of the opcode must have already been
|
10020 |
done.
|
10021 |
|
10022 |
"""
|
10023 |
instance = self.cfg.GetInstanceInfo(self.name) |
10024 |
if instance is None: |
10025 |
raise errors.ProgrammerError("Unknown instance '%s' passed to" |
10026 |
" IAllocator" % self.name) |
10027 |
|
10028 |
if instance.disk_template not in constants.DTS_NET_MIRROR: |
10029 |
raise errors.OpPrereqError("Can't relocate non-mirrored instances", |
10030 |
errors.ECODE_INVAL) |
10031 |
|
10032 |
if len(instance.secondary_nodes) != 1: |
10033 |
raise errors.OpPrereqError("Instance has not exactly one secondary node", |
10034 |
errors.ECODE_STATE) |
10035 |
|
10036 |
self.required_nodes = 1 |
10037 |
disk_sizes = [{'size': disk.size} for disk in instance.disks] |
10038 |
disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes) |
10039 |
|
10040 |
request = { |
10041 |
"name": self.name, |
10042 |
"disk_space_total": disk_space,
|
10043 |
"required_nodes": self.required_nodes, |
10044 |
"relocate_from": self.relocate_from, |
10045 |
} |
10046 |
return request
|
10047 |
|
10048 |
def _AddEvacuateNodes(self): |
10049 |
"""Add evacuate nodes data to allocator structure.
|
10050 |
|
10051 |
"""
|
10052 |
request = { |
10053 |
"evac_nodes": self.evac_nodes |
10054 |
} |
10055 |
return request
|
10056 |
|
10057 |
def _BuildInputData(self, fn): |
10058 |
"""Build input data structures.
|
10059 |
|
10060 |
"""
|
10061 |
self._ComputeClusterData()
|
10062 |
|
10063 |
request = fn() |
10064 |
request["type"] = self.mode |
10065 |
self.in_data["request"] = request |
10066 |
|
10067 |
self.in_text = serializer.Dump(self.in_data) |
10068 |
|
10069 |
def Run(self, name, validate=True, call_fn=None): |
10070 |
"""Run an instance allocator and return the results.
|
10071 |
|
10072 |
"""
|
10073 |
if call_fn is None: |
10074 |
call_fn = self.rpc.call_iallocator_runner
|
10075 |
|
10076 |
result = call_fn(self.cfg.GetMasterNode(), name, self.in_text) |
10077 |
result.Raise("Failure while running the iallocator script")
|
10078 |
|
10079 |
self.out_text = result.payload
|
10080 |
if validate:
|
10081 |
self._ValidateResult()
|
10082 |
|
10083 |
def _ValidateResult(self): |
10084 |
"""Process the allocator results.
|
10085 |
|
10086 |
This will process and if successful save the result in
|
10087 |
self.out_data and the other parameters.
|
10088 |
|
10089 |
"""
|
10090 |
try:
|
10091 |
rdict = serializer.Load(self.out_text)
|
10092 |
except Exception, err: |
10093 |
raise errors.OpExecError("Can't parse iallocator results: %s" % str(err)) |
10094 |
|
10095 |
if not isinstance(rdict, dict): |
10096 |
raise errors.OpExecError("Can't parse iallocator results: not a dict") |
10097 |
|
10098 |
# TODO: remove backwards compatiblity in later versions
|
10099 |
if "nodes" in rdict and "result" not in rdict: |
10100 |
rdict["result"] = rdict["nodes"] |
10101 |
del rdict["nodes"] |
10102 |
|
10103 |
for key in "success", "info", "result": |
10104 |
if key not in rdict: |
10105 |
raise errors.OpExecError("Can't parse iallocator results:" |
10106 |
" missing key '%s'" % key)
|
10107 |
setattr(self, key, rdict[key]) |
10108 |
|
10109 |
if not isinstance(rdict["result"], list): |
10110 |
raise errors.OpExecError("Can't parse iallocator results: 'result' key" |
10111 |
" is not a list")
|
10112 |
self.out_data = rdict
|
10113 |
|
10114 |
|
10115 |
class LUTestAllocator(NoHooksLU): |
10116 |
"""Run allocator tests.
|
10117 |
|
10118 |
This LU runs the allocator tests
|
10119 |
|
10120 |
"""
|
10121 |
_OP_PARAMS = [ |
10122 |
("direction", _NoDefault, _TElemOf(constants.VALID_IALLOCATOR_DIRECTIONS)),
|
10123 |
("mode", _NoDefault, _TElemOf(constants.VALID_IALLOCATOR_MODES)),
|
10124 |
("name", _NoDefault, _TNonEmptyString),
|
10125 |
("nics", _NoDefault, _TOr(_TNone, _TListOf(
|
10126 |
_TDictOf(_TElemOf(["mac", "ip", "bridge"]), |
10127 |
_TOr(_TNone, _TNonEmptyString))))), |
10128 |
("disks", _NoDefault, _TOr(_TNone, _TList)),
|
10129 |
("hypervisor", None, _TMaybeString), |
10130 |
("allocator", None, _TMaybeString), |
10131 |
("tags", _EmptyList, _TListOf(_TNonEmptyString)),
|
10132 |
("mem_size", None, _TOr(_TNone, _TPositiveInt)), |
10133 |
("vcpus", None, _TOr(_TNone, _TPositiveInt)), |
10134 |
("os", None, _TMaybeString), |
10135 |
("disk_template", None, _TMaybeString), |
10136 |
("evac_nodes", None, _TOr(_TNone, _TListOf(_TNonEmptyString))), |
10137 |
] |
10138 |
|
10139 |
def CheckPrereq(self): |
10140 |
"""Check prerequisites.
|
10141 |
|
10142 |
This checks the opcode parameters depending on the director and mode test.
|
10143 |
|
10144 |
"""
|
10145 |
if self.op.mode == constants.IALLOCATOR_MODE_ALLOC: |
10146 |
for attr in ["mem_size", "disks", "disk_template", |
10147 |
"os", "tags", "nics", "vcpus"]: |
10148 |
if not hasattr(self.op, attr): |
10149 |
raise errors.OpPrereqError("Missing attribute '%s' on opcode input" % |
10150 |
attr, errors.ECODE_INVAL) |
10151 |
iname = self.cfg.ExpandInstanceName(self.op.name) |
10152 |
if iname is not None: |
10153 |
raise errors.OpPrereqError("Instance '%s' already in the cluster" % |
10154 |
iname, errors.ECODE_EXISTS) |
10155 |
if not isinstance(self.op.nics, list): |
10156 |
raise errors.OpPrereqError("Invalid parameter 'nics'", |
10157 |
errors.ECODE_INVAL) |
10158 |
if not isinstance(self.op.disks, list): |
10159 |
raise errors.OpPrereqError("Invalid parameter 'disks'", |
10160 |
errors.ECODE_INVAL) |
10161 |
for row in self.op.disks: |
10162 |
if (not isinstance(row, dict) or |
10163 |
"size" not in row or |
10164 |
not isinstance(row["size"], int) or |
10165 |
"mode" not in row or |
10166 |
row["mode"] not in ['r', 'w']): |
10167 |
raise errors.OpPrereqError("Invalid contents of the 'disks'" |
10168 |
" parameter", errors.ECODE_INVAL)
|
10169 |
if self.op.hypervisor is None: |
10170 |
self.op.hypervisor = self.cfg.GetHypervisorType() |
10171 |
elif self.op.mode == constants.IALLOCATOR_MODE_RELOC: |
10172 |
fname = _ExpandInstanceName(self.cfg, self.op.name) |
10173 |
self.op.name = fname
|
10174 |
self.relocate_from = self.cfg.GetInstanceInfo(fname).secondary_nodes |
10175 |
elif self.op.mode == constants.IALLOCATOR_MODE_MEVAC: |
10176 |
if not hasattr(self.op, "evac_nodes"): |
10177 |
raise errors.OpPrereqError("Missing attribute 'evac_nodes' on" |
10178 |
" opcode input", errors.ECODE_INVAL)
|
10179 |
else:
|
10180 |
raise errors.OpPrereqError("Invalid test allocator mode '%s'" % |
10181 |
self.op.mode, errors.ECODE_INVAL)
|
10182 |
|
10183 |
if self.op.direction == constants.IALLOCATOR_DIR_OUT: |
10184 |
if self.op.allocator is None: |
10185 |
raise errors.OpPrereqError("Missing allocator name", |
10186 |
errors.ECODE_INVAL) |
10187 |
elif self.op.direction != constants.IALLOCATOR_DIR_IN: |
10188 |
raise errors.OpPrereqError("Wrong allocator test '%s'" % |
10189 |
self.op.direction, errors.ECODE_INVAL)
|
10190 |
|
10191 |
def Exec(self, feedback_fn): |
10192 |
"""Run the allocator test.
|
10193 |
|
10194 |
"""
|
10195 |
if self.op.mode == constants.IALLOCATOR_MODE_ALLOC: |
10196 |
ial = IAllocator(self.cfg, self.rpc, |
10197 |
mode=self.op.mode,
|
10198 |
name=self.op.name,
|
10199 |
mem_size=self.op.mem_size,
|
10200 |
disks=self.op.disks,
|
10201 |
disk_template=self.op.disk_template,
|
10202 |
os=self.op.os,
|
10203 |
tags=self.op.tags,
|
10204 |
nics=self.op.nics,
|
10205 |
vcpus=self.op.vcpus,
|
10206 |
hypervisor=self.op.hypervisor,
|
10207 |
) |
10208 |
elif self.op.mode == constants.IALLOCATOR_MODE_RELOC: |
10209 |
ial = IAllocator(self.cfg, self.rpc, |
10210 |
mode=self.op.mode,
|
10211 |
name=self.op.name,
|
10212 |
relocate_from=list(self.relocate_from), |
10213 |
) |
10214 |
elif self.op.mode == constants.IALLOCATOR_MODE_MEVAC: |
10215 |
ial = IAllocator(self.cfg, self.rpc, |
10216 |
mode=self.op.mode,
|
10217 |
evac_nodes=self.op.evac_nodes)
|
10218 |
else:
|
10219 |
raise errors.ProgrammerError("Uncatched mode %s in" |
10220 |
" LUTestAllocator.Exec", self.op.mode) |
10221 |
|
10222 |
if self.op.direction == constants.IALLOCATOR_DIR_IN: |
10223 |
result = ial.in_text |
10224 |
else:
|
10225 |
ial.Run(self.op.allocator, validate=False) |
10226 |
result = ial.out_text |
10227 |
return result
|