root / lib / backend.py @ 464f8daf
History | View | Annotate | Download (80.6 kB)
1 |
#
|
---|---|
2 |
#
|
3 |
|
4 |
# Copyright (C) 2006, 2007 Google Inc.
|
5 |
#
|
6 |
# This program is free software; you can redistribute it and/or modify
|
7 |
# it under the terms of the GNU General Public License as published by
|
8 |
# the Free Software Foundation; either version 2 of the License, or
|
9 |
# (at your option) any later version.
|
10 |
#
|
11 |
# This program is distributed in the hope that it will be useful, but
|
12 |
# WITHOUT ANY WARRANTY; without even the implied warranty of
|
13 |
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
14 |
# General Public License for more details.
|
15 |
#
|
16 |
# You should have received a copy of the GNU General Public License
|
17 |
# along with this program; if not, write to the Free Software
|
18 |
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
|
19 |
# 02110-1301, USA.
|
20 |
|
21 |
|
22 |
"""Functions used by the node daemon"""
|
23 |
|
24 |
|
25 |
import os |
26 |
import os.path |
27 |
import shutil |
28 |
import time |
29 |
import stat |
30 |
import errno |
31 |
import re |
32 |
import subprocess |
33 |
import random |
34 |
import logging |
35 |
import tempfile |
36 |
import zlib |
37 |
import base64 |
38 |
|
39 |
from ganeti import errors |
40 |
from ganeti import utils |
41 |
from ganeti import ssh |
42 |
from ganeti import hypervisor |
43 |
from ganeti import constants |
44 |
from ganeti import bdev |
45 |
from ganeti import objects |
46 |
from ganeti import ssconf |
47 |
|
48 |
|
49 |
def _GetConfig(): |
50 |
"""Simple wrapper to return a SimpleStore.
|
51 |
|
52 |
@rtype: L{ssconf.SimpleStore}
|
53 |
@return: a SimpleStore instance
|
54 |
|
55 |
"""
|
56 |
return ssconf.SimpleStore()
|
57 |
|
58 |
|
59 |
def _GetSshRunner(cluster_name): |
60 |
"""Simple wrapper to return an SshRunner.
|
61 |
|
62 |
@type cluster_name: str
|
63 |
@param cluster_name: the cluster name, which is needed
|
64 |
by the SshRunner constructor
|
65 |
@rtype: L{ssh.SshRunner}
|
66 |
@return: an SshRunner instance
|
67 |
|
68 |
"""
|
69 |
return ssh.SshRunner(cluster_name)
|
70 |
|
71 |
|
72 |
def _Decompress(data): |
73 |
"""Unpacks data compressed by the RPC client.
|
74 |
|
75 |
@type data: list or tuple
|
76 |
@param data: Data sent by RPC client
|
77 |
@rtype: str
|
78 |
@return: Decompressed data
|
79 |
|
80 |
"""
|
81 |
assert isinstance(data, (list, tuple)) |
82 |
assert len(data) == 2 |
83 |
(encoding, content) = data |
84 |
if encoding == constants.RPC_ENCODING_NONE:
|
85 |
return content
|
86 |
elif encoding == constants.RPC_ENCODING_ZLIB_BASE64:
|
87 |
return zlib.decompress(base64.b64decode(content))
|
88 |
else:
|
89 |
raise AssertionError("Unknown data encoding") |
90 |
|
91 |
|
92 |
def _CleanDirectory(path, exclude=None): |
93 |
"""Removes all regular files in a directory.
|
94 |
|
95 |
@type path: str
|
96 |
@param path: the directory to clean
|
97 |
@type exclude: list
|
98 |
@param exclude: list of files to be excluded, defaults
|
99 |
to the empty list
|
100 |
|
101 |
"""
|
102 |
if not os.path.isdir(path): |
103 |
return
|
104 |
if exclude is None: |
105 |
exclude = [] |
106 |
else:
|
107 |
# Normalize excluded paths
|
108 |
exclude = [os.path.normpath(i) for i in exclude] |
109 |
|
110 |
for rel_name in utils.ListVisibleFiles(path): |
111 |
full_name = os.path.normpath(os.path.join(path, rel_name)) |
112 |
if full_name in exclude: |
113 |
continue
|
114 |
if os.path.isfile(full_name) and not os.path.islink(full_name): |
115 |
utils.RemoveFile(full_name) |
116 |
|
117 |
|
118 |
def JobQueuePurge(): |
119 |
"""Removes job queue files and archived jobs.
|
120 |
|
121 |
@rtype: None
|
122 |
|
123 |
"""
|
124 |
_CleanDirectory(constants.QUEUE_DIR, exclude=[constants.JOB_QUEUE_LOCK_FILE]) |
125 |
_CleanDirectory(constants.JOB_QUEUE_ARCHIVE_DIR) |
126 |
|
127 |
|
128 |
def GetMasterInfo(): |
129 |
"""Returns master information.
|
130 |
|
131 |
This is an utility function to compute master information, either
|
132 |
for consumption here or from the node daemon.
|
133 |
|
134 |
@rtype: tuple
|
135 |
@return: (master_netdev, master_ip, master_name) if we have a good
|
136 |
configuration, otherwise (None, None, None)
|
137 |
|
138 |
"""
|
139 |
try:
|
140 |
cfg = _GetConfig() |
141 |
master_netdev = cfg.GetMasterNetdev() |
142 |
master_ip = cfg.GetMasterIP() |
143 |
master_node = cfg.GetMasterNode() |
144 |
except errors.ConfigurationError, err:
|
145 |
logging.exception("Cluster configuration incomplete")
|
146 |
return (None, None, None) |
147 |
return (master_netdev, master_ip, master_node)
|
148 |
|
149 |
|
150 |
def StartMaster(start_daemons): |
151 |
"""Activate local node as master node.
|
152 |
|
153 |
The function will always try activate the IP address of the master
|
154 |
(unless someone else has it). It will also start the master daemons,
|
155 |
based on the start_daemons parameter.
|
156 |
|
157 |
@type start_daemons: boolean
|
158 |
@param start_daemons: whther to also start the master
|
159 |
daemons (ganeti-masterd and ganeti-rapi)
|
160 |
@rtype: None
|
161 |
|
162 |
"""
|
163 |
ok = True
|
164 |
master_netdev, master_ip, _ = GetMasterInfo() |
165 |
if not master_netdev: |
166 |
return False |
167 |
|
168 |
if utils.TcpPing(master_ip, constants.DEFAULT_NODED_PORT):
|
169 |
if utils.OwnIpAddress(master_ip):
|
170 |
# we already have the ip:
|
171 |
logging.debug("Already started")
|
172 |
else:
|
173 |
logging.error("Someone else has the master ip, not activating")
|
174 |
ok = False
|
175 |
else:
|
176 |
result = utils.RunCmd(["ip", "address", "add", "%s/32" % master_ip, |
177 |
"dev", master_netdev, "label", |
178 |
"%s:0" % master_netdev])
|
179 |
if result.failed:
|
180 |
logging.error("Can't activate master IP: %s", result.output)
|
181 |
ok = False
|
182 |
|
183 |
result = utils.RunCmd(["arping", "-q", "-U", "-c 3", "-I", master_netdev, |
184 |
"-s", master_ip, master_ip])
|
185 |
# we'll ignore the exit code of arping
|
186 |
|
187 |
# and now start the master and rapi daemons
|
188 |
if start_daemons:
|
189 |
for daemon in 'ganeti-masterd', 'ganeti-rapi': |
190 |
result = utils.RunCmd([daemon]) |
191 |
if result.failed:
|
192 |
logging.error("Can't start daemon %s: %s", daemon, result.output)
|
193 |
ok = False
|
194 |
return ok
|
195 |
|
196 |
|
197 |
def StopMaster(stop_daemons): |
198 |
"""Deactivate this node as master.
|
199 |
|
200 |
The function will always try to deactivate the IP address of the
|
201 |
master. It will also stop the master daemons depending on the
|
202 |
stop_daemons parameter.
|
203 |
|
204 |
@type stop_daemons: boolean
|
205 |
@param stop_daemons: whether to also stop the master daemons
|
206 |
(ganeti-masterd and ganeti-rapi)
|
207 |
@rtype: None
|
208 |
|
209 |
"""
|
210 |
master_netdev, master_ip, _ = GetMasterInfo() |
211 |
if not master_netdev: |
212 |
return False |
213 |
|
214 |
result = utils.RunCmd(["ip", "address", "del", "%s/32" % master_ip, |
215 |
"dev", master_netdev])
|
216 |
if result.failed:
|
217 |
logging.error("Can't remove the master IP, error: %s", result.output)
|
218 |
# but otherwise ignore the failure
|
219 |
|
220 |
if stop_daemons:
|
221 |
# stop/kill the rapi and the master daemon
|
222 |
for daemon in constants.RAPI_PID, constants.MASTERD_PID: |
223 |
utils.KillProcess(utils.ReadPidFile(utils.DaemonPidFileName(daemon))) |
224 |
|
225 |
return True |
226 |
|
227 |
|
228 |
def AddNode(dsa, dsapub, rsa, rsapub, sshkey, sshpub): |
229 |
"""Joins this node to the cluster.
|
230 |
|
231 |
This does the following:
|
232 |
- updates the hostkeys of the machine (rsa and dsa)
|
233 |
- adds the ssh private key to the user
|
234 |
- adds the ssh public key to the users' authorized_keys file
|
235 |
|
236 |
@type dsa: str
|
237 |
@param dsa: the DSA private key to write
|
238 |
@type dsapub: str
|
239 |
@param dsapub: the DSA public key to write
|
240 |
@type rsa: str
|
241 |
@param rsa: the RSA private key to write
|
242 |
@type rsapub: str
|
243 |
@param rsapub: the RSA public key to write
|
244 |
@type sshkey: str
|
245 |
@param sshkey: the SSH private key to write
|
246 |
@type sshpub: str
|
247 |
@param sshpub: the SSH public key to write
|
248 |
@rtype: boolean
|
249 |
@return: the success of the operation
|
250 |
|
251 |
"""
|
252 |
sshd_keys = [(constants.SSH_HOST_RSA_PRIV, rsa, 0600),
|
253 |
(constants.SSH_HOST_RSA_PUB, rsapub, 0644),
|
254 |
(constants.SSH_HOST_DSA_PRIV, dsa, 0600),
|
255 |
(constants.SSH_HOST_DSA_PUB, dsapub, 0644)]
|
256 |
for name, content, mode in sshd_keys: |
257 |
utils.WriteFile(name, data=content, mode=mode) |
258 |
|
259 |
try:
|
260 |
priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS, |
261 |
mkdir=True)
|
262 |
except errors.OpExecError, err:
|
263 |
msg = "Error while processing user ssh files"
|
264 |
logging.exception(msg) |
265 |
return (False, "%s: %s" % (msg, err)) |
266 |
|
267 |
for name, content in [(priv_key, sshkey), (pub_key, sshpub)]: |
268 |
utils.WriteFile(name, data=content, mode=0600)
|
269 |
|
270 |
utils.AddAuthorizedKey(auth_keys, sshpub) |
271 |
|
272 |
utils.RunCmd([constants.SSH_INITD_SCRIPT, "restart"])
|
273 |
|
274 |
return (True, "Node added successfully") |
275 |
|
276 |
|
277 |
def LeaveCluster(): |
278 |
"""Cleans up and remove the current node.
|
279 |
|
280 |
This function cleans up and prepares the current node to be removed
|
281 |
from the cluster.
|
282 |
|
283 |
If processing is successful, then it raises an
|
284 |
L{errors.QuitGanetiException} which is used as a special case to
|
285 |
shutdown the node daemon.
|
286 |
|
287 |
"""
|
288 |
_CleanDirectory(constants.DATA_DIR) |
289 |
JobQueuePurge() |
290 |
|
291 |
try:
|
292 |
priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS) |
293 |
except errors.OpExecError:
|
294 |
logging.exception("Error while processing ssh files")
|
295 |
return
|
296 |
|
297 |
f = open(pub_key, 'r') |
298 |
try:
|
299 |
utils.RemoveAuthorizedKey(auth_keys, f.read(8192))
|
300 |
finally:
|
301 |
f.close() |
302 |
|
303 |
utils.RemoveFile(priv_key) |
304 |
utils.RemoveFile(pub_key) |
305 |
|
306 |
# Return a reassuring string to the caller, and quit
|
307 |
raise errors.QuitGanetiException(False, 'Shutdown scheduled') |
308 |
|
309 |
|
310 |
def GetNodeInfo(vgname, hypervisor_type): |
311 |
"""Gives back a hash with different informations about the node.
|
312 |
|
313 |
@type vgname: C{string}
|
314 |
@param vgname: the name of the volume group to ask for disk space information
|
315 |
@type hypervisor_type: C{str}
|
316 |
@param hypervisor_type: the name of the hypervisor to ask for
|
317 |
memory information
|
318 |
@rtype: C{dict}
|
319 |
@return: dictionary with the following keys:
|
320 |
- vg_size is the size of the configured volume group in MiB
|
321 |
- vg_free is the free size of the volume group in MiB
|
322 |
- memory_dom0 is the memory allocated for domain0 in MiB
|
323 |
- memory_free is the currently available (free) ram in MiB
|
324 |
- memory_total is the total number of ram in MiB
|
325 |
|
326 |
"""
|
327 |
outputarray = {} |
328 |
vginfo = _GetVGInfo(vgname) |
329 |
outputarray['vg_size'] = vginfo['vg_size'] |
330 |
outputarray['vg_free'] = vginfo['vg_free'] |
331 |
|
332 |
hyper = hypervisor.GetHypervisor(hypervisor_type) |
333 |
hyp_info = hyper.GetNodeInfo() |
334 |
if hyp_info is not None: |
335 |
outputarray.update(hyp_info) |
336 |
|
337 |
f = open("/proc/sys/kernel/random/boot_id", 'r') |
338 |
try:
|
339 |
outputarray["bootid"] = f.read(128).rstrip("\n") |
340 |
finally:
|
341 |
f.close() |
342 |
|
343 |
return outputarray
|
344 |
|
345 |
|
346 |
def VerifyNode(what, cluster_name): |
347 |
"""Verify the status of the local node.
|
348 |
|
349 |
Based on the input L{what} parameter, various checks are done on the
|
350 |
local node.
|
351 |
|
352 |
If the I{filelist} key is present, this list of
|
353 |
files is checksummed and the file/checksum pairs are returned.
|
354 |
|
355 |
If the I{nodelist} key is present, we check that we have
|
356 |
connectivity via ssh with the target nodes (and check the hostname
|
357 |
report).
|
358 |
|
359 |
If the I{node-net-test} key is present, we check that we have
|
360 |
connectivity to the given nodes via both primary IP and, if
|
361 |
applicable, secondary IPs.
|
362 |
|
363 |
@type what: C{dict}
|
364 |
@param what: a dictionary of things to check:
|
365 |
- filelist: list of files for which to compute checksums
|
366 |
- nodelist: list of nodes we should check ssh communication with
|
367 |
- node-net-test: list of nodes we should check node daemon port
|
368 |
connectivity with
|
369 |
- hypervisor: list with hypervisors to run the verify for
|
370 |
@rtype: dict
|
371 |
@return: a dictionary with the same keys as the input dict, and
|
372 |
values representing the result of the checks
|
373 |
|
374 |
"""
|
375 |
result = {} |
376 |
|
377 |
if constants.NV_HYPERVISOR in what: |
378 |
result[constants.NV_HYPERVISOR] = tmp = {} |
379 |
for hv_name in what[constants.NV_HYPERVISOR]: |
380 |
tmp[hv_name] = hypervisor.GetHypervisor(hv_name).Verify() |
381 |
|
382 |
if constants.NV_FILELIST in what: |
383 |
result[constants.NV_FILELIST] = utils.FingerprintFiles( |
384 |
what[constants.NV_FILELIST]) |
385 |
|
386 |
if constants.NV_NODELIST in what: |
387 |
result[constants.NV_NODELIST] = tmp = {} |
388 |
random.shuffle(what[constants.NV_NODELIST]) |
389 |
for node in what[constants.NV_NODELIST]: |
390 |
success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node) |
391 |
if not success: |
392 |
tmp[node] = message |
393 |
|
394 |
if constants.NV_NODENETTEST in what: |
395 |
result[constants.NV_NODENETTEST] = tmp = {} |
396 |
my_name = utils.HostInfo().name |
397 |
my_pip = my_sip = None
|
398 |
for name, pip, sip in what[constants.NV_NODENETTEST]: |
399 |
if name == my_name:
|
400 |
my_pip = pip |
401 |
my_sip = sip |
402 |
break
|
403 |
if not my_pip: |
404 |
tmp[my_name] = ("Can't find my own primary/secondary IP"
|
405 |
" in the node list")
|
406 |
else:
|
407 |
port = utils.GetNodeDaemonPort() |
408 |
for name, pip, sip in what[constants.NV_NODENETTEST]: |
409 |
fail = [] |
410 |
if not utils.TcpPing(pip, port, source=my_pip): |
411 |
fail.append("primary")
|
412 |
if sip != pip:
|
413 |
if not utils.TcpPing(sip, port, source=my_sip): |
414 |
fail.append("secondary")
|
415 |
if fail:
|
416 |
tmp[name] = ("failure using the %s interface(s)" %
|
417 |
" and ".join(fail))
|
418 |
|
419 |
if constants.NV_LVLIST in what: |
420 |
result[constants.NV_LVLIST] = GetVolumeList(what[constants.NV_LVLIST]) |
421 |
|
422 |
if constants.NV_INSTANCELIST in what: |
423 |
result[constants.NV_INSTANCELIST] = GetInstanceList( |
424 |
what[constants.NV_INSTANCELIST]) |
425 |
|
426 |
if constants.NV_VGLIST in what: |
427 |
result[constants.NV_VGLIST] = ListVolumeGroups() |
428 |
|
429 |
if constants.NV_VERSION in what: |
430 |
result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION, |
431 |
constants.RELEASE_VERSION) |
432 |
|
433 |
if constants.NV_HVINFO in what: |
434 |
hyper = hypervisor.GetHypervisor(what[constants.NV_HVINFO]) |
435 |
result[constants.NV_HVINFO] = hyper.GetNodeInfo() |
436 |
|
437 |
if constants.NV_DRBDLIST in what: |
438 |
try:
|
439 |
used_minors = bdev.DRBD8.GetUsedDevs().keys() |
440 |
except errors.BlockDeviceError, err:
|
441 |
logging.warning("Can't get used minors list", exc_info=True) |
442 |
used_minors = str(err)
|
443 |
result[constants.NV_DRBDLIST] = used_minors |
444 |
|
445 |
return result
|
446 |
|
447 |
|
448 |
def GetVolumeList(vg_name): |
449 |
"""Compute list of logical volumes and their size.
|
450 |
|
451 |
@type vg_name: str
|
452 |
@param vg_name: the volume group whose LVs we should list
|
453 |
@rtype: dict
|
454 |
@return:
|
455 |
dictionary of all partions (key) with value being a tuple of
|
456 |
their size (in MiB), inactive and online status::
|
457 |
|
458 |
{'test1': ('20.06', True, True)}
|
459 |
|
460 |
in case of errors, a string is returned with the error
|
461 |
details.
|
462 |
|
463 |
"""
|
464 |
lvs = {} |
465 |
sep = '|'
|
466 |
result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix", |
467 |
"--separator=%s" % sep,
|
468 |
"-olv_name,lv_size,lv_attr", vg_name])
|
469 |
if result.failed:
|
470 |
logging.error("Failed to list logical volumes, lvs output: %s",
|
471 |
result.output) |
472 |
return result.output
|
473 |
|
474 |
valid_line_re = re.compile("^ *([^|]+)\|([0-9.]+)\|([^|]{6})\|?$")
|
475 |
for line in result.stdout.splitlines(): |
476 |
line = line.strip() |
477 |
match = valid_line_re.match(line) |
478 |
if not match: |
479 |
logging.error("Invalid line returned from lvs output: '%s'", line)
|
480 |
continue
|
481 |
name, size, attr = match.groups() |
482 |
inactive = attr[4] == '-' |
483 |
online = attr[5] == 'o' |
484 |
lvs[name] = (size, inactive, online) |
485 |
|
486 |
return lvs
|
487 |
|
488 |
|
489 |
def ListVolumeGroups(): |
490 |
"""List the volume groups and their size.
|
491 |
|
492 |
@rtype: dict
|
493 |
@return: dictionary with keys volume name and values the
|
494 |
size of the volume
|
495 |
|
496 |
"""
|
497 |
return utils.ListVolumeGroups()
|
498 |
|
499 |
|
500 |
def NodeVolumes(): |
501 |
"""List all volumes on this node.
|
502 |
|
503 |
@rtype: list
|
504 |
@return:
|
505 |
A list of dictionaries, each having four keys:
|
506 |
- name: the logical volume name,
|
507 |
- size: the size of the logical volume
|
508 |
- dev: the physical device on which the LV lives
|
509 |
- vg: the volume group to which it belongs
|
510 |
|
511 |
In case of errors, we return an empty list and log the
|
512 |
error.
|
513 |
|
514 |
Note that since a logical volume can live on multiple physical
|
515 |
volumes, the resulting list might include a logical volume
|
516 |
multiple times.
|
517 |
|
518 |
"""
|
519 |
result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix", |
520 |
"--separator=|",
|
521 |
"--options=lv_name,lv_size,devices,vg_name"])
|
522 |
if result.failed:
|
523 |
logging.error("Failed to list logical volumes, lvs output: %s",
|
524 |
result.output) |
525 |
return []
|
526 |
|
527 |
def parse_dev(dev): |
528 |
if '(' in dev: |
529 |
return dev.split('(')[0] |
530 |
else:
|
531 |
return dev
|
532 |
|
533 |
def map_line(line): |
534 |
return {
|
535 |
'name': line[0].strip(), |
536 |
'size': line[1].strip(), |
537 |
'dev': parse_dev(line[2].strip()), |
538 |
'vg': line[3].strip(), |
539 |
} |
540 |
|
541 |
return [map_line(line.split('|')) for line in result.stdout.splitlines() |
542 |
if line.count('|') >= 3] |
543 |
|
544 |
|
545 |
def BridgesExist(bridges_list): |
546 |
"""Check if a list of bridges exist on the current node.
|
547 |
|
548 |
@rtype: boolean
|
549 |
@return: C{True} if all of them exist, C{False} otherwise
|
550 |
|
551 |
"""
|
552 |
for bridge in bridges_list: |
553 |
if not utils.BridgeExists(bridge): |
554 |
return False |
555 |
|
556 |
return True |
557 |
|
558 |
|
559 |
def GetInstanceList(hypervisor_list): |
560 |
"""Provides a list of instances.
|
561 |
|
562 |
@type hypervisor_list: list
|
563 |
@param hypervisor_list: the list of hypervisors to query information
|
564 |
|
565 |
@rtype: list
|
566 |
@return: a list of all running instances on the current node
|
567 |
- instance1.example.com
|
568 |
- instance2.example.com
|
569 |
|
570 |
"""
|
571 |
results = [] |
572 |
for hname in hypervisor_list: |
573 |
try:
|
574 |
names = hypervisor.GetHypervisor(hname).ListInstances() |
575 |
results.extend(names) |
576 |
except errors.HypervisorError, err:
|
577 |
logging.exception("Error enumerating instances for hypevisor %s", hname)
|
578 |
raise
|
579 |
|
580 |
return results
|
581 |
|
582 |
|
583 |
def GetInstanceInfo(instance, hname): |
584 |
"""Gives back the informations about an instance as a dictionary.
|
585 |
|
586 |
@type instance: string
|
587 |
@param instance: the instance name
|
588 |
@type hname: string
|
589 |
@param hname: the hypervisor type of the instance
|
590 |
|
591 |
@rtype: dict
|
592 |
@return: dictionary with the following keys:
|
593 |
- memory: memory size of instance (int)
|
594 |
- state: xen state of instance (string)
|
595 |
- time: cpu time of instance (float)
|
596 |
|
597 |
"""
|
598 |
output = {} |
599 |
|
600 |
iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance) |
601 |
if iinfo is not None: |
602 |
output['memory'] = iinfo[2] |
603 |
output['state'] = iinfo[4] |
604 |
output['time'] = iinfo[5] |
605 |
|
606 |
return output
|
607 |
|
608 |
|
609 |
def GetInstanceMigratable(instance): |
610 |
"""Gives whether an instance can be migrated.
|
611 |
|
612 |
@type instance: L{objects.Instance}
|
613 |
@param instance: object representing the instance to be checked.
|
614 |
|
615 |
@rtype: tuple
|
616 |
@return: tuple of (result, description) where:
|
617 |
- result: whether the instance can be migrated or not
|
618 |
- description: a description of the issue, if relevant
|
619 |
|
620 |
"""
|
621 |
hyper = hypervisor.GetHypervisor(instance.hypervisor) |
622 |
if instance.name not in hyper.ListInstances(): |
623 |
return (False, 'not running') |
624 |
|
625 |
for idx in range(len(instance.disks)): |
626 |
link_name = _GetBlockDevSymlinkPath(instance.name, idx) |
627 |
if not os.path.islink(link_name): |
628 |
return (False, 'not restarted since ganeti 1.2.5') |
629 |
|
630 |
return (True, '') |
631 |
|
632 |
|
633 |
def GetAllInstancesInfo(hypervisor_list): |
634 |
"""Gather data about all instances.
|
635 |
|
636 |
This is the equivalent of L{GetInstanceInfo}, except that it
|
637 |
computes data for all instances at once, thus being faster if one
|
638 |
needs data about more than one instance.
|
639 |
|
640 |
@type hypervisor_list: list
|
641 |
@param hypervisor_list: list of hypervisors to query for instance data
|
642 |
|
643 |
@rtype: dict
|
644 |
@return: dictionary of instance: data, with data having the following keys:
|
645 |
- memory: memory size of instance (int)
|
646 |
- state: xen state of instance (string)
|
647 |
- time: cpu time of instance (float)
|
648 |
- vcpus: the number of vcpus
|
649 |
|
650 |
"""
|
651 |
output = {} |
652 |
|
653 |
for hname in hypervisor_list: |
654 |
iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo() |
655 |
if iinfo:
|
656 |
for name, inst_id, memory, vcpus, state, times in iinfo: |
657 |
value = { |
658 |
'memory': memory,
|
659 |
'vcpus': vcpus,
|
660 |
'state': state,
|
661 |
'time': times,
|
662 |
} |
663 |
if name in output: |
664 |
# we only check static parameters, like memory and vcpus,
|
665 |
# and not state and time which can change between the
|
666 |
# invocations of the different hypervisors
|
667 |
for key in 'memory', 'vcpus': |
668 |
if value[key] != output[name][key]:
|
669 |
raise errors.HypervisorError("Instance %s is running twice" |
670 |
" with different parameters" % name)
|
671 |
output[name] = value |
672 |
|
673 |
return output
|
674 |
|
675 |
|
676 |
def InstanceOsAdd(instance): |
677 |
"""Add an OS to an instance.
|
678 |
|
679 |
@type instance: L{objects.Instance}
|
680 |
@param instance: Instance whose OS is to be installed
|
681 |
@rtype: boolean
|
682 |
@return: the success of the operation
|
683 |
|
684 |
"""
|
685 |
try:
|
686 |
inst_os = OSFromDisk(instance.os) |
687 |
except errors.InvalidOS, err:
|
688 |
os_name, os_dir, os_err = err.args |
689 |
if os_dir is None: |
690 |
return (False, "Can't find OS '%s': %s" % (os_name, os_err)) |
691 |
else:
|
692 |
return (False, "Error parsing OS '%s' in directory %s: %s" % |
693 |
(os_name, os_dir, os_err)) |
694 |
|
695 |
create_env = OSEnvironment(instance) |
696 |
|
697 |
logfile = "%s/add-%s-%s-%d.log" % (constants.LOG_OS_DIR, instance.os,
|
698 |
instance.name, int(time.time()))
|
699 |
|
700 |
result = utils.RunCmd([inst_os.create_script], env=create_env, |
701 |
cwd=inst_os.path, output=logfile,) |
702 |
if result.failed:
|
703 |
logging.error("os create command '%s' returned error: %s, logfile: %s,"
|
704 |
" output: %s", result.cmd, result.fail_reason, logfile,
|
705 |
result.output) |
706 |
lines = [utils.SafeEncode(val) |
707 |
for val in utils.TailFile(logfile, lines=20)] |
708 |
return (False, "OS create script failed (%s), last lines in the" |
709 |
" log file:\n%s" % (result.fail_reason, "\n".join(lines))) |
710 |
|
711 |
return (True, "Successfully installed") |
712 |
|
713 |
|
714 |
def RunRenameInstance(instance, old_name): |
715 |
"""Run the OS rename script for an instance.
|
716 |
|
717 |
@type instance: L{objects.Instance}
|
718 |
@param instance: Instance whose OS is to be installed
|
719 |
@type old_name: string
|
720 |
@param old_name: previous instance name
|
721 |
@rtype: boolean
|
722 |
@return: the success of the operation
|
723 |
|
724 |
"""
|
725 |
inst_os = OSFromDisk(instance.os) |
726 |
|
727 |
rename_env = OSEnvironment(instance) |
728 |
rename_env['OLD_INSTANCE_NAME'] = old_name
|
729 |
|
730 |
logfile = "%s/rename-%s-%s-%s-%d.log" % (constants.LOG_OS_DIR, instance.os,
|
731 |
old_name, |
732 |
instance.name, int(time.time()))
|
733 |
|
734 |
result = utils.RunCmd([inst_os.rename_script], env=rename_env, |
735 |
cwd=inst_os.path, output=logfile) |
736 |
|
737 |
if result.failed:
|
738 |
logging.error("os create command '%s' returned error: %s output: %s",
|
739 |
result.cmd, result.fail_reason, result.output) |
740 |
lines = [utils.SafeEncode(val) |
741 |
for val in utils.TailFile(logfile, lines=20)] |
742 |
return (False, "OS rename script failed (%s), last lines in the" |
743 |
" log file:\n%s" % (result.fail_reason, "\n".join(lines))) |
744 |
|
745 |
return (True, "Rename successful") |
746 |
|
747 |
|
748 |
def _GetVGInfo(vg_name): |
749 |
"""Get informations about the volume group.
|
750 |
|
751 |
@type vg_name: str
|
752 |
@param vg_name: the volume group which we query
|
753 |
@rtype: dict
|
754 |
@return:
|
755 |
A dictionary with the following keys:
|
756 |
- C{vg_size} is the total size of the volume group in MiB
|
757 |
- C{vg_free} is the free size of the volume group in MiB
|
758 |
- C{pv_count} are the number of physical disks in that VG
|
759 |
|
760 |
If an error occurs during gathering of data, we return the same dict
|
761 |
with keys all set to None.
|
762 |
|
763 |
"""
|
764 |
retdic = dict.fromkeys(["vg_size", "vg_free", "pv_count"]) |
765 |
|
766 |
retval = utils.RunCmd(["vgs", "-ovg_size,vg_free,pv_count", "--noheadings", |
767 |
"--nosuffix", "--units=m", "--separator=:", vg_name]) |
768 |
|
769 |
if retval.failed:
|
770 |
logging.error("volume group %s not present", vg_name)
|
771 |
return retdic
|
772 |
valarr = retval.stdout.strip().rstrip(':').split(':') |
773 |
if len(valarr) == 3: |
774 |
try:
|
775 |
retdic = { |
776 |
"vg_size": int(round(float(valarr[0]), 0)), |
777 |
"vg_free": int(round(float(valarr[1]), 0)), |
778 |
"pv_count": int(valarr[2]), |
779 |
} |
780 |
except ValueError, err: |
781 |
logging.exception("Fail to parse vgs output")
|
782 |
else:
|
783 |
logging.error("vgs output has the wrong number of fields (expected"
|
784 |
" three): %s", str(valarr)) |
785 |
return retdic
|
786 |
|
787 |
|
788 |
def _GetBlockDevSymlinkPath(instance_name, idx): |
789 |
return os.path.join(constants.DISK_LINKS_DIR,
|
790 |
"%s:%d" % (instance_name, idx))
|
791 |
|
792 |
|
793 |
def _SymlinkBlockDev(instance_name, device_path, idx): |
794 |
"""Set up symlinks to a instance's block device.
|
795 |
|
796 |
This is an auxiliary function run when an instance is start (on the primary
|
797 |
node) or when an instance is migrated (on the target node).
|
798 |
|
799 |
|
800 |
@param instance_name: the name of the target instance
|
801 |
@param device_path: path of the physical block device, on the node
|
802 |
@param idx: the disk index
|
803 |
@return: absolute path to the disk's symlink
|
804 |
|
805 |
"""
|
806 |
link_name = _GetBlockDevSymlinkPath(instance_name, idx) |
807 |
try:
|
808 |
os.symlink(device_path, link_name) |
809 |
except OSError, err: |
810 |
if err.errno == errno.EEXIST:
|
811 |
if (not os.path.islink(link_name) or |
812 |
os.readlink(link_name) != device_path): |
813 |
os.remove(link_name) |
814 |
os.symlink(device_path, link_name) |
815 |
else:
|
816 |
raise
|
817 |
|
818 |
return link_name
|
819 |
|
820 |
|
821 |
def _RemoveBlockDevLinks(instance_name, disks): |
822 |
"""Remove the block device symlinks belonging to the given instance.
|
823 |
|
824 |
"""
|
825 |
for idx, disk in enumerate(disks): |
826 |
link_name = _GetBlockDevSymlinkPath(instance_name, idx) |
827 |
if os.path.islink(link_name):
|
828 |
try:
|
829 |
os.remove(link_name) |
830 |
except OSError: |
831 |
logging.exception("Can't remove symlink '%s'", link_name)
|
832 |
|
833 |
|
834 |
def _GatherAndLinkBlockDevs(instance): |
835 |
"""Set up an instance's block device(s).
|
836 |
|
837 |
This is run on the primary node at instance startup. The block
|
838 |
devices must be already assembled.
|
839 |
|
840 |
@type instance: L{objects.Instance}
|
841 |
@param instance: the instance whose disks we shoul assemble
|
842 |
@rtype: list
|
843 |
@return: list of (disk_object, device_path)
|
844 |
|
845 |
"""
|
846 |
block_devices = [] |
847 |
for idx, disk in enumerate(instance.disks): |
848 |
device = _RecursiveFindBD(disk) |
849 |
if device is None: |
850 |
raise errors.BlockDeviceError("Block device '%s' is not set up." % |
851 |
str(disk))
|
852 |
device.Open() |
853 |
try:
|
854 |
link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx) |
855 |
except OSError, e: |
856 |
raise errors.BlockDeviceError("Cannot create block device symlink: %s" % |
857 |
e.strerror) |
858 |
|
859 |
block_devices.append((disk, link_name)) |
860 |
|
861 |
return block_devices
|
862 |
|
863 |
|
864 |
def StartInstance(instance): |
865 |
"""Start an instance.
|
866 |
|
867 |
@type instance: L{objects.Instance}
|
868 |
@param instance: the instance object
|
869 |
@rtype: boolean
|
870 |
@return: whether the startup was successful or not
|
871 |
|
872 |
"""
|
873 |
running_instances = GetInstanceList([instance.hypervisor]) |
874 |
|
875 |
if instance.name in running_instances: |
876 |
return (True, "Already running") |
877 |
|
878 |
try:
|
879 |
block_devices = _GatherAndLinkBlockDevs(instance) |
880 |
hyper = hypervisor.GetHypervisor(instance.hypervisor) |
881 |
hyper.StartInstance(instance, block_devices) |
882 |
except errors.BlockDeviceError, err:
|
883 |
logging.exception("Failed to start instance")
|
884 |
return (False, "Block device error: %s" % str(err)) |
885 |
except errors.HypervisorError, err:
|
886 |
logging.exception("Failed to start instance")
|
887 |
_RemoveBlockDevLinks(instance.name, instance.disks) |
888 |
return (False, "Hypervisor error: %s" % str(err)) |
889 |
|
890 |
return (True, "Instance started successfully") |
891 |
|
892 |
|
893 |
def InstanceShutdown(instance): |
894 |
"""Shut an instance down.
|
895 |
|
896 |
@note: this functions uses polling with a hardcoded timeout.
|
897 |
|
898 |
@type instance: L{objects.Instance}
|
899 |
@param instance: the instance object
|
900 |
@rtype: boolean
|
901 |
@return: whether the startup was successful or not
|
902 |
|
903 |
"""
|
904 |
hv_name = instance.hypervisor |
905 |
running_instances = GetInstanceList([hv_name]) |
906 |
|
907 |
if instance.name not in running_instances: |
908 |
return (True, "Instance already stopped") |
909 |
|
910 |
hyper = hypervisor.GetHypervisor(hv_name) |
911 |
try:
|
912 |
hyper.StopInstance(instance) |
913 |
except errors.HypervisorError, err:
|
914 |
msg = "Failed to stop instance %s: %s" % (instance.name, err)
|
915 |
logging.error(msg) |
916 |
return (False, msg) |
917 |
|
918 |
# test every 10secs for 2min
|
919 |
|
920 |
time.sleep(1)
|
921 |
for dummy in range(11): |
922 |
if instance.name not in GetInstanceList([hv_name]): |
923 |
break
|
924 |
time.sleep(10)
|
925 |
else:
|
926 |
# the shutdown did not succeed
|
927 |
logging.error("Shutdown of '%s' unsuccessful, using destroy",
|
928 |
instance.name) |
929 |
|
930 |
try:
|
931 |
hyper.StopInstance(instance, force=True)
|
932 |
except errors.HypervisorError, err:
|
933 |
msg = "Failed to force stop instance %s: %s" % (instance.name, err)
|
934 |
logging.error(msg) |
935 |
return (False, msg) |
936 |
|
937 |
time.sleep(1)
|
938 |
if instance.name in GetInstanceList([hv_name]): |
939 |
msg = ("Could not shutdown instance %s even by destroy" %
|
940 |
instance.name) |
941 |
logging.error(msg) |
942 |
return (False, msg) |
943 |
|
944 |
_RemoveBlockDevLinks(instance.name, instance.disks) |
945 |
|
946 |
return (True, "Instance has been shutdown successfully") |
947 |
|
948 |
|
949 |
def InstanceReboot(instance, reboot_type): |
950 |
"""Reboot an instance.
|
951 |
|
952 |
@type instance: L{objects.Instance}
|
953 |
@param instance: the instance object to reboot
|
954 |
@type reboot_type: str
|
955 |
@param reboot_type: the type of reboot, one the following
|
956 |
constants:
|
957 |
- L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
|
958 |
instance OS, do not recreate the VM
|
959 |
- L{constants.INSTANCE_REBOOT_HARD}: tear down and
|
960 |
restart the VM (at the hypervisor level)
|
961 |
- the other reboot type (L{constants.INSTANCE_REBOOT_HARD})
|
962 |
is not accepted here, since that mode is handled
|
963 |
differently
|
964 |
@rtype: boolean
|
965 |
@return: the success of the operation
|
966 |
|
967 |
"""
|
968 |
running_instances = GetInstanceList([instance.hypervisor]) |
969 |
|
970 |
if instance.name not in running_instances: |
971 |
msg = "Cannot reboot instance %s that is not running" % instance.name
|
972 |
logging.error(msg) |
973 |
return (False, msg) |
974 |
|
975 |
hyper = hypervisor.GetHypervisor(instance.hypervisor) |
976 |
if reboot_type == constants.INSTANCE_REBOOT_SOFT:
|
977 |
try:
|
978 |
hyper.RebootInstance(instance) |
979 |
except errors.HypervisorError, err:
|
980 |
msg = "Failed to soft reboot instance %s: %s" % (instance.name, err)
|
981 |
logging.error(msg) |
982 |
return (False, msg) |
983 |
elif reboot_type == constants.INSTANCE_REBOOT_HARD:
|
984 |
try:
|
985 |
stop_result = InstanceShutdown(instance) |
986 |
if not stop_result[0]: |
987 |
return stop_result
|
988 |
return StartInstance(instance)
|
989 |
except errors.HypervisorError, err:
|
990 |
msg = "Failed to hard reboot instance %s: %s" % (instance.name, err)
|
991 |
logging.error(msg) |
992 |
return (False, msg) |
993 |
else:
|
994 |
return (False, "Invalid reboot_type received: %s" % (reboot_type,)) |
995 |
|
996 |
return (True, "Reboot successful") |
997 |
|
998 |
|
999 |
def MigrationInfo(instance): |
1000 |
"""Gather information about an instance to be migrated.
|
1001 |
|
1002 |
@type instance: L{objects.Instance}
|
1003 |
@param instance: the instance definition
|
1004 |
|
1005 |
"""
|
1006 |
hyper = hypervisor.GetHypervisor(instance.hypervisor) |
1007 |
try:
|
1008 |
info = hyper.MigrationInfo(instance) |
1009 |
except errors.HypervisorError, err:
|
1010 |
msg = "Failed to fetch migration information"
|
1011 |
logging.exception(msg) |
1012 |
return (False, '%s: %s' % (msg, err)) |
1013 |
return (True, info) |
1014 |
|
1015 |
|
1016 |
def AcceptInstance(instance, info, target): |
1017 |
"""Prepare the node to accept an instance.
|
1018 |
|
1019 |
@type instance: L{objects.Instance}
|
1020 |
@param instance: the instance definition
|
1021 |
@type info: string/data (opaque)
|
1022 |
@param info: migration information, from the source node
|
1023 |
@type target: string
|
1024 |
@param target: target host (usually ip), on this node
|
1025 |
|
1026 |
"""
|
1027 |
hyper = hypervisor.GetHypervisor(instance.hypervisor) |
1028 |
try:
|
1029 |
hyper.AcceptInstance(instance, info, target) |
1030 |
except errors.HypervisorError, err:
|
1031 |
msg = "Failed to accept instance"
|
1032 |
logging.exception(msg) |
1033 |
return (False, '%s: %s' % (msg, err)) |
1034 |
return (True, "Accept successfull") |
1035 |
|
1036 |
|
1037 |
def FinalizeMigration(instance, info, success): |
1038 |
"""Finalize any preparation to accept an instance.
|
1039 |
|
1040 |
@type instance: L{objects.Instance}
|
1041 |
@param instance: the instance definition
|
1042 |
@type info: string/data (opaque)
|
1043 |
@param info: migration information, from the source node
|
1044 |
@type success: boolean
|
1045 |
@param success: whether the migration was a success or a failure
|
1046 |
|
1047 |
"""
|
1048 |
hyper = hypervisor.GetHypervisor(instance.hypervisor) |
1049 |
try:
|
1050 |
hyper.FinalizeMigration(instance, info, success) |
1051 |
except errors.HypervisorError, err:
|
1052 |
msg = "Failed to finalize migration"
|
1053 |
logging.exception(msg) |
1054 |
return (False, '%s: %s' % (msg, err)) |
1055 |
return (True, "Migration Finalized") |
1056 |
|
1057 |
|
1058 |
def MigrateInstance(instance, target, live): |
1059 |
"""Migrates an instance to another node.
|
1060 |
|
1061 |
@type instance: L{objects.Instance}
|
1062 |
@param instance: the instance definition
|
1063 |
@type target: string
|
1064 |
@param target: the target node name
|
1065 |
@type live: boolean
|
1066 |
@param live: whether the migration should be done live or not (the
|
1067 |
interpretation of this parameter is left to the hypervisor)
|
1068 |
@rtype: tuple
|
1069 |
@return: a tuple of (success, msg) where:
|
1070 |
- succes is a boolean denoting the success/failure of the operation
|
1071 |
- msg is a string with details in case of failure
|
1072 |
|
1073 |
"""
|
1074 |
hyper = hypervisor.GetHypervisor(instance.hypervisor) |
1075 |
|
1076 |
try:
|
1077 |
hyper.MigrateInstance(instance.name, target, live) |
1078 |
except errors.HypervisorError, err:
|
1079 |
msg = "Failed to migrate instance"
|
1080 |
logging.exception(msg) |
1081 |
return (False, "%s: %s" % (msg, err)) |
1082 |
return (True, "Migration successfull") |
1083 |
|
1084 |
|
1085 |
def BlockdevCreate(disk, size, owner, on_primary, info): |
1086 |
"""Creates a block device for an instance.
|
1087 |
|
1088 |
@type disk: L{objects.Disk}
|
1089 |
@param disk: the object describing the disk we should create
|
1090 |
@type size: int
|
1091 |
@param size: the size of the physical underlying device, in MiB
|
1092 |
@type owner: str
|
1093 |
@param owner: the name of the instance for which disk is created,
|
1094 |
used for device cache data
|
1095 |
@type on_primary: boolean
|
1096 |
@param on_primary: indicates if it is the primary node or not
|
1097 |
@type info: string
|
1098 |
@param info: string that will be sent to the physical device
|
1099 |
creation, used for example to set (LVM) tags on LVs
|
1100 |
|
1101 |
@return: the new unique_id of the device (this can sometime be
|
1102 |
computed only after creation), or None. On secondary nodes,
|
1103 |
it's not required to return anything.
|
1104 |
|
1105 |
"""
|
1106 |
clist = [] |
1107 |
if disk.children:
|
1108 |
for child in disk.children: |
1109 |
try:
|
1110 |
crdev = _RecursiveAssembleBD(child, owner, on_primary) |
1111 |
except errors.BlockDeviceError, err:
|
1112 |
errmsg = "Can't assemble device %s: %s" % (child, err)
|
1113 |
logging.error(errmsg) |
1114 |
return False, errmsg |
1115 |
if on_primary or disk.AssembleOnSecondary(): |
1116 |
# we need the children open in case the device itself has to
|
1117 |
# be assembled
|
1118 |
try:
|
1119 |
crdev.Open() |
1120 |
except errors.BlockDeviceError, err:
|
1121 |
errmsg = "Can't make child '%s' read-write: %s" % (child, err)
|
1122 |
logging.error(errmsg) |
1123 |
return False, errmsg |
1124 |
clist.append(crdev) |
1125 |
|
1126 |
try:
|
1127 |
device = bdev.Create(disk.dev_type, disk.physical_id, clist, disk.size) |
1128 |
except errors.BlockDeviceError, err:
|
1129 |
return False, "Can't create block device: %s" % str(err) |
1130 |
|
1131 |
if on_primary or disk.AssembleOnSecondary(): |
1132 |
try:
|
1133 |
device.Assemble() |
1134 |
except errors.BlockDeviceError, err:
|
1135 |
errmsg = ("Can't assemble device after creation, very"
|
1136 |
" unusual event: %s" % str(err)) |
1137 |
logging.error(errmsg) |
1138 |
return False, errmsg |
1139 |
device.SetSyncSpeed(constants.SYNC_SPEED) |
1140 |
if on_primary or disk.OpenOnSecondary(): |
1141 |
try:
|
1142 |
device.Open(force=True)
|
1143 |
except errors.BlockDeviceError, err:
|
1144 |
errmsg = ("Can't make device r/w after creation, very"
|
1145 |
" unusual event: %s" % str(err)) |
1146 |
logging.error(errmsg) |
1147 |
return False, errmsg |
1148 |
DevCacheManager.UpdateCache(device.dev_path, owner, |
1149 |
on_primary, disk.iv_name) |
1150 |
|
1151 |
device.SetInfo(info) |
1152 |
|
1153 |
physical_id = device.unique_id |
1154 |
return True, physical_id |
1155 |
|
1156 |
|
1157 |
def BlockdevRemove(disk): |
1158 |
"""Remove a block device.
|
1159 |
|
1160 |
@note: This is intended to be called recursively.
|
1161 |
|
1162 |
@type disk: L{objects.Disk}
|
1163 |
@param disk: the disk object we should remove
|
1164 |
@rtype: boolean
|
1165 |
@return: the success of the operation
|
1166 |
|
1167 |
"""
|
1168 |
msgs = [] |
1169 |
result = True
|
1170 |
try:
|
1171 |
rdev = _RecursiveFindBD(disk) |
1172 |
except errors.BlockDeviceError, err:
|
1173 |
# probably can't attach
|
1174 |
logging.info("Can't attach to device %s in remove", disk)
|
1175 |
rdev = None
|
1176 |
if rdev is not None: |
1177 |
r_path = rdev.dev_path |
1178 |
try:
|
1179 |
rdev.Remove() |
1180 |
except errors.BlockDeviceError, err:
|
1181 |
msgs.append(str(err))
|
1182 |
result = False
|
1183 |
if result:
|
1184 |
DevCacheManager.RemoveCache(r_path) |
1185 |
|
1186 |
if disk.children:
|
1187 |
for child in disk.children: |
1188 |
c_status, c_msg = BlockdevRemove(child) |
1189 |
result = result and c_status
|
1190 |
if c_msg: # not an empty message |
1191 |
msgs.append(c_msg) |
1192 |
|
1193 |
return (result, "; ".join(msgs)) |
1194 |
|
1195 |
|
1196 |
def _RecursiveAssembleBD(disk, owner, as_primary): |
1197 |
"""Activate a block device for an instance.
|
1198 |
|
1199 |
This is run on the primary and secondary nodes for an instance.
|
1200 |
|
1201 |
@note: this function is called recursively.
|
1202 |
|
1203 |
@type disk: L{objects.Disk}
|
1204 |
@param disk: the disk we try to assemble
|
1205 |
@type owner: str
|
1206 |
@param owner: the name of the instance which owns the disk
|
1207 |
@type as_primary: boolean
|
1208 |
@param as_primary: if we should make the block device
|
1209 |
read/write
|
1210 |
|
1211 |
@return: the assembled device or None (in case no device
|
1212 |
was assembled)
|
1213 |
@raise errors.BlockDeviceError: in case there is an error
|
1214 |
during the activation of the children or the device
|
1215 |
itself
|
1216 |
|
1217 |
"""
|
1218 |
children = [] |
1219 |
if disk.children:
|
1220 |
mcn = disk.ChildrenNeeded() |
1221 |
if mcn == -1: |
1222 |
mcn = 0 # max number of Nones allowed |
1223 |
else:
|
1224 |
mcn = len(disk.children) - mcn # max number of Nones |
1225 |
for chld_disk in disk.children: |
1226 |
try:
|
1227 |
cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary) |
1228 |
except errors.BlockDeviceError, err:
|
1229 |
if children.count(None) >= mcn: |
1230 |
raise
|
1231 |
cdev = None
|
1232 |
logging.error("Error in child activation (but continuing): %s",
|
1233 |
str(err))
|
1234 |
children.append(cdev) |
1235 |
|
1236 |
if as_primary or disk.AssembleOnSecondary(): |
1237 |
r_dev = bdev.Assemble(disk.dev_type, disk.physical_id, children, disk.size) |
1238 |
r_dev.SetSyncSpeed(constants.SYNC_SPEED) |
1239 |
result = r_dev |
1240 |
if as_primary or disk.OpenOnSecondary(): |
1241 |
r_dev.Open() |
1242 |
DevCacheManager.UpdateCache(r_dev.dev_path, owner, |
1243 |
as_primary, disk.iv_name) |
1244 |
|
1245 |
else:
|
1246 |
result = True
|
1247 |
return result
|
1248 |
|
1249 |
|
1250 |
def BlockdevAssemble(disk, owner, as_primary): |
1251 |
"""Activate a block device for an instance.
|
1252 |
|
1253 |
This is a wrapper over _RecursiveAssembleBD.
|
1254 |
|
1255 |
@rtype: str or boolean
|
1256 |
@return: a C{/dev/...} path for primary nodes, and
|
1257 |
C{True} for secondary nodes
|
1258 |
|
1259 |
"""
|
1260 |
status = True
|
1261 |
result = "no error information"
|
1262 |
try:
|
1263 |
result = _RecursiveAssembleBD(disk, owner, as_primary) |
1264 |
if isinstance(result, bdev.BlockDev): |
1265 |
result = result.dev_path |
1266 |
except errors.BlockDeviceError, err:
|
1267 |
result = "Error while assembling disk: %s" % str(err) |
1268 |
status = False
|
1269 |
return (status, result)
|
1270 |
|
1271 |
|
1272 |
def BlockdevShutdown(disk): |
1273 |
"""Shut down a block device.
|
1274 |
|
1275 |
First, if the device is assembled (Attach() is successfull), then
|
1276 |
the device is shutdown. Then the children of the device are
|
1277 |
shutdown.
|
1278 |
|
1279 |
This function is called recursively. Note that we don't cache the
|
1280 |
children or such, as oppossed to assemble, shutdown of different
|
1281 |
devices doesn't require that the upper device was active.
|
1282 |
|
1283 |
@type disk: L{objects.Disk}
|
1284 |
@param disk: the description of the disk we should
|
1285 |
shutdown
|
1286 |
@rtype: boolean
|
1287 |
@return: the success of the operation
|
1288 |
|
1289 |
"""
|
1290 |
msgs = [] |
1291 |
result = True
|
1292 |
r_dev = _RecursiveFindBD(disk) |
1293 |
if r_dev is not None: |
1294 |
r_path = r_dev.dev_path |
1295 |
try:
|
1296 |
r_dev.Shutdown() |
1297 |
DevCacheManager.RemoveCache(r_path) |
1298 |
except errors.BlockDeviceError, err:
|
1299 |
msgs.append(str(err))
|
1300 |
result = False
|
1301 |
|
1302 |
if disk.children:
|
1303 |
for child in disk.children: |
1304 |
c_status, c_msg = BlockdevShutdown(child) |
1305 |
result = result and c_status
|
1306 |
if c_msg: # not an empty message |
1307 |
msgs.append(c_msg) |
1308 |
|
1309 |
return (result, "; ".join(msgs)) |
1310 |
|
1311 |
|
1312 |
def BlockdevAddchildren(parent_cdev, new_cdevs): |
1313 |
"""Extend a mirrored block device.
|
1314 |
|
1315 |
@type parent_cdev: L{objects.Disk}
|
1316 |
@param parent_cdev: the disk to which we should add children
|
1317 |
@type new_cdevs: list of L{objects.Disk}
|
1318 |
@param new_cdevs: the list of children which we should add
|
1319 |
@rtype: boolean
|
1320 |
@return: the success of the operation
|
1321 |
|
1322 |
"""
|
1323 |
parent_bdev = _RecursiveFindBD(parent_cdev) |
1324 |
if parent_bdev is None: |
1325 |
logging.error("Can't find parent device")
|
1326 |
return False |
1327 |
new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs] |
1328 |
if new_bdevs.count(None) > 0: |
1329 |
logging.error("Can't find new device(s) to add: %s:%s",
|
1330 |
new_bdevs, new_cdevs) |
1331 |
return False |
1332 |
parent_bdev.AddChildren(new_bdevs) |
1333 |
return True |
1334 |
|
1335 |
|
1336 |
def BlockdevRemovechildren(parent_cdev, new_cdevs): |
1337 |
"""Shrink a mirrored block device.
|
1338 |
|
1339 |
@type parent_cdev: L{objects.Disk}
|
1340 |
@param parent_cdev: the disk from which we should remove children
|
1341 |
@type new_cdevs: list of L{objects.Disk}
|
1342 |
@param new_cdevs: the list of children which we should remove
|
1343 |
@rtype: boolean
|
1344 |
@return: the success of the operation
|
1345 |
|
1346 |
"""
|
1347 |
parent_bdev = _RecursiveFindBD(parent_cdev) |
1348 |
if parent_bdev is None: |
1349 |
logging.error("Can't find parent in remove children: %s", parent_cdev)
|
1350 |
return False |
1351 |
devs = [] |
1352 |
for disk in new_cdevs: |
1353 |
rpath = disk.StaticDevPath() |
1354 |
if rpath is None: |
1355 |
bd = _RecursiveFindBD(disk) |
1356 |
if bd is None: |
1357 |
logging.error("Can't find dynamic device %s while removing children",
|
1358 |
disk) |
1359 |
return False |
1360 |
else:
|
1361 |
devs.append(bd.dev_path) |
1362 |
else:
|
1363 |
devs.append(rpath) |
1364 |
parent_bdev.RemoveChildren(devs) |
1365 |
return True |
1366 |
|
1367 |
|
1368 |
def BlockdevGetmirrorstatus(disks): |
1369 |
"""Get the mirroring status of a list of devices.
|
1370 |
|
1371 |
@type disks: list of L{objects.Disk}
|
1372 |
@param disks: the list of disks which we should query
|
1373 |
@rtype: disk
|
1374 |
@return:
|
1375 |
a list of (mirror_done, estimated_time) tuples, which
|
1376 |
are the result of L{bdev.BlockDev.CombinedSyncStatus}
|
1377 |
@raise errors.BlockDeviceError: if any of the disks cannot be
|
1378 |
found
|
1379 |
|
1380 |
"""
|
1381 |
stats = [] |
1382 |
for dsk in disks: |
1383 |
rbd = _RecursiveFindBD(dsk) |
1384 |
if rbd is None: |
1385 |
raise errors.BlockDeviceError("Can't find device %s" % str(dsk)) |
1386 |
stats.append(rbd.CombinedSyncStatus()) |
1387 |
return stats
|
1388 |
|
1389 |
|
1390 |
def _RecursiveFindBD(disk): |
1391 |
"""Check if a device is activated.
|
1392 |
|
1393 |
If so, return informations about the real device.
|
1394 |
|
1395 |
@type disk: L{objects.Disk}
|
1396 |
@param disk: the disk object we need to find
|
1397 |
|
1398 |
@return: None if the device can't be found,
|
1399 |
otherwise the device instance
|
1400 |
|
1401 |
"""
|
1402 |
children = [] |
1403 |
if disk.children:
|
1404 |
for chdisk in disk.children: |
1405 |
children.append(_RecursiveFindBD(chdisk)) |
1406 |
|
1407 |
return bdev.FindDevice(disk.dev_type, disk.physical_id, children, disk.size)
|
1408 |
|
1409 |
|
1410 |
def BlockdevFind(disk): |
1411 |
"""Check if a device is activated.
|
1412 |
|
1413 |
If it is, return informations about the real device.
|
1414 |
|
1415 |
@type disk: L{objects.Disk}
|
1416 |
@param disk: the disk to find
|
1417 |
@rtype: None or tuple
|
1418 |
@return: None if the disk cannot be found, otherwise a
|
1419 |
tuple (device_path, major, minor, sync_percent,
|
1420 |
estimated_time, is_degraded)
|
1421 |
|
1422 |
"""
|
1423 |
try:
|
1424 |
rbd = _RecursiveFindBD(disk) |
1425 |
except errors.BlockDeviceError, err:
|
1426 |
return (False, str(err)) |
1427 |
if rbd is None: |
1428 |
return (True, None) |
1429 |
return (True, (rbd.dev_path, rbd.major, rbd.minor) + rbd.GetSyncStatus()) |
1430 |
|
1431 |
|
1432 |
def UploadFile(file_name, data, mode, uid, gid, atime, mtime): |
1433 |
"""Write a file to the filesystem.
|
1434 |
|
1435 |
This allows the master to overwrite(!) a file. It will only perform
|
1436 |
the operation if the file belongs to a list of configuration files.
|
1437 |
|
1438 |
@type file_name: str
|
1439 |
@param file_name: the target file name
|
1440 |
@type data: str
|
1441 |
@param data: the new contents of the file
|
1442 |
@type mode: int
|
1443 |
@param mode: the mode to give the file (can be None)
|
1444 |
@type uid: int
|
1445 |
@param uid: the owner of the file (can be -1 for default)
|
1446 |
@type gid: int
|
1447 |
@param gid: the group of the file (can be -1 for default)
|
1448 |
@type atime: float
|
1449 |
@param atime: the atime to set on the file (can be None)
|
1450 |
@type mtime: float
|
1451 |
@param mtime: the mtime to set on the file (can be None)
|
1452 |
@rtype: boolean
|
1453 |
@return: the success of the operation; errors are logged
|
1454 |
in the node daemon log
|
1455 |
|
1456 |
"""
|
1457 |
if not os.path.isabs(file_name): |
1458 |
logging.error("Filename passed to UploadFile is not absolute: '%s'",
|
1459 |
file_name) |
1460 |
return False |
1461 |
|
1462 |
allowed_files = [ |
1463 |
constants.CLUSTER_CONF_FILE, |
1464 |
constants.ETC_HOSTS, |
1465 |
constants.SSH_KNOWN_HOSTS_FILE, |
1466 |
constants.VNC_PASSWORD_FILE, |
1467 |
] |
1468 |
|
1469 |
if file_name not in allowed_files: |
1470 |
logging.error("Filename passed to UploadFile not in allowed"
|
1471 |
" upload targets: '%s'", file_name)
|
1472 |
return False |
1473 |
|
1474 |
raw_data = _Decompress(data) |
1475 |
|
1476 |
utils.WriteFile(file_name, data=raw_data, mode=mode, uid=uid, gid=gid, |
1477 |
atime=atime, mtime=mtime) |
1478 |
return True |
1479 |
|
1480 |
|
1481 |
def WriteSsconfFiles(values): |
1482 |
"""Update all ssconf files.
|
1483 |
|
1484 |
Wrapper around the SimpleStore.WriteFiles.
|
1485 |
|
1486 |
"""
|
1487 |
ssconf.SimpleStore().WriteFiles(values) |
1488 |
|
1489 |
|
1490 |
def _ErrnoOrStr(err): |
1491 |
"""Format an EnvironmentError exception.
|
1492 |
|
1493 |
If the L{err} argument has an errno attribute, it will be looked up
|
1494 |
and converted into a textual C{E...} description. Otherwise the
|
1495 |
string representation of the error will be returned.
|
1496 |
|
1497 |
@type err: L{EnvironmentError}
|
1498 |
@param err: the exception to format
|
1499 |
|
1500 |
"""
|
1501 |
if hasattr(err, 'errno'): |
1502 |
detail = errno.errorcode[err.errno] |
1503 |
else:
|
1504 |
detail = str(err)
|
1505 |
return detail
|
1506 |
|
1507 |
|
1508 |
def _OSOndiskVersion(name, os_dir): |
1509 |
"""Compute and return the API version of a given OS.
|
1510 |
|
1511 |
This function will try to read the API version of the OS given by
|
1512 |
the 'name' parameter and residing in the 'os_dir' directory.
|
1513 |
|
1514 |
@type name: str
|
1515 |
@param name: the OS name we should look for
|
1516 |
@type os_dir: str
|
1517 |
@param os_dir: the directory inwhich we should look for the OS
|
1518 |
@rtype: int or None
|
1519 |
@return:
|
1520 |
Either an integer denoting the version or None in the
|
1521 |
case when this is not a valid OS name.
|
1522 |
@raise errors.InvalidOS: if the OS cannot be found
|
1523 |
|
1524 |
"""
|
1525 |
api_file = os.path.sep.join([os_dir, "ganeti_api_version"])
|
1526 |
|
1527 |
try:
|
1528 |
st = os.stat(api_file) |
1529 |
except EnvironmentError, err: |
1530 |
raise errors.InvalidOS(name, os_dir, "'ganeti_api_version' file not" |
1531 |
" found (%s)" % _ErrnoOrStr(err))
|
1532 |
|
1533 |
if not stat.S_ISREG(stat.S_IFMT(st.st_mode)): |
1534 |
raise errors.InvalidOS(name, os_dir, "'ganeti_api_version' file is not" |
1535 |
" a regular file")
|
1536 |
|
1537 |
try:
|
1538 |
f = open(api_file)
|
1539 |
try:
|
1540 |
api_versions = f.readlines() |
1541 |
finally:
|
1542 |
f.close() |
1543 |
except EnvironmentError, err: |
1544 |
raise errors.InvalidOS(name, os_dir, "error while reading the" |
1545 |
" API version (%s)" % _ErrnoOrStr(err))
|
1546 |
|
1547 |
api_versions = [version.strip() for version in api_versions] |
1548 |
try:
|
1549 |
api_versions = [int(version) for version in api_versions] |
1550 |
except (TypeError, ValueError), err: |
1551 |
raise errors.InvalidOS(name, os_dir,
|
1552 |
"API version is not integer (%s)" % str(err)) |
1553 |
|
1554 |
return api_versions
|
1555 |
|
1556 |
|
1557 |
def DiagnoseOS(top_dirs=None): |
1558 |
"""Compute the validity for all OSes.
|
1559 |
|
1560 |
@type top_dirs: list
|
1561 |
@param top_dirs: the list of directories in which to
|
1562 |
search (if not given defaults to
|
1563 |
L{constants.OS_SEARCH_PATH})
|
1564 |
@rtype: list of L{objects.OS}
|
1565 |
@return: an OS object for each name in all the given
|
1566 |
directories
|
1567 |
|
1568 |
"""
|
1569 |
if top_dirs is None: |
1570 |
top_dirs = constants.OS_SEARCH_PATH |
1571 |
|
1572 |
result = [] |
1573 |
for dir_name in top_dirs: |
1574 |
if os.path.isdir(dir_name):
|
1575 |
try:
|
1576 |
f_names = utils.ListVisibleFiles(dir_name) |
1577 |
except EnvironmentError, err: |
1578 |
logging.exception("Can't list the OS directory %s", dir_name)
|
1579 |
break
|
1580 |
for name in f_names: |
1581 |
try:
|
1582 |
os_inst = OSFromDisk(name, base_dir=dir_name) |
1583 |
result.append(os_inst) |
1584 |
except errors.InvalidOS, err:
|
1585 |
result.append(objects.OS.FromInvalidOS(err)) |
1586 |
|
1587 |
return result
|
1588 |
|
1589 |
|
1590 |
def OSFromDisk(name, base_dir=None): |
1591 |
"""Create an OS instance from disk.
|
1592 |
|
1593 |
This function will return an OS instance if the given name is a
|
1594 |
valid OS name. Otherwise, it will raise an appropriate
|
1595 |
L{errors.InvalidOS} exception, detailing why this is not a valid OS.
|
1596 |
|
1597 |
@type base_dir: string
|
1598 |
@keyword base_dir: Base directory containing OS installations.
|
1599 |
Defaults to a search in all the OS_SEARCH_PATH dirs.
|
1600 |
@rtype: L{objects.OS}
|
1601 |
@return: the OS instance if we find a valid one
|
1602 |
@raise errors.InvalidOS: if we don't find a valid OS
|
1603 |
|
1604 |
"""
|
1605 |
if base_dir is None: |
1606 |
os_dir = utils.FindFile(name, constants.OS_SEARCH_PATH, os.path.isdir) |
1607 |
if os_dir is None: |
1608 |
raise errors.InvalidOS(name, None, "OS dir not found in search path") |
1609 |
else:
|
1610 |
os_dir = os.path.sep.join([base_dir, name]) |
1611 |
|
1612 |
api_versions = _OSOndiskVersion(name, os_dir) |
1613 |
|
1614 |
if constants.OS_API_VERSION not in api_versions: |
1615 |
raise errors.InvalidOS(name, os_dir, "API version mismatch" |
1616 |
" (found %s want %s)"
|
1617 |
% (api_versions, constants.OS_API_VERSION)) |
1618 |
|
1619 |
# OS Scripts dictionary, we will populate it with the actual script names
|
1620 |
os_scripts = dict.fromkeys(constants.OS_SCRIPTS)
|
1621 |
|
1622 |
for script in os_scripts: |
1623 |
os_scripts[script] = os.path.sep.join([os_dir, script]) |
1624 |
|
1625 |
try:
|
1626 |
st = os.stat(os_scripts[script]) |
1627 |
except EnvironmentError, err: |
1628 |
raise errors.InvalidOS(name, os_dir, "'%s' script missing (%s)" % |
1629 |
(script, _ErrnoOrStr(err))) |
1630 |
|
1631 |
if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
|
1632 |
raise errors.InvalidOS(name, os_dir, "'%s' script not executable" % |
1633 |
script) |
1634 |
|
1635 |
if not stat.S_ISREG(stat.S_IFMT(st.st_mode)): |
1636 |
raise errors.InvalidOS(name, os_dir, "'%s' is not a regular file" % |
1637 |
script) |
1638 |
|
1639 |
|
1640 |
return objects.OS(name=name, path=os_dir, status=constants.OS_VALID_STATUS,
|
1641 |
create_script=os_scripts[constants.OS_SCRIPT_CREATE], |
1642 |
export_script=os_scripts[constants.OS_SCRIPT_EXPORT], |
1643 |
import_script=os_scripts[constants.OS_SCRIPT_IMPORT], |
1644 |
rename_script=os_scripts[constants.OS_SCRIPT_RENAME], |
1645 |
api_versions=api_versions) |
1646 |
|
1647 |
def OSEnvironment(instance, debug=0): |
1648 |
"""Calculate the environment for an os script.
|
1649 |
|
1650 |
@type instance: L{objects.Instance}
|
1651 |
@param instance: target instance for the os script run
|
1652 |
@type debug: integer
|
1653 |
@param debug: debug level (0 or 1, for OS Api 10)
|
1654 |
@rtype: dict
|
1655 |
@return: dict of environment variables
|
1656 |
@raise errors.BlockDeviceError: if the block device
|
1657 |
cannot be found
|
1658 |
|
1659 |
"""
|
1660 |
result = {} |
1661 |
result['OS_API_VERSION'] = '%d' % constants.OS_API_VERSION |
1662 |
result['INSTANCE_NAME'] = instance.name
|
1663 |
result['INSTANCE_OS'] = instance.os
|
1664 |
result['HYPERVISOR'] = instance.hypervisor
|
1665 |
result['DISK_COUNT'] = '%d' % len(instance.disks) |
1666 |
result['NIC_COUNT'] = '%d' % len(instance.nics) |
1667 |
result['DEBUG_LEVEL'] = '%d' % debug |
1668 |
for idx, disk in enumerate(instance.disks): |
1669 |
real_disk = _RecursiveFindBD(disk) |
1670 |
if real_disk is None: |
1671 |
raise errors.BlockDeviceError("Block device '%s' is not set up" % |
1672 |
str(disk))
|
1673 |
real_disk.Open() |
1674 |
result['DISK_%d_PATH' % idx] = real_disk.dev_path
|
1675 |
result['DISK_%d_ACCESS' % idx] = disk.mode
|
1676 |
if constants.HV_DISK_TYPE in instance.hvparams: |
1677 |
result['DISK_%d_FRONTEND_TYPE' % idx] = \
|
1678 |
instance.hvparams[constants.HV_DISK_TYPE] |
1679 |
if disk.dev_type in constants.LDS_BLOCK: |
1680 |
result['DISK_%d_BACKEND_TYPE' % idx] = 'block' |
1681 |
elif disk.dev_type == constants.LD_FILE:
|
1682 |
result['DISK_%d_BACKEND_TYPE' % idx] = \
|
1683 |
'file:%s' % disk.physical_id[0] |
1684 |
for idx, nic in enumerate(instance.nics): |
1685 |
result['NIC_%d_MAC' % idx] = nic.mac
|
1686 |
if nic.ip:
|
1687 |
result['NIC_%d_IP' % idx] = nic.ip
|
1688 |
result['NIC_%d_BRIDGE' % idx] = nic.bridge
|
1689 |
if constants.HV_NIC_TYPE in instance.hvparams: |
1690 |
result['NIC_%d_FRONTEND_TYPE' % idx] = \
|
1691 |
instance.hvparams[constants.HV_NIC_TYPE] |
1692 |
|
1693 |
for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]: |
1694 |
for key, value in source.items(): |
1695 |
result["INSTANCE_%s_%s" % (kind, key)] = str(value) |
1696 |
|
1697 |
return result
|
1698 |
|
1699 |
def BlockdevGrow(disk, amount): |
1700 |
"""Grow a stack of block devices.
|
1701 |
|
1702 |
This function is called recursively, with the childrens being the
|
1703 |
first ones to resize.
|
1704 |
|
1705 |
@type disk: L{objects.Disk}
|
1706 |
@param disk: the disk to be grown
|
1707 |
@rtype: (status, result)
|
1708 |
@return: a tuple with the status of the operation
|
1709 |
(True/False), and the errors message if status
|
1710 |
is False
|
1711 |
|
1712 |
"""
|
1713 |
r_dev = _RecursiveFindBD(disk) |
1714 |
if r_dev is None: |
1715 |
return False, "Cannot find block device %s" % (disk,) |
1716 |
|
1717 |
try:
|
1718 |
r_dev.Grow(amount) |
1719 |
except errors.BlockDeviceError, err:
|
1720 |
return False, str(err) |
1721 |
|
1722 |
return True, None |
1723 |
|
1724 |
|
1725 |
def BlockdevSnapshot(disk): |
1726 |
"""Create a snapshot copy of a block device.
|
1727 |
|
1728 |
This function is called recursively, and the snapshot is actually created
|
1729 |
just for the leaf lvm backend device.
|
1730 |
|
1731 |
@type disk: L{objects.Disk}
|
1732 |
@param disk: the disk to be snapshotted
|
1733 |
@rtype: string
|
1734 |
@return: snapshot disk path
|
1735 |
|
1736 |
"""
|
1737 |
if disk.children:
|
1738 |
if len(disk.children) == 1: |
1739 |
# only one child, let's recurse on it
|
1740 |
return BlockdevSnapshot(disk.children[0]) |
1741 |
else:
|
1742 |
# more than one child, choose one that matches
|
1743 |
for child in disk.children: |
1744 |
if child.size == disk.size:
|
1745 |
# return implies breaking the loop
|
1746 |
return BlockdevSnapshot(child)
|
1747 |
elif disk.dev_type == constants.LD_LV:
|
1748 |
r_dev = _RecursiveFindBD(disk) |
1749 |
if r_dev is not None: |
1750 |
# let's stay on the safe side and ask for the full size, for now
|
1751 |
return r_dev.Snapshot(disk.size)
|
1752 |
else:
|
1753 |
return None |
1754 |
else:
|
1755 |
raise errors.ProgrammerError("Cannot snapshot non-lvm block device" |
1756 |
" '%s' of type '%s'" %
|
1757 |
(disk.unique_id, disk.dev_type)) |
1758 |
|
1759 |
|
1760 |
def ExportSnapshot(disk, dest_node, instance, cluster_name, idx): |
1761 |
"""Export a block device snapshot to a remote node.
|
1762 |
|
1763 |
@type disk: L{objects.Disk}
|
1764 |
@param disk: the description of the disk to export
|
1765 |
@type dest_node: str
|
1766 |
@param dest_node: the destination node to export to
|
1767 |
@type instance: L{objects.Instance}
|
1768 |
@param instance: the instance object to whom the disk belongs
|
1769 |
@type cluster_name: str
|
1770 |
@param cluster_name: the cluster name, needed for SSH hostalias
|
1771 |
@type idx: int
|
1772 |
@param idx: the index of the disk in the instance's disk list,
|
1773 |
used to export to the OS scripts environment
|
1774 |
@rtype: boolean
|
1775 |
@return: the success of the operation
|
1776 |
|
1777 |
"""
|
1778 |
export_env = OSEnvironment(instance) |
1779 |
|
1780 |
inst_os = OSFromDisk(instance.os) |
1781 |
export_script = inst_os.export_script |
1782 |
|
1783 |
logfile = "%s/exp-%s-%s-%s.log" % (constants.LOG_OS_DIR, inst_os.name,
|
1784 |
instance.name, int(time.time()))
|
1785 |
if not os.path.exists(constants.LOG_OS_DIR): |
1786 |
os.mkdir(constants.LOG_OS_DIR, 0750)
|
1787 |
real_disk = _RecursiveFindBD(disk) |
1788 |
if real_disk is None: |
1789 |
raise errors.BlockDeviceError("Block device '%s' is not set up" % |
1790 |
str(disk))
|
1791 |
real_disk.Open() |
1792 |
|
1793 |
export_env['EXPORT_DEVICE'] = real_disk.dev_path
|
1794 |
export_env['EXPORT_INDEX'] = str(idx) |
1795 |
|
1796 |
destdir = os.path.join(constants.EXPORT_DIR, instance.name + ".new")
|
1797 |
destfile = disk.physical_id[1]
|
1798 |
|
1799 |
# the target command is built out of three individual commands,
|
1800 |
# which are joined by pipes; we check each individual command for
|
1801 |
# valid parameters
|
1802 |
expcmd = utils.BuildShellCmd("cd %s; %s 2>%s", inst_os.path,
|
1803 |
export_script, logfile) |
1804 |
|
1805 |
comprcmd = "gzip"
|
1806 |
|
1807 |
destcmd = utils.BuildShellCmd("mkdir -p %s && cat > %s/%s",
|
1808 |
destdir, destdir, destfile) |
1809 |
remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node, |
1810 |
constants.GANETI_RUNAS, |
1811 |
destcmd) |
1812 |
|
1813 |
# all commands have been checked, so we're safe to combine them
|
1814 |
command = '|'.join([expcmd, comprcmd, utils.ShellQuoteArgs(remotecmd)])
|
1815 |
|
1816 |
result = utils.RunCmd(command, env=export_env) |
1817 |
|
1818 |
if result.failed:
|
1819 |
logging.error("os snapshot export command '%s' returned error: %s"
|
1820 |
" output: %s", command, result.fail_reason, result.output)
|
1821 |
return False |
1822 |
|
1823 |
return True |
1824 |
|
1825 |
|
1826 |
def FinalizeExport(instance, snap_disks): |
1827 |
"""Write out the export configuration information.
|
1828 |
|
1829 |
@type instance: L{objects.Instance}
|
1830 |
@param instance: the instance which we export, used for
|
1831 |
saving configuration
|
1832 |
@type snap_disks: list of L{objects.Disk}
|
1833 |
@param snap_disks: list of snapshot block devices, which
|
1834 |
will be used to get the actual name of the dump file
|
1835 |
|
1836 |
@rtype: boolean
|
1837 |
@return: the success of the operation
|
1838 |
|
1839 |
"""
|
1840 |
destdir = os.path.join(constants.EXPORT_DIR, instance.name + ".new")
|
1841 |
finaldestdir = os.path.join(constants.EXPORT_DIR, instance.name) |
1842 |
|
1843 |
config = objects.SerializableConfigParser() |
1844 |
|
1845 |
config.add_section(constants.INISECT_EXP) |
1846 |
config.set(constants.INISECT_EXP, 'version', '0') |
1847 |
config.set(constants.INISECT_EXP, 'timestamp', '%d' % int(time.time())) |
1848 |
config.set(constants.INISECT_EXP, 'source', instance.primary_node)
|
1849 |
config.set(constants.INISECT_EXP, 'os', instance.os)
|
1850 |
config.set(constants.INISECT_EXP, 'compression', 'gzip') |
1851 |
|
1852 |
config.add_section(constants.INISECT_INS) |
1853 |
config.set(constants.INISECT_INS, 'name', instance.name)
|
1854 |
config.set(constants.INISECT_INS, 'memory', '%d' % |
1855 |
instance.beparams[constants.BE_MEMORY]) |
1856 |
config.set(constants.INISECT_INS, 'vcpus', '%d' % |
1857 |
instance.beparams[constants.BE_VCPUS]) |
1858 |
config.set(constants.INISECT_INS, 'disk_template', instance.disk_template)
|
1859 |
|
1860 |
nic_total = 0
|
1861 |
for nic_count, nic in enumerate(instance.nics): |
1862 |
nic_total += 1
|
1863 |
config.set(constants.INISECT_INS, 'nic%d_mac' %
|
1864 |
nic_count, '%s' % nic.mac)
|
1865 |
config.set(constants.INISECT_INS, 'nic%d_ip' % nic_count, '%s' % nic.ip) |
1866 |
config.set(constants.INISECT_INS, 'nic%d_bridge' % nic_count,
|
1867 |
'%s' % nic.bridge)
|
1868 |
# TODO: redundant: on load can read nics until it doesn't exist
|
1869 |
config.set(constants.INISECT_INS, 'nic_count' , '%d' % nic_total) |
1870 |
|
1871 |
disk_total = 0
|
1872 |
for disk_count, disk in enumerate(snap_disks): |
1873 |
if disk:
|
1874 |
disk_total += 1
|
1875 |
config.set(constants.INISECT_INS, 'disk%d_ivname' % disk_count,
|
1876 |
('%s' % disk.iv_name))
|
1877 |
config.set(constants.INISECT_INS, 'disk%d_dump' % disk_count,
|
1878 |
('%s' % disk.physical_id[1])) |
1879 |
config.set(constants.INISECT_INS, 'disk%d_size' % disk_count,
|
1880 |
('%d' % disk.size))
|
1881 |
|
1882 |
config.set(constants.INISECT_INS, 'disk_count' , '%d' % disk_total) |
1883 |
|
1884 |
utils.WriteFile(os.path.join(destdir, constants.EXPORT_CONF_FILE), |
1885 |
data=config.Dumps()) |
1886 |
shutil.rmtree(finaldestdir, True)
|
1887 |
shutil.move(destdir, finaldestdir) |
1888 |
|
1889 |
return True |
1890 |
|
1891 |
|
1892 |
def ExportInfo(dest): |
1893 |
"""Get export configuration information.
|
1894 |
|
1895 |
@type dest: str
|
1896 |
@param dest: directory containing the export
|
1897 |
|
1898 |
@rtype: L{objects.SerializableConfigParser}
|
1899 |
@return: a serializable config file containing the
|
1900 |
export info
|
1901 |
|
1902 |
"""
|
1903 |
cff = os.path.join(dest, constants.EXPORT_CONF_FILE) |
1904 |
|
1905 |
config = objects.SerializableConfigParser() |
1906 |
config.read(cff) |
1907 |
|
1908 |
if (not config.has_section(constants.INISECT_EXP) or |
1909 |
not config.has_section(constants.INISECT_INS)):
|
1910 |
return None |
1911 |
|
1912 |
return config
|
1913 |
|
1914 |
|
1915 |
def ImportOSIntoInstance(instance, src_node, src_images, cluster_name): |
1916 |
"""Import an os image into an instance.
|
1917 |
|
1918 |
@type instance: L{objects.Instance}
|
1919 |
@param instance: instance to import the disks into
|
1920 |
@type src_node: string
|
1921 |
@param src_node: source node for the disk images
|
1922 |
@type src_images: list of string
|
1923 |
@param src_images: absolute paths of the disk images
|
1924 |
@rtype: list of boolean
|
1925 |
@return: each boolean represent the success of importing the n-th disk
|
1926 |
|
1927 |
"""
|
1928 |
import_env = OSEnvironment(instance) |
1929 |
inst_os = OSFromDisk(instance.os) |
1930 |
import_script = inst_os.import_script |
1931 |
|
1932 |
logfile = "%s/import-%s-%s-%s.log" % (constants.LOG_OS_DIR, instance.os,
|
1933 |
instance.name, int(time.time()))
|
1934 |
if not os.path.exists(constants.LOG_OS_DIR): |
1935 |
os.mkdir(constants.LOG_OS_DIR, 0750)
|
1936 |
|
1937 |
comprcmd = "gunzip"
|
1938 |
impcmd = utils.BuildShellCmd("(cd %s; %s >%s 2>&1)", inst_os.path,
|
1939 |
import_script, logfile) |
1940 |
|
1941 |
final_result = [] |
1942 |
for idx, image in enumerate(src_images): |
1943 |
if image:
|
1944 |
destcmd = utils.BuildShellCmd('cat %s', image)
|
1945 |
remotecmd = _GetSshRunner(cluster_name).BuildCmd(src_node, |
1946 |
constants.GANETI_RUNAS, |
1947 |
destcmd) |
1948 |
command = '|'.join([utils.ShellQuoteArgs(remotecmd), comprcmd, impcmd])
|
1949 |
import_env['IMPORT_DEVICE'] = import_env['DISK_%d_PATH' % idx] |
1950 |
import_env['IMPORT_INDEX'] = str(idx) |
1951 |
result = utils.RunCmd(command, env=import_env) |
1952 |
if result.failed:
|
1953 |
logging.error("Disk import command '%s' returned error: %s"
|
1954 |
" output: %s", command, result.fail_reason,
|
1955 |
result.output) |
1956 |
final_result.append(False)
|
1957 |
else:
|
1958 |
final_result.append(True)
|
1959 |
else:
|
1960 |
final_result.append(True)
|
1961 |
|
1962 |
return final_result
|
1963 |
|
1964 |
|
1965 |
def ListExports(): |
1966 |
"""Return a list of exports currently available on this machine.
|
1967 |
|
1968 |
@rtype: list
|
1969 |
@return: list of the exports
|
1970 |
|
1971 |
"""
|
1972 |
if os.path.isdir(constants.EXPORT_DIR):
|
1973 |
return utils.ListVisibleFiles(constants.EXPORT_DIR)
|
1974 |
else:
|
1975 |
return []
|
1976 |
|
1977 |
|
1978 |
def RemoveExport(export): |
1979 |
"""Remove an existing export from the node.
|
1980 |
|
1981 |
@type export: str
|
1982 |
@param export: the name of the export to remove
|
1983 |
@rtype: boolean
|
1984 |
@return: the success of the operation
|
1985 |
|
1986 |
"""
|
1987 |
target = os.path.join(constants.EXPORT_DIR, export) |
1988 |
|
1989 |
shutil.rmtree(target) |
1990 |
# TODO: catch some of the relevant exceptions and provide a pretty
|
1991 |
# error message if rmtree fails.
|
1992 |
|
1993 |
return True |
1994 |
|
1995 |
|
1996 |
def BlockdevRename(devlist): |
1997 |
"""Rename a list of block devices.
|
1998 |
|
1999 |
@type devlist: list of tuples
|
2000 |
@param devlist: list of tuples of the form (disk,
|
2001 |
new_logical_id, new_physical_id); disk is an
|
2002 |
L{objects.Disk} object describing the current disk,
|
2003 |
and new logical_id/physical_id is the name we
|
2004 |
rename it to
|
2005 |
@rtype: boolean
|
2006 |
@return: True if all renames succeeded, False otherwise
|
2007 |
|
2008 |
"""
|
2009 |
result = True
|
2010 |
for disk, unique_id in devlist: |
2011 |
dev = _RecursiveFindBD(disk) |
2012 |
if dev is None: |
2013 |
result = False
|
2014 |
continue
|
2015 |
try:
|
2016 |
old_rpath = dev.dev_path |
2017 |
dev.Rename(unique_id) |
2018 |
new_rpath = dev.dev_path |
2019 |
if old_rpath != new_rpath:
|
2020 |
DevCacheManager.RemoveCache(old_rpath) |
2021 |
# FIXME: we should add the new cache information here, like:
|
2022 |
# DevCacheManager.UpdateCache(new_rpath, owner, ...)
|
2023 |
# but we don't have the owner here - maybe parse from existing
|
2024 |
# cache? for now, we only lose lvm data when we rename, which
|
2025 |
# is less critical than DRBD or MD
|
2026 |
except errors.BlockDeviceError, err:
|
2027 |
logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
|
2028 |
result = False
|
2029 |
return result
|
2030 |
|
2031 |
|
2032 |
def _TransformFileStorageDir(file_storage_dir): |
2033 |
"""Checks whether given file_storage_dir is valid.
|
2034 |
|
2035 |
Checks wheter the given file_storage_dir is within the cluster-wide
|
2036 |
default file_storage_dir stored in SimpleStore. Only paths under that
|
2037 |
directory are allowed.
|
2038 |
|
2039 |
@type file_storage_dir: str
|
2040 |
@param file_storage_dir: the path to check
|
2041 |
|
2042 |
@return: the normalized path if valid, None otherwise
|
2043 |
|
2044 |
"""
|
2045 |
cfg = _GetConfig() |
2046 |
file_storage_dir = os.path.normpath(file_storage_dir) |
2047 |
base_file_storage_dir = cfg.GetFileStorageDir() |
2048 |
if (not os.path.commonprefix([file_storage_dir, base_file_storage_dir]) == |
2049 |
base_file_storage_dir): |
2050 |
logging.error("file storage directory '%s' is not under base file"
|
2051 |
" storage directory '%s'",
|
2052 |
file_storage_dir, base_file_storage_dir) |
2053 |
return None |
2054 |
return file_storage_dir
|
2055 |
|
2056 |
|
2057 |
def CreateFileStorageDir(file_storage_dir): |
2058 |
"""Create file storage directory.
|
2059 |
|
2060 |
@type file_storage_dir: str
|
2061 |
@param file_storage_dir: directory to create
|
2062 |
|
2063 |
@rtype: tuple
|
2064 |
@return: tuple with first element a boolean indicating wheter dir
|
2065 |
creation was successful or not
|
2066 |
|
2067 |
"""
|
2068 |
file_storage_dir = _TransformFileStorageDir(file_storage_dir) |
2069 |
result = True,
|
2070 |
if not file_storage_dir: |
2071 |
result = False,
|
2072 |
else:
|
2073 |
if os.path.exists(file_storage_dir):
|
2074 |
if not os.path.isdir(file_storage_dir): |
2075 |
logging.error("'%s' is not a directory", file_storage_dir)
|
2076 |
result = False,
|
2077 |
else:
|
2078 |
try:
|
2079 |
os.makedirs(file_storage_dir, 0750)
|
2080 |
except OSError, err: |
2081 |
logging.error("Cannot create file storage directory '%s': %s",
|
2082 |
file_storage_dir, err) |
2083 |
result = False,
|
2084 |
return result
|
2085 |
|
2086 |
|
2087 |
def RemoveFileStorageDir(file_storage_dir): |
2088 |
"""Remove file storage directory.
|
2089 |
|
2090 |
Remove it only if it's empty. If not log an error and return.
|
2091 |
|
2092 |
@type file_storage_dir: str
|
2093 |
@param file_storage_dir: the directory we should cleanup
|
2094 |
@rtype: tuple (success,)
|
2095 |
@return: tuple of one element, C{success}, denoting
|
2096 |
whether the operation was successfull
|
2097 |
|
2098 |
"""
|
2099 |
file_storage_dir = _TransformFileStorageDir(file_storage_dir) |
2100 |
result = True,
|
2101 |
if not file_storage_dir: |
2102 |
result = False,
|
2103 |
else:
|
2104 |
if os.path.exists(file_storage_dir):
|
2105 |
if not os.path.isdir(file_storage_dir): |
2106 |
logging.error("'%s' is not a directory", file_storage_dir)
|
2107 |
result = False,
|
2108 |
# deletes dir only if empty, otherwise we want to return False
|
2109 |
try:
|
2110 |
os.rmdir(file_storage_dir) |
2111 |
except OSError, err: |
2112 |
logging.exception("Cannot remove file storage directory '%s'",
|
2113 |
file_storage_dir) |
2114 |
result = False,
|
2115 |
return result
|
2116 |
|
2117 |
|
2118 |
def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir): |
2119 |
"""Rename the file storage directory.
|
2120 |
|
2121 |
@type old_file_storage_dir: str
|
2122 |
@param old_file_storage_dir: the current path
|
2123 |
@type new_file_storage_dir: str
|
2124 |
@param new_file_storage_dir: the name we should rename to
|
2125 |
@rtype: tuple (success,)
|
2126 |
@return: tuple of one element, C{success}, denoting
|
2127 |
whether the operation was successful
|
2128 |
|
2129 |
"""
|
2130 |
old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir) |
2131 |
new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir) |
2132 |
result = True,
|
2133 |
if not old_file_storage_dir or not new_file_storage_dir: |
2134 |
result = False,
|
2135 |
else:
|
2136 |
if not os.path.exists(new_file_storage_dir): |
2137 |
if os.path.isdir(old_file_storage_dir):
|
2138 |
try:
|
2139 |
os.rename(old_file_storage_dir, new_file_storage_dir) |
2140 |
except OSError, err: |
2141 |
logging.exception("Cannot rename '%s' to '%s'",
|
2142 |
old_file_storage_dir, new_file_storage_dir) |
2143 |
result = False,
|
2144 |
else:
|
2145 |
logging.error("'%s' is not a directory", old_file_storage_dir)
|
2146 |
result = False,
|
2147 |
else:
|
2148 |
if os.path.exists(old_file_storage_dir):
|
2149 |
logging.error("Cannot rename '%s' to '%s'. Both locations exist.",
|
2150 |
old_file_storage_dir, new_file_storage_dir) |
2151 |
result = False,
|
2152 |
return result
|
2153 |
|
2154 |
|
2155 |
def _IsJobQueueFile(file_name): |
2156 |
"""Checks whether the given filename is in the queue directory.
|
2157 |
|
2158 |
@type file_name: str
|
2159 |
@param file_name: the file name we should check
|
2160 |
@rtype: boolean
|
2161 |
@return: whether the file is under the queue directory
|
2162 |
|
2163 |
"""
|
2164 |
queue_dir = os.path.normpath(constants.QUEUE_DIR) |
2165 |
result = (os.path.commonprefix([queue_dir, file_name]) == queue_dir) |
2166 |
|
2167 |
if not result: |
2168 |
logging.error("'%s' is not a file in the queue directory",
|
2169 |
file_name) |
2170 |
|
2171 |
return result
|
2172 |
|
2173 |
|
2174 |
def JobQueueUpdate(file_name, content): |
2175 |
"""Updates a file in the queue directory.
|
2176 |
|
2177 |
This is just a wrapper over L{utils.WriteFile}, with proper
|
2178 |
checking.
|
2179 |
|
2180 |
@type file_name: str
|
2181 |
@param file_name: the job file name
|
2182 |
@type content: str
|
2183 |
@param content: the new job contents
|
2184 |
@rtype: boolean
|
2185 |
@return: the success of the operation
|
2186 |
|
2187 |
"""
|
2188 |
if not _IsJobQueueFile(file_name): |
2189 |
return False |
2190 |
|
2191 |
# Write and replace the file atomically
|
2192 |
utils.WriteFile(file_name, data=_Decompress(content)) |
2193 |
|
2194 |
return True |
2195 |
|
2196 |
|
2197 |
def JobQueueRename(old, new): |
2198 |
"""Renames a job queue file.
|
2199 |
|
2200 |
This is just a wrapper over os.rename with proper checking.
|
2201 |
|
2202 |
@type old: str
|
2203 |
@param old: the old (actual) file name
|
2204 |
@type new: str
|
2205 |
@param new: the desired file name
|
2206 |
@rtype: boolean
|
2207 |
@return: the success of the operation
|
2208 |
|
2209 |
"""
|
2210 |
if not (_IsJobQueueFile(old) and _IsJobQueueFile(new)): |
2211 |
return False |
2212 |
|
2213 |
utils.RenameFile(old, new, mkdir=True)
|
2214 |
|
2215 |
return True |
2216 |
|
2217 |
|
2218 |
def JobQueueSetDrainFlag(drain_flag): |
2219 |
"""Set the drain flag for the queue.
|
2220 |
|
2221 |
This will set or unset the queue drain flag.
|
2222 |
|
2223 |
@type drain_flag: boolean
|
2224 |
@param drain_flag: if True, will set the drain flag, otherwise reset it.
|
2225 |
@rtype: boolean
|
2226 |
@return: always True
|
2227 |
@warning: the function always returns True
|
2228 |
|
2229 |
"""
|
2230 |
if drain_flag:
|
2231 |
utils.WriteFile(constants.JOB_QUEUE_DRAIN_FILE, data="", close=True) |
2232 |
else:
|
2233 |
utils.RemoveFile(constants.JOB_QUEUE_DRAIN_FILE) |
2234 |
|
2235 |
return True |
2236 |
|
2237 |
|
2238 |
def BlockdevClose(instance_name, disks): |
2239 |
"""Closes the given block devices.
|
2240 |
|
2241 |
This means they will be switched to secondary mode (in case of
|
2242 |
DRBD).
|
2243 |
|
2244 |
@param instance_name: if the argument is not empty, the symlinks
|
2245 |
of this instance will be removed
|
2246 |
@type disks: list of L{objects.Disk}
|
2247 |
@param disks: the list of disks to be closed
|
2248 |
@rtype: tuple (success, message)
|
2249 |
@return: a tuple of success and message, where success
|
2250 |
indicates the succes of the operation, and message
|
2251 |
which will contain the error details in case we
|
2252 |
failed
|
2253 |
|
2254 |
"""
|
2255 |
bdevs = [] |
2256 |
for cf in disks: |
2257 |
rd = _RecursiveFindBD(cf) |
2258 |
if rd is None: |
2259 |
return (False, "Can't find device %s" % cf) |
2260 |
bdevs.append(rd) |
2261 |
|
2262 |
msg = [] |
2263 |
for rd in bdevs: |
2264 |
try:
|
2265 |
rd.Close() |
2266 |
except errors.BlockDeviceError, err:
|
2267 |
msg.append(str(err))
|
2268 |
if msg:
|
2269 |
return (False, "Can't make devices secondary: %s" % ",".join(msg)) |
2270 |
else:
|
2271 |
if instance_name:
|
2272 |
_RemoveBlockDevLinks(instance_name, disks) |
2273 |
return (True, "All devices secondary") |
2274 |
|
2275 |
|
2276 |
def ValidateHVParams(hvname, hvparams): |
2277 |
"""Validates the given hypervisor parameters.
|
2278 |
|
2279 |
@type hvname: string
|
2280 |
@param hvname: the hypervisor name
|
2281 |
@type hvparams: dict
|
2282 |
@param hvparams: the hypervisor parameters to be validated
|
2283 |
@rtype: tuple (success, message)
|
2284 |
@return: a tuple of success and message, where success
|
2285 |
indicates the succes of the operation, and message
|
2286 |
which will contain the error details in case we
|
2287 |
failed
|
2288 |
|
2289 |
"""
|
2290 |
try:
|
2291 |
hv_type = hypervisor.GetHypervisor(hvname) |
2292 |
hv_type.ValidateParameters(hvparams) |
2293 |
return (True, "Validation passed") |
2294 |
except errors.HypervisorError, err:
|
2295 |
return (False, str(err)) |
2296 |
|
2297 |
|
2298 |
def DemoteFromMC(): |
2299 |
"""Demotes the current node from master candidate role.
|
2300 |
|
2301 |
"""
|
2302 |
# try to ensure we're not the master by mistake
|
2303 |
master, myself = ssconf.GetMasterAndMyself() |
2304 |
if master == myself:
|
2305 |
return (False, "ssconf status shows I'm the master node, will not demote") |
2306 |
pid_file = utils.DaemonPidFileName(constants.MASTERD_PID) |
2307 |
if utils.IsProcessAlive(utils.ReadPidFile(pid_file)):
|
2308 |
return (False, "The master daemon is running, will not demote") |
2309 |
try:
|
2310 |
utils.CreateBackup(constants.CLUSTER_CONF_FILE) |
2311 |
except EnvironmentError, err: |
2312 |
if err.errno != errno.ENOENT:
|
2313 |
return (False, "Error while backing up cluster file: %s" % str(err)) |
2314 |
utils.RemoveFile(constants.CLUSTER_CONF_FILE) |
2315 |
return (True, "Done") |
2316 |
|
2317 |
|
2318 |
def _FindDisks(nodes_ip, disks): |
2319 |
"""Sets the physical ID on disks and returns the block devices.
|
2320 |
|
2321 |
"""
|
2322 |
# set the correct physical ID
|
2323 |
my_name = utils.HostInfo().name |
2324 |
for cf in disks: |
2325 |
cf.SetPhysicalID(my_name, nodes_ip) |
2326 |
|
2327 |
bdevs = [] |
2328 |
|
2329 |
for cf in disks: |
2330 |
rd = _RecursiveFindBD(cf) |
2331 |
if rd is None: |
2332 |
return (False, "Can't find device %s" % cf) |
2333 |
bdevs.append(rd) |
2334 |
return (True, bdevs) |
2335 |
|
2336 |
|
2337 |
def DrbdDisconnectNet(nodes_ip, disks): |
2338 |
"""Disconnects the network on a list of drbd devices.
|
2339 |
|
2340 |
"""
|
2341 |
status, bdevs = _FindDisks(nodes_ip, disks) |
2342 |
if not status: |
2343 |
return status, bdevs
|
2344 |
|
2345 |
# disconnect disks
|
2346 |
for rd in bdevs: |
2347 |
try:
|
2348 |
rd.DisconnectNet() |
2349 |
except errors.BlockDeviceError, err:
|
2350 |
logging.exception("Failed to go into standalone mode")
|
2351 |
return (False, "Can't change network configuration: %s" % str(err)) |
2352 |
return (True, "All disks are now disconnected") |
2353 |
|
2354 |
|
2355 |
def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster): |
2356 |
"""Attaches the network on a list of drbd devices.
|
2357 |
|
2358 |
"""
|
2359 |
status, bdevs = _FindDisks(nodes_ip, disks) |
2360 |
if not status: |
2361 |
return status, bdevs
|
2362 |
|
2363 |
if multimaster:
|
2364 |
for idx, rd in enumerate(bdevs): |
2365 |
try:
|
2366 |
_SymlinkBlockDev(instance_name, rd.dev_path, idx) |
2367 |
except EnvironmentError, err: |
2368 |
return (False, "Can't create symlink: %s" % str(err)) |
2369 |
# reconnect disks, switch to new master configuration and if
|
2370 |
# needed primary mode
|
2371 |
for rd in bdevs: |
2372 |
try:
|
2373 |
rd.AttachNet(multimaster) |
2374 |
except errors.BlockDeviceError, err:
|
2375 |
return (False, "Can't change network configuration: %s" % str(err)) |
2376 |
# wait until the disks are connected; we need to retry the re-attach
|
2377 |
# if the device becomes standalone, as this might happen if the one
|
2378 |
# node disconnects and reconnects in a different mode before the
|
2379 |
# other node reconnects; in this case, one or both of the nodes will
|
2380 |
# decide it has wrong configuration and switch to standalone
|
2381 |
RECONNECT_TIMEOUT = 2 * 60 |
2382 |
sleep_time = 0.100 # start with 100 miliseconds |
2383 |
timeout_limit = time.time() + RECONNECT_TIMEOUT |
2384 |
while time.time() < timeout_limit:
|
2385 |
all_connected = True
|
2386 |
for rd in bdevs: |
2387 |
stats = rd.GetProcStatus() |
2388 |
if not (stats.is_connected or stats.is_in_resync): |
2389 |
all_connected = False
|
2390 |
if stats.is_standalone:
|
2391 |
# peer had different config info and this node became
|
2392 |
# standalone, even though this should not happen with the
|
2393 |
# new staged way of changing disk configs
|
2394 |
try:
|
2395 |
rd.ReAttachNet(multimaster) |
2396 |
except errors.BlockDeviceError, err:
|
2397 |
return (False, "Can't change network configuration: %s" % str(err)) |
2398 |
if all_connected:
|
2399 |
break
|
2400 |
time.sleep(sleep_time) |
2401 |
sleep_time = min(5, sleep_time * 1.5) |
2402 |
if not all_connected: |
2403 |
return (False, "Timeout in disk reconnecting") |
2404 |
if multimaster:
|
2405 |
# change to primary mode
|
2406 |
for rd in bdevs: |
2407 |
try:
|
2408 |
rd.Open() |
2409 |
except errors.BlockDeviceError, err:
|
2410 |
return (False, "Can't change to primary mode: %s" % str(err)) |
2411 |
if multimaster:
|
2412 |
msg = "multi-master and primary"
|
2413 |
else:
|
2414 |
msg = "single-master"
|
2415 |
return (True, "Disks are now configured as %s" % msg) |
2416 |
|
2417 |
|
2418 |
def DrbdWaitSync(nodes_ip, disks): |
2419 |
"""Wait until DRBDs have synchronized.
|
2420 |
|
2421 |
"""
|
2422 |
status, bdevs = _FindDisks(nodes_ip, disks) |
2423 |
if not status: |
2424 |
return status, bdevs
|
2425 |
|
2426 |
min_resync = 100
|
2427 |
alldone = True
|
2428 |
failure = False
|
2429 |
for rd in bdevs: |
2430 |
stats = rd.GetProcStatus() |
2431 |
if not (stats.is_connected or stats.is_in_resync): |
2432 |
failure = True
|
2433 |
break
|
2434 |
alldone = alldone and (not stats.is_in_resync) |
2435 |
if stats.sync_percent is not None: |
2436 |
min_resync = min(min_resync, stats.sync_percent)
|
2437 |
return (not failure, (alldone, min_resync)) |
2438 |
|
2439 |
|
2440 |
class HooksRunner(object): |
2441 |
"""Hook runner.
|
2442 |
|
2443 |
This class is instantiated on the node side (ganeti-noded) and not
|
2444 |
on the master side.
|
2445 |
|
2446 |
"""
|
2447 |
RE_MASK = re.compile("^[a-zA-Z0-9_-]+$")
|
2448 |
|
2449 |
def __init__(self, hooks_base_dir=None): |
2450 |
"""Constructor for hooks runner.
|
2451 |
|
2452 |
@type hooks_base_dir: str or None
|
2453 |
@param hooks_base_dir: if not None, this overrides the
|
2454 |
L{constants.HOOKS_BASE_DIR} (useful for unittests)
|
2455 |
|
2456 |
"""
|
2457 |
if hooks_base_dir is None: |
2458 |
hooks_base_dir = constants.HOOKS_BASE_DIR |
2459 |
self._BASE_DIR = hooks_base_dir
|
2460 |
|
2461 |
@staticmethod
|
2462 |
def ExecHook(script, env): |
2463 |
"""Exec one hook script.
|
2464 |
|
2465 |
@type script: str
|
2466 |
@param script: the full path to the script
|
2467 |
@type env: dict
|
2468 |
@param env: the environment with which to exec the script
|
2469 |
@rtype: tuple (success, message)
|
2470 |
@return: a tuple of success and message, where success
|
2471 |
indicates the succes of the operation, and message
|
2472 |
which will contain the error details in case we
|
2473 |
failed
|
2474 |
|
2475 |
"""
|
2476 |
# exec the process using subprocess and log the output
|
2477 |
fdstdin = None
|
2478 |
try:
|
2479 |
fdstdin = open("/dev/null", "r") |
2480 |
child = subprocess.Popen([script], stdin=fdstdin, stdout=subprocess.PIPE, |
2481 |
stderr=subprocess.STDOUT, close_fds=True,
|
2482 |
shell=False, cwd="/", env=env) |
2483 |
output = ""
|
2484 |
try:
|
2485 |
output = child.stdout.read(4096)
|
2486 |
child.stdout.close() |
2487 |
except EnvironmentError, err: |
2488 |
output += "Hook script error: %s" % str(err) |
2489 |
|
2490 |
while True: |
2491 |
try:
|
2492 |
result = child.wait() |
2493 |
break
|
2494 |
except EnvironmentError, err: |
2495 |
if err.errno == errno.EINTR:
|
2496 |
continue
|
2497 |
raise
|
2498 |
finally:
|
2499 |
# try not to leak fds
|
2500 |
for fd in (fdstdin, ): |
2501 |
if fd is not None: |
2502 |
try:
|
2503 |
fd.close() |
2504 |
except EnvironmentError, err: |
2505 |
# just log the error
|
2506 |
#logging.exception("Error while closing fd %s", fd)
|
2507 |
pass
|
2508 |
|
2509 |
return result == 0, utils.SafeEncode(output.strip()) |
2510 |
|
2511 |
def RunHooks(self, hpath, phase, env): |
2512 |
"""Run the scripts in the hooks directory.
|
2513 |
|
2514 |
@type hpath: str
|
2515 |
@param hpath: the path to the hooks directory which
|
2516 |
holds the scripts
|
2517 |
@type phase: str
|
2518 |
@param phase: either L{constants.HOOKS_PHASE_PRE} or
|
2519 |
L{constants.HOOKS_PHASE_POST}
|
2520 |
@type env: dict
|
2521 |
@param env: dictionary with the environment for the hook
|
2522 |
@rtype: list
|
2523 |
@return: list of 3-element tuples:
|
2524 |
- script path
|
2525 |
- script result, either L{constants.HKR_SUCCESS} or
|
2526 |
L{constants.HKR_FAIL}
|
2527 |
- output of the script
|
2528 |
|
2529 |
@raise errors.ProgrammerError: for invalid input
|
2530 |
parameters
|
2531 |
|
2532 |
"""
|
2533 |
if phase == constants.HOOKS_PHASE_PRE:
|
2534 |
suffix = "pre"
|
2535 |
elif phase == constants.HOOKS_PHASE_POST:
|
2536 |
suffix = "post"
|
2537 |
else:
|
2538 |
raise errors.ProgrammerError("Unknown hooks phase: '%s'" % phase) |
2539 |
rr = [] |
2540 |
|
2541 |
subdir = "%s-%s.d" % (hpath, suffix)
|
2542 |
dir_name = "%s/%s" % (self._BASE_DIR, subdir) |
2543 |
try:
|
2544 |
dir_contents = utils.ListVisibleFiles(dir_name) |
2545 |
except OSError, err: |
2546 |
# FIXME: must log output in case of failures
|
2547 |
return rr
|
2548 |
|
2549 |
# we use the standard python sort order,
|
2550 |
# so 00name is the recommended naming scheme
|
2551 |
dir_contents.sort() |
2552 |
for relname in dir_contents: |
2553 |
fname = os.path.join(dir_name, relname) |
2554 |
if not (os.path.isfile(fname) and os.access(fname, os.X_OK) and |
2555 |
self.RE_MASK.match(relname) is not None): |
2556 |
rrval = constants.HKR_SKIP |
2557 |
output = ""
|
2558 |
else:
|
2559 |
result, output = self.ExecHook(fname, env)
|
2560 |
if not result: |
2561 |
rrval = constants.HKR_FAIL |
2562 |
else:
|
2563 |
rrval = constants.HKR_SUCCESS |
2564 |
rr.append(("%s/%s" % (subdir, relname), rrval, output))
|
2565 |
|
2566 |
return rr
|
2567 |
|
2568 |
|
2569 |
class IAllocatorRunner(object): |
2570 |
"""IAllocator runner.
|
2571 |
|
2572 |
This class is instantiated on the node side (ganeti-noded) and not on
|
2573 |
the master side.
|
2574 |
|
2575 |
"""
|
2576 |
def Run(self, name, idata): |
2577 |
"""Run an iallocator script.
|
2578 |
|
2579 |
@type name: str
|
2580 |
@param name: the iallocator script name
|
2581 |
@type idata: str
|
2582 |
@param idata: the allocator input data
|
2583 |
|
2584 |
@rtype: tuple
|
2585 |
@return: four element tuple of:
|
2586 |
- run status (one of the IARUN_ constants)
|
2587 |
- stdout
|
2588 |
- stderr
|
2589 |
- fail reason (as from L{utils.RunResult})
|
2590 |
|
2591 |
"""
|
2592 |
alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH, |
2593 |
os.path.isfile) |
2594 |
if alloc_script is None: |
2595 |
return (constants.IARUN_NOTFOUND, None, None, None) |
2596 |
|
2597 |
fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
|
2598 |
try:
|
2599 |
os.write(fd, idata) |
2600 |
os.close(fd) |
2601 |
result = utils.RunCmd([alloc_script, fin_name]) |
2602 |
if result.failed:
|
2603 |
return (constants.IARUN_FAILURE, result.stdout, result.stderr,
|
2604 |
result.fail_reason) |
2605 |
finally:
|
2606 |
os.unlink(fin_name) |
2607 |
|
2608 |
return (constants.IARUN_SUCCESS, result.stdout, result.stderr, None) |
2609 |
|
2610 |
|
2611 |
class DevCacheManager(object): |
2612 |
"""Simple class for managing a cache of block device information.
|
2613 |
|
2614 |
"""
|
2615 |
_DEV_PREFIX = "/dev/"
|
2616 |
_ROOT_DIR = constants.BDEV_CACHE_DIR |
2617 |
|
2618 |
@classmethod
|
2619 |
def _ConvertPath(cls, dev_path): |
2620 |
"""Converts a /dev/name path to the cache file name.
|
2621 |
|
2622 |
This replaces slashes with underscores and strips the /dev
|
2623 |
prefix. It then returns the full path to the cache file.
|
2624 |
|
2625 |
@type dev_path: str
|
2626 |
@param dev_path: the C{/dev/} path name
|
2627 |
@rtype: str
|
2628 |
@return: the converted path name
|
2629 |
|
2630 |
"""
|
2631 |
if dev_path.startswith(cls._DEV_PREFIX):
|
2632 |
dev_path = dev_path[len(cls._DEV_PREFIX):]
|
2633 |
dev_path = dev_path.replace("/", "_") |
2634 |
fpath = "%s/bdev_%s" % (cls._ROOT_DIR, dev_path)
|
2635 |
return fpath
|
2636 |
|
2637 |
@classmethod
|
2638 |
def UpdateCache(cls, dev_path, owner, on_primary, iv_name): |
2639 |
"""Updates the cache information for a given device.
|
2640 |
|
2641 |
@type dev_path: str
|
2642 |
@param dev_path: the pathname of the device
|
2643 |
@type owner: str
|
2644 |
@param owner: the owner (instance name) of the device
|
2645 |
@type on_primary: bool
|
2646 |
@param on_primary: whether this is the primary
|
2647 |
node nor not
|
2648 |
@type iv_name: str
|
2649 |
@param iv_name: the instance-visible name of the
|
2650 |
device, as in objects.Disk.iv_name
|
2651 |
|
2652 |
@rtype: None
|
2653 |
|
2654 |
"""
|
2655 |
if dev_path is None: |
2656 |
logging.error("DevCacheManager.UpdateCache got a None dev_path")
|
2657 |
return
|
2658 |
fpath = cls._ConvertPath(dev_path) |
2659 |
if on_primary:
|
2660 |
state = "primary"
|
2661 |
else:
|
2662 |
state = "secondary"
|
2663 |
if iv_name is None: |
2664 |
iv_name = "not_visible"
|
2665 |
fdata = "%s %s %s\n" % (str(owner), state, iv_name) |
2666 |
try:
|
2667 |
utils.WriteFile(fpath, data=fdata) |
2668 |
except EnvironmentError, err: |
2669 |
logging.exception("Can't update bdev cache for %s", dev_path)
|
2670 |
|
2671 |
@classmethod
|
2672 |
def RemoveCache(cls, dev_path): |
2673 |
"""Remove data for a dev_path.
|
2674 |
|
2675 |
This is just a wrapper over L{utils.RemoveFile} with a converted
|
2676 |
path name and logging.
|
2677 |
|
2678 |
@type dev_path: str
|
2679 |
@param dev_path: the pathname of the device
|
2680 |
|
2681 |
@rtype: None
|
2682 |
|
2683 |
"""
|
2684 |
if dev_path is None: |
2685 |
logging.error("DevCacheManager.RemoveCache got a None dev_path")
|
2686 |
return
|
2687 |
fpath = cls._ConvertPath(dev_path) |
2688 |
try:
|
2689 |
utils.RemoveFile(fpath) |
2690 |
except EnvironmentError, err: |
2691 |
logging.exception("Can't update bdev cache for %s", dev_path)
|