root / scripts / gnt-cluster @ 207a6c74
History | View | Annotate | Download (15.1 kB)
1 |
#!/usr/bin/python |
---|---|
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 |
import sys |
23 |
from optparse import make_option |
24 |
import pprint |
25 |
import os.path |
26 |
|
27 |
from ganeti.cli import * |
28 |
from ganeti import opcodes |
29 |
from ganeti import constants |
30 |
from ganeti import errors |
31 |
from ganeti import utils |
32 |
from ganeti import bootstrap |
33 |
from ganeti import ssh |
34 |
from ganeti import ssconf |
35 |
|
36 |
|
37 |
def InitCluster(opts, args): |
38 |
"""Initialize the cluster. |
39 |
|
40 |
Args: |
41 |
opts - class with options as members |
42 |
args - list of arguments, expected to be [clustername] |
43 |
|
44 |
""" |
45 |
if not opts.lvm_storage and opts.vg_name: |
46 |
print ("Options --no-lvm-storage and --vg-name conflict.") |
47 |
return 1 |
48 |
|
49 |
vg_name = opts.vg_name |
50 |
if opts.lvm_storage and not opts.vg_name: |
51 |
vg_name = constants.DEFAULT_VG |
52 |
|
53 |
bootstrap.InitCluster(cluster_name=args[0], |
54 |
secondary_ip=opts.secondary_ip, |
55 |
hypervisor_type=opts.hypervisor_type, |
56 |
vg_name=vg_name, |
57 |
mac_prefix=opts.mac_prefix, |
58 |
def_bridge=opts.def_bridge, |
59 |
master_netdev=opts.master_netdev, |
60 |
file_storage_dir=opts.file_storage_dir) |
61 |
return 0 |
62 |
|
63 |
|
64 |
def DestroyCluster(opts, args): |
65 |
"""Destroy the cluster. |
66 |
|
67 |
Args: |
68 |
opts - class with options as members |
69 |
|
70 |
""" |
71 |
if not opts.yes_do_it: |
72 |
print ("Destroying a cluster is irreversibly. If you really want destroy" |
73 |
" this cluster, supply the --yes-do-it option.") |
74 |
return 1 |
75 |
|
76 |
op = opcodes.OpDestroyCluster() |
77 |
master = SubmitOpCode(op) |
78 |
# if we reached this, the opcode didn't fail; we can proceed to |
79 |
# shutdown all the daemons |
80 |
bootstrap.FinalizeClusterDestroy(master) |
81 |
return 0 |
82 |
|
83 |
|
84 |
def RenameCluster(opts, args): |
85 |
"""Rename the cluster. |
86 |
|
87 |
Args: |
88 |
opts - class with options as members, we use force only |
89 |
args - list of arguments, expected to be [new_name] |
90 |
|
91 |
""" |
92 |
name = args[0] |
93 |
if not opts.force: |
94 |
usertext = ("This will rename the cluster to '%s'. If you are connected" |
95 |
" over the network to the cluster name, the operation is very" |
96 |
" dangerous as the IP address will be removed from the node" |
97 |
" and the change may not go through. Continue?") % name |
98 |
if not AskUser(usertext): |
99 |
return 1 |
100 |
|
101 |
op = opcodes.OpRenameCluster(name=name) |
102 |
SubmitOpCode(op) |
103 |
return 0 |
104 |
|
105 |
|
106 |
def ShowClusterVersion(opts, args): |
107 |
"""Write version of ganeti software to the standard output. |
108 |
|
109 |
Args: |
110 |
opts - class with options as members |
111 |
|
112 |
""" |
113 |
op = opcodes.OpQueryClusterInfo() |
114 |
result = SubmitOpCode(op) |
115 |
print ("Software version: %s" % result["software_version"]) |
116 |
print ("Internode protocol: %s" % result["protocol_version"]) |
117 |
print ("Configuration format: %s" % result["config_version"]) |
118 |
print ("OS api version: %s" % result["os_api_version"]) |
119 |
print ("Export interface: %s" % result["export_version"]) |
120 |
return 0 |
121 |
|
122 |
|
123 |
def ShowClusterMaster(opts, args): |
124 |
"""Write name of master node to the standard output. |
125 |
|
126 |
Args: |
127 |
opts - class with options as members |
128 |
|
129 |
""" |
130 |
sstore = ssconf.SimpleStore() |
131 |
print sstore.GetMasterNode() |
132 |
return 0 |
133 |
|
134 |
|
135 |
def ShowClusterConfig(opts, args): |
136 |
"""Shows cluster information. |
137 |
|
138 |
""" |
139 |
op = opcodes.OpQueryClusterInfo() |
140 |
result = SubmitOpCode(op) |
141 |
|
142 |
print ("Cluster name: %s" % result["name"]) |
143 |
|
144 |
print ("Master node: %s" % result["master"]) |
145 |
|
146 |
print ("Architecture (this node): %s (%s)" % |
147 |
(result["architecture"][0], result["architecture"][1])) |
148 |
|
149 |
print ("Cluster hypervisor: %s" % result["hypervisor_type"]) |
150 |
|
151 |
return 0 |
152 |
|
153 |
|
154 |
def ClusterCopyFile(opts, args): |
155 |
"""Copy a file from master to some nodes. |
156 |
|
157 |
Args: |
158 |
opts - class with options as members |
159 |
args - list containing a single element, the file name |
160 |
Opts used: |
161 |
nodes - list containing the name of target nodes; if empty, all nodes |
162 |
|
163 |
""" |
164 |
filename = args[0] |
165 |
if not os.path.exists(filename): |
166 |
raise errors.OpPrereqError("No such filename '%s'" % filename) |
167 |
|
168 |
myname = utils.HostInfo().name |
169 |
|
170 |
op = opcodes.OpQueryNodes(output_fields=["name"], names=opts.nodes) |
171 |
results = [row[0] for row in SubmitOpCode(op) if row[0] != myname] |
172 |
srun = ssh.SshRunner() |
173 |
for node in results: |
174 |
if not srun.CopyFileToNode(node, filename): |
175 |
print >> sys.stderr, ("Copy of file %s to node %s failed" % |
176 |
(filename, node)) |
177 |
|
178 |
return 0 |
179 |
|
180 |
|
181 |
def RunClusterCommand(opts, args): |
182 |
"""Run a command on some nodes. |
183 |
|
184 |
Args: |
185 |
opts - class with options as members |
186 |
args - the command list as a list |
187 |
Opts used: |
188 |
nodes: list containing the name of target nodes; if empty, all nodes |
189 |
|
190 |
""" |
191 |
command = " ".join(args) |
192 |
op = opcodes.OpQueryNodes(output_fields=["name"], names=opts.nodes) |
193 |
nodes = [row[0] for row in SubmitOpCode(op)] |
194 |
|
195 |
sstore = ssconf.SimpleStore() |
196 |
master_node = sstore.GetMasterNode() |
197 |
srun = ssh.SshRunner(sstore=sstore) |
198 |
|
199 |
if master_node in nodes: |
200 |
nodes.remove(master_node) |
201 |
nodes.append(master_node) |
202 |
|
203 |
for name in nodes: |
204 |
result = srun.Run(name, "root", command) |
205 |
print ("------------------------------------------------") |
206 |
print ("node: %s" % name) |
207 |
print ("%s" % result.output) |
208 |
print ("return code = %s" % result.exit_code) |
209 |
|
210 |
return 0 |
211 |
|
212 |
|
213 |
def VerifyCluster(opts, args): |
214 |
"""Verify integrity of cluster, performing various test on nodes. |
215 |
|
216 |
Args: |
217 |
opts - class with options as members |
218 |
|
219 |
""" |
220 |
skip_checks = [] |
221 |
if opts.skip_nplusone_mem: |
222 |
skip_checks.append(constants.VERIFY_NPLUSONE_MEM) |
223 |
op = opcodes.OpVerifyCluster(skip_checks=skip_checks) |
224 |
if SubmitOpCode(op): |
225 |
return 0 |
226 |
else: |
227 |
return 1 |
228 |
|
229 |
|
230 |
def VerifyDisks(opts, args): |
231 |
"""Verify integrity of cluster disks. |
232 |
|
233 |
Args: |
234 |
opts - class with options as members |
235 |
|
236 |
""" |
237 |
op = opcodes.OpVerifyDisks() |
238 |
result = SubmitOpCode(op) |
239 |
if not isinstance(result, (list, tuple)) or len(result) != 4: |
240 |
raise errors.ProgrammerError("Unknown result type for OpVerifyDisks") |
241 |
|
242 |
nodes, nlvm, instances, missing = result |
243 |
|
244 |
if nodes: |
245 |
print "Nodes unreachable or with bad data:" |
246 |
for name in nodes: |
247 |
print "\t%s" % name |
248 |
retcode = constants.EXIT_SUCCESS |
249 |
|
250 |
if nlvm: |
251 |
for node, text in nlvm.iteritems(): |
252 |
print ("Error on node %s: LVM error: %s" % |
253 |
(node, text[-400:].encode('string_escape'))) |
254 |
retcode |= 1 |
255 |
print "You need to fix these nodes first before fixing instances" |
256 |
|
257 |
if instances: |
258 |
for iname in instances: |
259 |
if iname in missing: |
260 |
continue |
261 |
op = opcodes.OpActivateInstanceDisks(instance_name=iname) |
262 |
try: |
263 |
print "Activating disks for instance '%s'" % iname |
264 |
SubmitOpCode(op) |
265 |
except errors.GenericError, err: |
266 |
nret, msg = FormatError(err) |
267 |
retcode |= nret |
268 |
print >> sys.stderr, ("Error activating disks for instance %s: %s" % |
269 |
(iname, msg)) |
270 |
|
271 |
if missing: |
272 |
for iname, ival in missing.iteritems(): |
273 |
all_missing = utils.all(ival, lambda x: x[0] in nlvm) |
274 |
if all_missing: |
275 |
print ("Instance %s cannot be verified as it lives on" |
276 |
" broken nodes" % iname) |
277 |
else: |
278 |
print "Instance %s has missing logical volumes:" % iname |
279 |
ival.sort() |
280 |
for node, vol in ival: |
281 |
if node in nlvm: |
282 |
print ("\tbroken node %s /dev/xenvg/%s" % (node, vol)) |
283 |
else: |
284 |
print ("\t%s /dev/xenvg/%s" % (node, vol)) |
285 |
print ("You need to run replace_disks for all the above" |
286 |
" instances, if this message persist after fixing nodes.") |
287 |
retcode |= 1 |
288 |
|
289 |
return retcode |
290 |
|
291 |
|
292 |
def MasterFailover(opts, args): |
293 |
"""Failover the master node. |
294 |
|
295 |
This command, when run on a non-master node, will cause the current |
296 |
master to cease being master, and the non-master to become new |
297 |
master. |
298 |
|
299 |
""" |
300 |
return bootstrap.MasterFailover() |
301 |
|
302 |
|
303 |
def SearchTags(opts, args): |
304 |
"""Searches the tags on all the cluster. |
305 |
|
306 |
""" |
307 |
op = opcodes.OpSearchTags(pattern=args[0]) |
308 |
result = SubmitOpCode(op) |
309 |
if not result: |
310 |
return 1 |
311 |
result = list(result) |
312 |
result.sort() |
313 |
for path, tag in result: |
314 |
print "%s %s" % (path, tag) |
315 |
|
316 |
|
317 |
def SetClusterParams(opts, args): |
318 |
"""Modify the cluster. |
319 |
|
320 |
Args: |
321 |
opts - class with options as members |
322 |
|
323 |
""" |
324 |
if not (not opts.lvm_storage or opts.vg_name): |
325 |
print "Please give at least one of the parameters." |
326 |
return 1 |
327 |
|
328 |
vg_name = opts.vg_name |
329 |
if not opts.lvm_storage and opts.vg_name: |
330 |
print ("Options --no-lvm-storage and --vg-name conflict.") |
331 |
return 1 |
332 |
|
333 |
op = opcodes.OpSetClusterParams(vg_name=opts.vg_name) |
334 |
SubmitOpCode(op) |
335 |
return 0 |
336 |
|
337 |
|
338 |
# this is an option common to more than one command, so we declare |
339 |
# it here and reuse it |
340 |
node_option = make_option("-n", "--node", action="append", dest="nodes", |
341 |
help="Node to copy to (if not given, all nodes)," |
342 |
" can be given multiple times", |
343 |
metavar="<node>", default=[]) |
344 |
|
345 |
commands = { |
346 |
'init': (InitCluster, ARGS_ONE, |
347 |
[DEBUG_OPT, |
348 |
make_option("-s", "--secondary-ip", dest="secondary_ip", |
349 |
help="Specify the secondary ip for this node;" |
350 |
" if given, the entire cluster must have secondary" |
351 |
" addresses", |
352 |
metavar="ADDRESS", default=None), |
353 |
make_option("-t", "--hypervisor-type", dest="hypervisor_type", |
354 |
help="Specify the hypervisor type " |
355 |
"(xen-3.0, kvm, fake, xen-hvm-3.1)", |
356 |
metavar="TYPE", choices=["xen-3.0", |
357 |
"kvm", |
358 |
"fake", |
359 |
"xen-hvm-3.1"], |
360 |
default="xen-3.0",), |
361 |
make_option("-m", "--mac-prefix", dest="mac_prefix", |
362 |
help="Specify the mac prefix for the instance IP" |
363 |
" addresses, in the format XX:XX:XX", |
364 |
metavar="PREFIX", |
365 |
default="aa:00:00",), |
366 |
make_option("-g", "--vg-name", dest="vg_name", |
367 |
help="Specify the volume group name " |
368 |
" (cluster-wide) for disk allocation [xenvg]", |
369 |
metavar="VG", |
370 |
default=None,), |
371 |
make_option("-b", "--bridge", dest="def_bridge", |
372 |
help="Specify the default bridge name (cluster-wide)" |
373 |
" to connect the instances to [%s]" % |
374 |
constants.DEFAULT_BRIDGE, |
375 |
metavar="BRIDGE", |
376 |
default=constants.DEFAULT_BRIDGE,), |
377 |
make_option("--master-netdev", dest="master_netdev", |
378 |
help="Specify the node interface (cluster-wide)" |
379 |
" on which the master IP address will be added " |
380 |
" [%s]" % constants.DEFAULT_BRIDGE, |
381 |
metavar="NETDEV", |
382 |
default=constants.DEFAULT_BRIDGE,), |
383 |
make_option("--file-storage-dir", dest="file_storage_dir", |
384 |
help="Specify the default directory (cluster-wide)" |
385 |
" for storing the file-based disks [%s]" % |
386 |
constants.DEFAULT_FILE_STORAGE_DIR, |
387 |
metavar="DIR", |
388 |
default=constants.DEFAULT_FILE_STORAGE_DIR,), |
389 |
make_option("--no-lvm-storage", dest="lvm_storage", |
390 |
help="No support for lvm based instances" |
391 |
" (cluster-wide)", |
392 |
action="store_false", default=True,), |
393 |
], |
394 |
"[opts...] <cluster_name>", |
395 |
"Initialises a new cluster configuration"), |
396 |
'destroy': (DestroyCluster, ARGS_NONE, |
397 |
[DEBUG_OPT, |
398 |
make_option("--yes-do-it", dest="yes_do_it", |
399 |
help="Destroy cluster", |
400 |
action="store_true"), |
401 |
], |
402 |
"", "Destroy cluster"), |
403 |
'rename': (RenameCluster, ARGS_ONE, [DEBUG_OPT, FORCE_OPT], |
404 |
"<new_name>", |
405 |
"Renames the cluster"), |
406 |
'verify': (VerifyCluster, ARGS_NONE, [DEBUG_OPT, |
407 |
make_option("--no-nplus1-mem", dest="skip_nplusone_mem", |
408 |
help="Skip N+1 memory redundancy tests", |
409 |
action="store_true", |
410 |
default=False,), |
411 |
], |
412 |
"", "Does a check on the cluster configuration"), |
413 |
'verify-disks': (VerifyDisks, ARGS_NONE, [DEBUG_OPT], |
414 |
"", "Does a check on the cluster disk status"), |
415 |
'masterfailover': (MasterFailover, ARGS_NONE, [DEBUG_OPT], |
416 |
"", "Makes the current node the master"), |
417 |
'version': (ShowClusterVersion, ARGS_NONE, [DEBUG_OPT], |
418 |
"", "Shows the cluster version"), |
419 |
'getmaster': (ShowClusterMaster, ARGS_NONE, [DEBUG_OPT], |
420 |
"", "Shows the cluster master"), |
421 |
'copyfile': (ClusterCopyFile, ARGS_ONE, [DEBUG_OPT, node_option], |
422 |
"[-n node...] <filename>", |
423 |
"Copies a file to all (or only some) nodes"), |
424 |
'command': (RunClusterCommand, ARGS_ATLEAST(1), [DEBUG_OPT, node_option], |
425 |
"[-n node...] <command>", |
426 |
"Runs a command on all (or only some) nodes"), |
427 |
'info': (ShowClusterConfig, ARGS_NONE, [DEBUG_OPT], |
428 |
"", "Show cluster configuration"), |
429 |
'list-tags': (ListTags, ARGS_NONE, |
430 |
[DEBUG_OPT], "", "List the tags of the cluster"), |
431 |
'add-tags': (AddTags, ARGS_ANY, [DEBUG_OPT, TAG_SRC_OPT], |
432 |
"tag...", "Add tags to the cluster"), |
433 |
'remove-tags': (RemoveTags, ARGS_ANY, [DEBUG_OPT, TAG_SRC_OPT], |
434 |
"tag...", "Remove tags from the cluster"), |
435 |
'search-tags': (SearchTags, ARGS_ONE, |
436 |
[DEBUG_OPT], "", "Searches the tags on all objects on" |
437 |
" the cluster for a given pattern (regex)"), |
438 |
'modify': (SetClusterParams, ARGS_NONE, |
439 |
[DEBUG_OPT, |
440 |
make_option("-g", "--vg-name", dest="vg_name", |
441 |
help="Specify the volume group name " |
442 |
" (cluster-wide) for disk allocation " |
443 |
"and enable lvm based storage", |
444 |
metavar="VG",), |
445 |
make_option("--no-lvm-storage", dest="lvm_storage", |
446 |
help="Disable support for lvm based instances" |
447 |
" (cluster-wide)", |
448 |
action="store_false", default=True,), |
449 |
], |
450 |
"[opts...]", |
451 |
"Alters the parameters of the cluster"), |
452 |
} |
453 |
|
454 |
if __name__ == '__main__': |
455 |
sys.exit(GenericMain(commands, override={"tag_type": constants.TAG_CLUSTER})) |