root / tools / burnin @ d7b47a77
History | View | Annotate | Download (13.5 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 |
"""Burnin program |
23 |
|
24 |
""" |
25 |
|
26 |
import os |
27 |
import sys |
28 |
import optparse |
29 |
from itertools import izip, islice, cycle |
30 |
from cStringIO import StringIO |
31 |
|
32 |
from ganeti import opcodes |
33 |
from ganeti import mcpu |
34 |
from ganeti import constants |
35 |
from ganeti import cli |
36 |
from ganeti import logger |
37 |
from ganeti import errors |
38 |
from ganeti import utils |
39 |
|
40 |
|
41 |
USAGE = ("\tburnin -o OS_NAME [options...] instance_name ...") |
42 |
|
43 |
|
44 |
def Usage(): |
45 |
"""Shows program usage information and exits the program.""" |
46 |
|
47 |
print >> sys.stderr, "Usage:" |
48 |
print >> sys.stderr, USAGE |
49 |
sys.exit(2) |
50 |
|
51 |
|
52 |
def Log(msg): |
53 |
"""Simple function that prints out its argument. |
54 |
|
55 |
""" |
56 |
print msg |
57 |
sys.stdout.flush() |
58 |
|
59 |
|
60 |
class Burner(object): |
61 |
"""Burner class.""" |
62 |
|
63 |
def __init__(self): |
64 |
"""Constructor.""" |
65 |
logger.SetupLogging(debug=False, program="ganeti/burnin") |
66 |
self._feed_buf = StringIO() |
67 |
self.proc = mcpu.Processor(feedback=self.Feedback) |
68 |
self.nodes = [] |
69 |
self.instances = [] |
70 |
self.to_rem = [] |
71 |
self.opts = None |
72 |
self.ParseOptions() |
73 |
self.GetState() |
74 |
|
75 |
def ClearFeedbackBuf(self): |
76 |
"""Clear the feedback buffer.""" |
77 |
self._feed_buf.truncate(0) |
78 |
|
79 |
def GetFeedbackBuf(self): |
80 |
"""Return the contents of the buffer.""" |
81 |
return self._feed_buf.getvalue() |
82 |
|
83 |
def Feedback(self, msg): |
84 |
"""Acumulate feedback in our buffer.""" |
85 |
self._feed_buf.write(msg) |
86 |
self._feed_buf.write("\n") |
87 |
if self.opts.verbose: |
88 |
Log(msg) |
89 |
|
90 |
def ExecOp(self, op): |
91 |
"""Execute an opcode and manage the exec buffer.""" |
92 |
self.ClearFeedbackBuf() |
93 |
return self.proc.ExecOpCode(op) |
94 |
|
95 |
def ParseOptions(self): |
96 |
"""Parses the command line options. |
97 |
|
98 |
In case of command line errors, it will show the usage and exit the |
99 |
program. |
100 |
|
101 |
""" |
102 |
|
103 |
parser = optparse.OptionParser(usage="\n%s" % USAGE, |
104 |
version="%%prog (ganeti) %s" % |
105 |
constants.RELEASE_VERSION, |
106 |
option_class=cli.CliOption) |
107 |
|
108 |
parser.add_option("-o", "--os", dest="os", default=None, |
109 |
help="OS to use during burnin", |
110 |
metavar="<OS>") |
111 |
parser.add_option("--os-size", dest="os_size", help="Disk size", |
112 |
default=4 * 1024, type="unit", metavar="<size>") |
113 |
parser.add_option("--swap-size", dest="swap_size", help="Swap size", |
114 |
default=4 * 1024, type="unit", metavar="<size>") |
115 |
parser.add_option("-v", "--verbose", |
116 |
action="store_true", dest="verbose", default=False, |
117 |
help="print command execution messages to stdout") |
118 |
parser.add_option("--no-replace1", dest="do_replace1", |
119 |
help="Skip disk replacement with the same secondary", |
120 |
action="store_false", default=True) |
121 |
parser.add_option("--no-replace2", dest="do_replace2", |
122 |
help="Skip disk replacement with a different secondary", |
123 |
action="store_false", default=True) |
124 |
parser.add_option("--no-failover", dest="do_failover", |
125 |
help="Skip instance failovers", action="store_false", |
126 |
default=True) |
127 |
parser.add_option("--no-importexport", dest="do_importexport", |
128 |
help="Skip instance export/import", action="store_false", |
129 |
default=True) |
130 |
parser.add_option("--no-startstop", dest="do_startstop", |
131 |
help="Skip instance stop/start", action="store_false", |
132 |
default=True) |
133 |
parser.add_option("-t", "--disk-template", dest="disk_template", |
134 |
choices=("diskless", "plain", "remote_raid1", "drbd"), |
135 |
default="remote_raid1", |
136 |
help="Template type for network mirroring (remote_raid1" |
137 |
" or drbd) [remote_raid1]") |
138 |
parser.add_option("-n", "--nodes", dest="nodes", default="", |
139 |
help="Comma separated list of nodes to perform" |
140 |
" the burnin on (defaults to all nodes)") |
141 |
|
142 |
options, args = parser.parse_args() |
143 |
if len(args) < 1 or options.os is None: |
144 |
Usage() |
145 |
|
146 |
supported_disk_templates = (constants.DT_DISKLESS, constants.DT_PLAIN, |
147 |
constants.DT_REMOTE_RAID1, |
148 |
constants.DT_DRBD8) |
149 |
if options.disk_template not in supported_disk_templates: |
150 |
Log("Unknown disk template '%s'" % options.disk_template) |
151 |
sys.exit(1) |
152 |
|
153 |
self.opts = options |
154 |
self.instances = args |
155 |
|
156 |
def GetState(self): |
157 |
"""Read the cluster state from the config.""" |
158 |
if self.opts.nodes: |
159 |
names = self.opts.nodes.split(",") |
160 |
else: |
161 |
names = [] |
162 |
try: |
163 |
op = opcodes.OpQueryNodes(output_fields=["name"], names=names) |
164 |
result = self.ExecOp(op) |
165 |
except errors.GenericError, err: |
166 |
err_code, msg = cli.FormatError(err) |
167 |
Log(msg) |
168 |
sys.exit(err_code) |
169 |
self.nodes = [data[0] for data in result] |
170 |
|
171 |
result = self.ExecOp(opcodes.OpDiagnoseOS()) |
172 |
|
173 |
if not result: |
174 |
Log("Can't get the OS list") |
175 |
sys.exit(1) |
176 |
|
177 |
# filter non-valid OS-es |
178 |
oses = {} |
179 |
for node_name in result: |
180 |
oses[node_name] = [obj for obj in result[node_name] if obj] |
181 |
|
182 |
fnode = oses.keys()[0] |
183 |
os_set = set([os_inst.name for os_inst in oses[fnode]]) |
184 |
del oses[fnode] |
185 |
for node in oses: |
186 |
os_set &= set([os_inst.name for os_inst in oses[node]]) |
187 |
|
188 |
if self.opts.os not in os_set: |
189 |
Log("OS '%s' not found" % self.opts.os) |
190 |
sys.exit(1) |
191 |
|
192 |
def CreateInstances(self): |
193 |
"""Create the given instances. |
194 |
|
195 |
""" |
196 |
self.to_rem = [] |
197 |
mytor = izip(cycle(self.nodes), |
198 |
islice(cycle(self.nodes), 1, None), |
199 |
self.instances) |
200 |
for pnode, snode, instance in mytor: |
201 |
op = opcodes.OpCreateInstance(instance_name=instance, |
202 |
mem_size=128, |
203 |
disk_size=self.opts.os_size, |
204 |
swap_size=self.opts.swap_size, |
205 |
disk_template=self.opts.disk_template, |
206 |
mode=constants.INSTANCE_CREATE, |
207 |
os_type=self.opts.os, |
208 |
pnode=pnode, |
209 |
snode=snode, |
210 |
vcpus=1, |
211 |
start=True, |
212 |
ip_check=True, |
213 |
wait_for_sync=True, |
214 |
mac="auto", |
215 |
kernel_path=None, |
216 |
initrd_path=None, |
217 |
hvm_boot_order=None) |
218 |
Log("- Add instance %s on nodes %s/%s" % (instance, pnode, snode)) |
219 |
self.ExecOp(op) |
220 |
self.to_rem.append(instance) |
221 |
|
222 |
def ReplaceDisks1R1(self): |
223 |
"""Replace disks with the same secondary for rr1.""" |
224 |
# replace all, both disks |
225 |
for instance in self.instances: |
226 |
op = opcodes.OpReplaceDisks(instance_name=instance, |
227 |
remote_node=None, |
228 |
mode=constants.REPLACE_DISK_ALL, |
229 |
disks=["sda", "sdb"]) |
230 |
|
231 |
Log("- Replace disks for instance %s" % (instance)) |
232 |
self.ExecOp(op) |
233 |
|
234 |
def ReplaceDisks1D8(self): |
235 |
"""Replace disks on primary and secondary for drbd8.""" |
236 |
for instance in self.instances: |
237 |
for mode in constants.REPLACE_DISK_SEC, constants.REPLACE_DISK_PRI: |
238 |
op = opcodes.OpReplaceDisks(instance_name=instance, |
239 |
mode=mode, |
240 |
disks=["sda", "sdb"]) |
241 |
Log("- Replace disks (%s) for instance %s" % (mode, instance)) |
242 |
self.ExecOp(op) |
243 |
|
244 |
def ReplaceDisks2(self): |
245 |
"""Replace secondary node.""" |
246 |
if self.opts.disk_template == constants.DT_REMOTE_RAID1: |
247 |
mode = constants.REPLACE_DISK_ALL |
248 |
else: |
249 |
mode = constants.REPLACE_DISK_SEC |
250 |
|
251 |
mytor = izip(islice(cycle(self.nodes), 2, None), |
252 |
self.instances) |
253 |
for tnode, instance in mytor: |
254 |
op = opcodes.OpReplaceDisks(instance_name=instance, |
255 |
mode=mode, |
256 |
remote_node=tnode, |
257 |
disks=["sda", "sdb"]) |
258 |
Log("- Replace secondary (%s) for instance %s" % (mode, instance)) |
259 |
self.ExecOp(op) |
260 |
|
261 |
def Failover(self): |
262 |
"""Failover the instances.""" |
263 |
|
264 |
for instance in self.instances: |
265 |
op = opcodes.OpFailoverInstance(instance_name=instance, |
266 |
ignore_consistency=False) |
267 |
|
268 |
Log("- Failover instance %s" % (instance)) |
269 |
self.ExecOp(op) |
270 |
|
271 |
def ImportExport(self): |
272 |
"""Export the instance, delete it, and import it back. |
273 |
|
274 |
""" |
275 |
|
276 |
mytor = izip(cycle(self.nodes), |
277 |
islice(cycle(self.nodes), 1, None), |
278 |
islice(cycle(self.nodes), 2, None), |
279 |
self.instances) |
280 |
|
281 |
for pnode, snode, enode, instance in mytor: |
282 |
exp_op = opcodes.OpExportInstance(instance_name=instance, |
283 |
target_node=enode, |
284 |
shutdown=True) |
285 |
rem_op = opcodes.OpRemoveInstance(instance_name=instance) |
286 |
nam_op = opcodes.OpQueryInstances(output_fields=["name"], |
287 |
names=[instance]) |
288 |
full_name = self.ExecOp(nam_op)[0][0] |
289 |
imp_dir = os.path.join(constants.EXPORT_DIR, full_name) |
290 |
imp_op = opcodes.OpCreateInstance(instance_name=instance, |
291 |
mem_size=128, |
292 |
disk_size=self.opts.os_size, |
293 |
swap_size=self.opts.swap_size, |
294 |
disk_template=self.opts.disk_template, |
295 |
mode=constants.INSTANCE_IMPORT, |
296 |
src_node=enode, |
297 |
src_path=imp_dir, |
298 |
pnode=pnode, |
299 |
snode=snode, |
300 |
vcpus=1, |
301 |
start=True, |
302 |
ip_check=True, |
303 |
wait_for_sync=True, |
304 |
mac="auto") |
305 |
|
306 |
Log("- Export instance %s to node %s" % (instance, enode)) |
307 |
self.ExecOp(exp_op) |
308 |
Log("- Remove instance %s" % (instance)) |
309 |
self.ExecOp(rem_op) |
310 |
self.to_rem.remove(instance) |
311 |
Log("- Import instance %s from node %s to node %s" % |
312 |
(instance, enode, pnode)) |
313 |
self.ExecOp(imp_op) |
314 |
self.to_rem.append(instance) |
315 |
|
316 |
def StopStart(self): |
317 |
"""Stop/start the instances.""" |
318 |
for instance in self.instances: |
319 |
op = opcodes.OpShutdownInstance(instance_name=instance) |
320 |
Log("- Shutdown instance %s" % instance) |
321 |
self.ExecOp(op) |
322 |
op = opcodes.OpStartupInstance(instance_name=instance, force=False) |
323 |
Log("- Start instance %s" % instance) |
324 |
self.ExecOp(op) |
325 |
|
326 |
def Remove(self): |
327 |
"""Remove the instances.""" |
328 |
for instance in self.to_rem: |
329 |
op = opcodes.OpRemoveInstance(instance_name=instance) |
330 |
Log("- Remove instance %s" % instance) |
331 |
self.ExecOp(op) |
332 |
|
333 |
def BurninCluster(self): |
334 |
"""Test a cluster intensively. |
335 |
|
336 |
This will create instances and then start/stop/failover them. |
337 |
It is safe for existing instances but could impact performance. |
338 |
|
339 |
""" |
340 |
|
341 |
opts = self.opts |
342 |
|
343 |
Log("- Testing global parameters") |
344 |
|
345 |
if (len(self.nodes) == 1 and |
346 |
opts.disk_template not in (constants.DT_DISKLESS, constants.DT_PLAIN)): |
347 |
Log("When one node is available/selected the disk template must" |
348 |
" be 'plain' or 'diskless'") |
349 |
sys.exit(1) |
350 |
|
351 |
has_err = True |
352 |
try: |
353 |
self.CreateInstances() |
354 |
if opts.do_replace1 and opts.disk_template in constants.DTS_NET_MIRROR: |
355 |
if opts.disk_template == constants.DT_REMOTE_RAID1: |
356 |
self.ReplaceDisks1R1() |
357 |
elif opts.disk_template == constants.DT_DRBD8: |
358 |
self.ReplaceDisks1D8() |
359 |
if (opts.do_replace2 and len(self.nodes) > 2 and |
360 |
opts.disk_template in constants.DTS_NET_MIRROR) : |
361 |
self.ReplaceDisks2() |
362 |
|
363 |
if opts.do_failover and opts.disk_template in constants.DTS_NET_MIRROR: |
364 |
self.Failover() |
365 |
|
366 |
if opts.do_importexport: |
367 |
self.ImportExport() |
368 |
|
369 |
if opts.do_startstop: |
370 |
self.StopStart() |
371 |
|
372 |
has_err = False |
373 |
finally: |
374 |
if has_err: |
375 |
Log("Error detected: opcode buffer follows:\n\n") |
376 |
Log(self.GetFeedbackBuf()) |
377 |
Log("\n\n") |
378 |
self.Remove() |
379 |
|
380 |
return 0 |
381 |
|
382 |
|
383 |
def main(): |
384 |
"""Main function""" |
385 |
|
386 |
burner = Burner() |
387 |
try: |
388 |
utils.Lock('cmd', max_retries=15, debug=True) |
389 |
except errors.LockError, err: |
390 |
logger.ToStderr(str(err)) |
391 |
return 1 |
392 |
try: |
393 |
retval = burner.BurninCluster() |
394 |
finally: |
395 |
utils.Unlock('cmd') |
396 |
utils.LockCleanup() |
397 |
return retval |
398 |
|
399 |
|
400 |
if __name__ == "__main__": |
401 |
main() |