root / tools / burnin @ 4501af56
History | View | Annotate | Download (10.8 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 sys |
27 |
import optparse |
28 |
from itertools import izip, islice, cycle |
29 |
from cStringIO import StringIO |
30 |
|
31 |
from ganeti import opcodes |
32 |
from ganeti import mcpu |
33 |
from ganeti import constants |
34 |
from ganeti import cli |
35 |
from ganeti import logger |
36 |
from ganeti import errors |
37 |
from ganeti import utils |
38 |
|
39 |
|
40 |
USAGE = ("\tburnin -o OS_NAME [options...] instance_name ...") |
41 |
|
42 |
|
43 |
def Usage(): |
44 |
"""Shows program usage information and exits the program.""" |
45 |
|
46 |
print >> sys.stderr, "Usage:" |
47 |
print >> sys.stderr, USAGE |
48 |
sys.exit(2) |
49 |
|
50 |
|
51 |
def Log(msg): |
52 |
"""Simple function that prints out its argument. |
53 |
|
54 |
""" |
55 |
print msg |
56 |
sys.stdout.flush() |
57 |
|
58 |
|
59 |
class Burner(object): |
60 |
"""Burner class.""" |
61 |
|
62 |
def __init__(self): |
63 |
"""Constructor.""" |
64 |
logger.SetupLogging(debug=False, program="ganeti/burnin") |
65 |
self._feed_buf = StringIO() |
66 |
self.proc = mcpu.Processor(feedback=self.Feedback) |
67 |
self.nodes = [] |
68 |
self.instances = [] |
69 |
self.to_rem = [] |
70 |
self.opts = None |
71 |
self.ParseOptions() |
72 |
self.GetState() |
73 |
|
74 |
def ClearFeedbackBuf(self): |
75 |
"""Clear the feedback buffer.""" |
76 |
self._feed_buf.truncate(0) |
77 |
|
78 |
def GetFeedbackBuf(self): |
79 |
"""Return the contents of the buffer.""" |
80 |
return self._feed_buf.getvalue() |
81 |
|
82 |
def Feedback(self, msg): |
83 |
"""Acumulate feedback in our buffer.""" |
84 |
self._feed_buf.write(msg) |
85 |
self._feed_buf.write("\n") |
86 |
|
87 |
def ExecOp(self, op): |
88 |
"""Execute an opcode and manage the exec buffer.""" |
89 |
self.ClearFeedbackBuf() |
90 |
return self.proc.ExecOpCode(op) |
91 |
|
92 |
def ParseOptions(self): |
93 |
"""Parses the command line options. |
94 |
|
95 |
In case of command line errors, it will show the usage and exit the |
96 |
program. |
97 |
|
98 |
""" |
99 |
|
100 |
parser = optparse.OptionParser(usage="\n%s" % USAGE, |
101 |
version="%%prog (ganeti) %s" % |
102 |
constants.RELEASE_VERSION, |
103 |
option_class=cli.CliOption) |
104 |
|
105 |
parser.add_option("-o", "--os", dest="os", default=None, |
106 |
help="OS to use during burnin", |
107 |
metavar="<OS>") |
108 |
parser.add_option("--os-size", dest="os_size", help="Disk size", |
109 |
default=4 * 1024, type="unit", metavar="<size>") |
110 |
parser.add_option("--swap-size", dest="swap_size", help="Swap size", |
111 |
default=4 * 1024, type="unit", metavar="<size>") |
112 |
parser.add_option("-v", "--verbose", |
113 |
action="store_true", dest="verbose", default=False, |
114 |
help="print command execution messages to stdout") |
115 |
parser.add_option("--no-replace1", dest="do_replace1", |
116 |
help="Skip disk replacement with the same secondary", |
117 |
action="store_false", default=True) |
118 |
parser.add_option("--no-replace2", dest="do_replace2", |
119 |
help="Skip disk replacement with a different secondary", |
120 |
action="store_false", default=True) |
121 |
parser.add_option("--no-failover", dest="do_failover", |
122 |
help="Skip instance failovers", action="store_false", |
123 |
default=True) |
124 |
parser.add_option("-t", "--disk-template", dest="disk_template", |
125 |
choices=("remote_raid1", "drbd"), |
126 |
default="remote_raid1", |
127 |
help="Template type for network mirroring (remote_raid1" |
128 |
" or drbd) [remote_raid1]") |
129 |
parser.add_option("-n", "--nodes", dest="nodes", default="", |
130 |
help="Comma separated list of nodes to perform" |
131 |
" the burnin on (defaults to all nodes)") |
132 |
|
133 |
options, args = parser.parse_args() |
134 |
if len(args) < 1 or options.os is None: |
135 |
Usage() |
136 |
|
137 |
supported_disk_templates = (constants.DT_PLAIN, constants.DT_REMOTE_RAID1, |
138 |
constants.DT_DRBD8) |
139 |
if options.disk_template not in supported_disk_templates: |
140 |
Log("Unknown disk template '%s'" % options.disk_template) |
141 |
sys.exit(1) |
142 |
|
143 |
self.opts = options |
144 |
self.instances = args |
145 |
|
146 |
def GetState(self): |
147 |
"""Read the cluster state from the config.""" |
148 |
if self.opts.nodes: |
149 |
names = self.opts.nodes.split(",") |
150 |
else: |
151 |
names = [] |
152 |
try: |
153 |
op = opcodes.OpQueryNodes(output_fields=["name"], names=names) |
154 |
result = self.ExecOp(op) |
155 |
except errors.GenericError, err: |
156 |
err_code, msg = cli.FormatError(err) |
157 |
Log(msg) |
158 |
sys.exit(err_code) |
159 |
self.nodes = [data[0] for data in result] |
160 |
|
161 |
result = self.ExecOp(opcodes.OpDiagnoseOS()) |
162 |
|
163 |
if not result: |
164 |
Log("Can't get the OS list") |
165 |
sys.exit(1) |
166 |
|
167 |
# filter non-valid OS-es |
168 |
oses = {} |
169 |
for node_name in result: |
170 |
oses[node_name] = [obj for obj in result[node_name] if obj] |
171 |
|
172 |
fnode = oses.keys()[0] |
173 |
os_set = set([os_inst.name for os_inst in oses[fnode]]) |
174 |
del oses[fnode] |
175 |
for node in oses: |
176 |
os_set &= set([os_inst.name for os_inst in oses[node]]) |
177 |
|
178 |
if self.opts.os not in os_set: |
179 |
Log("OS '%s' not found" % self.opts.os) |
180 |
sys.exit(1) |
181 |
|
182 |
def CreateInstances(self): |
183 |
"""Create the given instances. |
184 |
|
185 |
""" |
186 |
self.to_rem = [] |
187 |
mytor = izip(cycle(self.nodes), |
188 |
islice(cycle(self.nodes), 1, None), |
189 |
self.instances) |
190 |
for pnode, snode, instance in mytor: |
191 |
op = opcodes.OpCreateInstance(instance_name=instance, |
192 |
mem_size=128, |
193 |
disk_size=self.opts.os_size, |
194 |
swap_size=self.opts.swap_size, |
195 |
disk_template=self.opts.disk_template, |
196 |
mode=constants.INSTANCE_CREATE, |
197 |
os_type=self.opts.os, |
198 |
pnode=pnode, |
199 |
snode=snode, |
200 |
vcpus=1, |
201 |
start=True, |
202 |
ip_check=True, |
203 |
wait_for_sync=True, |
204 |
mac="auto", |
205 |
kernel_path=None, |
206 |
initrd_path=None) |
207 |
Log("- Add instance %s on node %s" % (instance, pnode)) |
208 |
self.ExecOp(op) |
209 |
self.to_rem.append(instance) |
210 |
|
211 |
def ReplaceDisks1R1(self): |
212 |
"""Replace disks with the same secondary for rr1.""" |
213 |
# replace all, both disks |
214 |
for instance in self.instances: |
215 |
op = opcodes.OpReplaceDisks(instance_name=instance, |
216 |
remote_node=None, |
217 |
mode=constants.REPLACE_DISK_ALL, |
218 |
disks=["sda", "sdb"]) |
219 |
|
220 |
Log("- Replace disks for instance %s" % (instance)) |
221 |
self.ExecOp(op) |
222 |
|
223 |
def ReplaceDisks1D8(self): |
224 |
"""Replace disks on primary and secondary for drbd8.""" |
225 |
for instance in self.instances: |
226 |
for mode in constants.REPLACE_DISK_SEC, constants.REPLACE_DISK_PRI: |
227 |
op = opcodes.OpReplaceDisks(instance_name=instance, |
228 |
mode=mode, |
229 |
disks=["sda", "sdb"]) |
230 |
Log("- Replace disks (%s) for instance %s" % (mode, instance)) |
231 |
self.ExecOp(op) |
232 |
|
233 |
def ReplaceDisks2(self): |
234 |
"""Replace secondary node.""" |
235 |
if self.opts.disk_template == constants.DT_REMOTE_RAID1: |
236 |
mode = constants.REPLACE_DISK_ALL |
237 |
else: |
238 |
mode = constants.REPLACE_DISK_SEC |
239 |
|
240 |
mytor = izip(islice(cycle(self.nodes), 2, None), |
241 |
self.instances) |
242 |
for tnode, instance in mytor: |
243 |
op = opcodes.OpReplaceDisks(instance_name=instance, |
244 |
mode=mode, |
245 |
remote_node=tnode, |
246 |
disks=["sda", "sdb"]) |
247 |
Log("- Replace secondary (%s) for instance %s" % (mode, instance)) |
248 |
self.ExecOp(op) |
249 |
|
250 |
def Failover(self): |
251 |
"""Failover the instances.""" |
252 |
|
253 |
for instance in self.instances: |
254 |
op = opcodes.OpFailoverInstance(instance_name=instance, |
255 |
ignore_consistency=False) |
256 |
|
257 |
Log("- Failover instance %s" % (instance)) |
258 |
self.ExecOp(op) |
259 |
|
260 |
def StopStart(self): |
261 |
"""Stop/start the instances.""" |
262 |
for instance in self.instances: |
263 |
op = opcodes.OpShutdownInstance(instance_name=instance) |
264 |
Log("- Shutdown instance %s" % instance) |
265 |
self.ExecOp(op) |
266 |
op = opcodes.OpStartupInstance(instance_name=instance, force=False) |
267 |
Log("- Start instance %s" % instance) |
268 |
self.ExecOp(op) |
269 |
|
270 |
def Remove(self): |
271 |
"""Remove the instances.""" |
272 |
for instance in self.to_rem: |
273 |
op = opcodes.OpRemoveInstance(instance_name=instance) |
274 |
Log("- Remove instance %s" % instance) |
275 |
self.ExecOp(op) |
276 |
|
277 |
def BurninCluster(self): |
278 |
"""Test a cluster intensively. |
279 |
|
280 |
This will create instances and then start/stop/failover them. |
281 |
It is safe for existing instances but could impact performance. |
282 |
|
283 |
""" |
284 |
|
285 |
opts = self.opts |
286 |
|
287 |
Log("- Testing global parameters") |
288 |
|
289 |
if len(self.nodes) == 1 and opts.disk_template != constants.DT_PLAIN: |
290 |
Log("When one node is available/selected the disk template must" |
291 |
" be 'plain'") |
292 |
sys.exit(1) |
293 |
|
294 |
has_err = True |
295 |
try: |
296 |
self.CreateInstances() |
297 |
if opts.do_replace1 and opts.disk_template in constants.DTS_NET_MIRROR: |
298 |
if opts.disk_template == constants.DT_REMOTE_RAID1: |
299 |
self.ReplaceDisks1R1() |
300 |
elif opts.disk_template == constants.DT_DRBD8: |
301 |
self.ReplaceDisks1D8() |
302 |
if (opts.do_replace2 and len(self.nodes) > 2 and |
303 |
opts.disk_template in constants.DTS_NET_MIRROR) : |
304 |
self.ReplaceDisks2() |
305 |
|
306 |
if opts.do_failover and opts.disk_template in constants.DTS_NET_MIRROR: |
307 |
self.Failover() |
308 |
|
309 |
self.StopStart() |
310 |
has_err = False |
311 |
finally: |
312 |
if has_err: |
313 |
Log("Error detected: opcode buffer follows:\n\n") |
314 |
Log(self.GetFeedbackBuf()) |
315 |
Log("\n\n") |
316 |
self.Remove() |
317 |
|
318 |
return 0 |
319 |
|
320 |
|
321 |
def main(): |
322 |
"""Main function""" |
323 |
|
324 |
burner = Burner() |
325 |
try: |
326 |
utils.Lock('cmd', max_retries=15, debug=True) |
327 |
except errors.LockError, err: |
328 |
logger.ToStderr(str(err)) |
329 |
return 1 |
330 |
try: |
331 |
retval = burner.BurninCluster() |
332 |
finally: |
333 |
utils.Unlock('cmd') |
334 |
utils.LockCleanup() |
335 |
return retval |
336 |
|
337 |
|
338 |
if __name__ == "__main__": |
339 |
main() |