[Buildbot] #3330: LatentBuildSlave not starting or properly queuing builds
Buildbot trac
trac at buildbot.net
Sun Aug 30 02:32:44 UTC 2015
#3330: LatentBuildSlave not starting or properly queuing builds
-------------------+------------------------
Reporter: devoid | Owner:
Type: defect | Status: new
Priority: major | Milestone: undecided
Version: 0.8.12 | Resolution:
Keywords: |
-------------------+------------------------
Comment (by dustin):
So from the master twistd.log, I see
{{{
2015-08-30 02:24:34+0000 [-] <Build builder>.startBuild
2015-08-30 02:24:34+0000 [-]
ShellCommand.startCommand(cmd=<RemoteShellCommand '['sleep', '10']'>)
2015-08-30 02:24:34+0000 [-] cmd.args = {'workdir': 'build', 'env':
None, 'want_stderr': 1, 'logEnviron': True, 'sigtermTime': None,
'logfiles': {}, 'timeout': 1200, 'want_stdout': 1, 'usePTY': 'slave-
config', 'maxTime': None, 'initial_stdin': None}
2015-08-30 02:24:34+0000 [-] <RemoteShellCommand '['sleep', '10']'>:
RemoteCommand.run [0]
2015-08-30 02:24:34+0000 [-] command '['sleep', '10']' in dir 'build'
2015-08-30 02:24:44+0000 [Broker,0,172.17.1.76] <RemoteShellCommand
'['sleep', '10']'> rc=0
2015-08-30 02:24:44+0000 [-] closing log <buildbot.status.logfile.LogFile
instance at 0x7ffa49b8d440>
2015-08-30 02:24:44+0000 [-]
releaseLocks(<buildbot.steps.shell.ShellCommand object at
0x7ffa4abb2250>): []
2015-08-30 02:24:44+0000 [-] step 'shell' complete: success
2015-08-30 02:24:44+0000 [-] <Build builder>: build finished
2015-08-30 02:24:44+0000 [-] setting expectations for next time
2015-08-30 02:24:44+0000 [-] new expectations: 10.0166060925 seconds
2015-08-30 02:24:44+0000 [-] Stopping container 7b8276...
2015-08-30 02:24:44+0000 [-] releaseLocks(<DockerLatentBuildSlave
'builder-01'>): []
2015-08-30 02:24:54+0000 [Broker,0,172.17.1.76]
BuildSlave.detached(builder-01)
2015-08-30 02:24:54+0000 [Broker,0,172.17.1.76]
releaseLocks(<DockerLatentBuildSlave 'builder-01'>): []
}}}
The stop occurs right after the step completes, since the
build_wait_timeout is set to 0.
What *should* happen when the slave disconnects, is another scan for
pending build requests. Since two remain, the slave should be started
back up immediately. This should happen when the locks are released.
--
Ticket URL: <http://trac.buildbot.net/ticket/3330#comment:5>
Buildbot <http://buildbot.net/>
Buildbot: build/test automation
More information about the bugs
mailing list