> On March 20, 2015, 3:52 p.m., Joe Smith wrote: > > > > Joe Smith wrote: > It seems like the `self.quitquitquit` is the important part (on line 340 > of the runner)- doesn't decreasing the timeout not give `quitquitquit` the > time it needs? > > Joe Smith wrote: > In `src/main/python/apache/aurora/executor/thermos_task_runner.py` > > ``` > 331 waited = Amount(0, Time.SECONDS) > > 332 while self.is_alive and waited < timeout: > > 333 self._clock.sleep(self.POLL_INTERVAL.as_(Time.SECONDS)) > > 334 waited += self.POLL_INTERVAL > > 335 > > 336 if not self.is_alive and self.task_state() != TaskState.ACTIVE: > > 337 return > > 338 > > 339 log.info('Thermos task did not shut down cleanly, rebinding to > kill.') > 340 self.quitquitquit() > > 341 > > 342 while not self._monitor.finished and waited < timeout: > > 343 self._clock.sleep(self.POLL_INTERVAL.as_(Time.SECONDS)) > > 344 waited += self.POLL_INTERVAL > ``` > > Is it that we need to reset waited to Amount(0, Time.SECONDS) ?
friendly ping - Joe ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/32221/#review77296 ----------------------------------------------------------- On March 18, 2015, 6:20 p.m., Brian Wickman wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/32221/ > ----------------------------------------------------------- > > (Updated March 18, 2015, 6:20 p.m.) > > > Review request for Aurora and Bill Farner. > > > Bugs: AURORA-1054 > https://issues.apache.org/jira/browse/AURORA-1054 > > > Repository: aurora > > > Description > ------- > > Remove excessively low timeout in SIGTERM swallowing test. > > > Diffs > ----- > > src/test/python/apache/aurora/executor/test_thermos_task_runner.py > 6b24bbb2ab7ca16f97961aabeed945b61e5b5908 > > Diff: https://reviews.apache.org/r/32221/diff/ > > > Testing > ------- > > Cannot reproduce locally, but 5 seconds is an impossibly small timeout, even > if we aren't testing SIGTERM swallowing. If this fails, we will get tripped > by 60s timeout instead. > > > Thanks, > > Brian Wickman > >