[
https://issues.apache.org/jira/browse/STORM-1191?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14998633#comment-14998633
]
ASF GitHub Bot commented on STORM-1191:
---------------------------------------
Github user revans2 commented on the pull request:
https://github.com/apache/storm/pull/869#issuecomment-155431246
@ashnazg the test "might" "logically" pass. The code is a bit complex.
When debugging these in the past the tests use simulated time, which is a great
thing it lets us run tests that would take a long time very quickly, but to
make it work properly there are some critical points where we wait for the
simulated cluster to be idle. Typically we see timeouts happen here, waiting
for the entire cluster to stop. But for some reason on really slow hardware,
that I don't totally understand yet, this does not always happen. and
increasing the timeout does not always induce it to be in that state.
I am +1 on merging this in, but I am not totally sure that it will "fix"
the problem.
> Random test failures on backtype.storm.transactional-test
> ---------------------------------------------------------
>
> Key: STORM-1191
> URL: https://issues.apache.org/jira/browse/STORM-1191
> Project: Apache Storm
> Issue Type: Bug
> Reporter: Chuck Burgess
> Assignee: Chuck Burgess
> Priority: Minor
>
> Travis builds show occasional backtype.storm.transactional-test failures due
> to the 100s timeout (https://travis-ci.org/apache/storm/jobs/90105000).
> Assuming this is not a performance test, I'll submit PR to bump timeout by
> 50% to cut back on these failures.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
