[
https://issues.apache.org/jira/browse/STORM-1377?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15060401#comment-15060401
]
ASF GitHub Bot commented on STORM-1377:
---------------------------------------
Github user d2r commented on the pull request:
https://github.com/apache/storm/pull/941#issuecomment-165193668
> It's an unlikely scenario and I might be paranoid, but what if the tests
were running on a cluster under heavy traffic? For instance, what if the tests
were to be run on OpenStack, and the test just happened to run when there is
unrelated heavy traffic (say, Hadoop shuffle) pressuring the switches?
This should be using the loopback device. I think the main problem here is
the CPU was too busy to have the thrift server reply within 30ms. If traffic
were the issue, we would see many more of our tests fail similarly.
> nimbus_auth_test: very short timeouts causing spurious failures
> ---------------------------------------------------------------
>
> Key: STORM-1377
> URL: https://issues.apache.org/jira/browse/STORM-1377
> Project: Apache Storm
> Issue Type: Bug
> Components: storm-core
> Affects Versions: 0.10.0, 0.11.0
> Reporter: Derek Dagit
> Assignee: Derek Dagit
> Priority: Minor
>
> This is caused by a units mismatch. We are waiting 30 ms for the thrift
> server to reply when we thought we were waiting 30s. This means that
> sometimes when we expect NotAliveException, we instead get
> TTransportException(SocketTimeoutException), and this fails the assertions.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
