[jira] [Commented] (MESOS-1222) Turning on GTEST_SHUFFLE on the master branch triggers many failures in SlaveRecoveryTest/0.*

["Ian Downes (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/MESOS-1222?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13978387#comment-13978387 ] 

Ian Downes commented on MESOS-1222:
-----------------------------------

The root cause of all test failures described here was because libev was waiting on all children (patch was not applied, see MESOS-1229). MesosContainerizer was unable to get the exit status from mesos-fetcher and failed to launch containers leading to tests failing in different ways.

> Turning on GTEST_SHUFFLE on the master branch triggers many failures in SlaveRecoveryTest/0.*
> ---------------------------------------------------------------------------------------------
>
>                 Key: MESOS-1222
>                 URL: https://issues.apache.org/jira/browse/MESOS-1222
>             Project: Mesos
>          Issue Type: Bug
>            Reporter: Chi Zhang
>            Assignee: Ian Downes
>         Attachments: gmock_warning.tar.gz, mock_function_called_more_times.tar.gz, registerExecutor_or_status_or_shutdownFramework_msg_lost.tar.gz, task_lost.tar.gz
>
>
> I pulled from the master branch a moment ago and used the following command to run recovery tests.
> MESOS_VERBOSE=1 GLOG_v=1 ./bin/mesos-tests.sh --gtest_filter='SlaveRecoveryTest/0.*' --gtest_shuffle --gtest_repeat=50 --gtest_break_on_failure 2>&1 | tee /tmp/recover_apache.log.$$
> Ran it 12 times. The errors seemed to show up pretty randomly. I roughly put them into 4 categories and attached separately.
> I think the 3 patches from Ian are likely very useful: 20428 20429 20430.
> Also, would including GTEST_SHUFFLE as default increase our confidence for make check?



--
This message was sent by Atlassian JIRA
(v6.2#6252)