[ 
https://issues.apache.org/jira/browse/MESOS-960?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13889822#comment-13889822
 ] 

Benjamin Mahler commented on MESOS-960:
---------------------------------------

I found one test that produces zombies so far: 
SlaveRecovery.ReconcileTasksMissingFromSlave.

This test launches a task on the slave but wipes the meta directory, and so the 
subsequent slave run does not capture the pid to reap. When process isolation 
is in use, this results in an "leak" of the child process.

One fix is to kill the child process by reading the pid file from the meta 
directory. Any better ideas?

> Un-reaped executor during test teardown.
> ----------------------------------------
>
>                 Key: MESOS-960
>                 URL: https://issues.apache.org/jira/browse/MESOS-960
>             Project: Mesos
>          Issue Type: Bug
>          Components: test
>            Reporter: Benjamin Mahler
>
> Now that I've added a check for zombie processes during the test envronment 
> tear down, there was an un-reaped executor caught during CI:
> [----------] Global test environment tear-down
> ../../src/tests/environment.cpp:247: Failure
> Failed
> Tests completed with child processes remaining:
> -+- 1958 
> /x1/jenkins/jenkins-slave/workspace/Mesos-Trunk-Ubuntu-Build-Out-Of-Src-Disable-Java-Disable-Python-Disable-Webui/build/src/.libs/lt-mesos-tests
>  \-+- 5894 sh -c 
> /x1/jenkins/jenkins-slave/workspace/Mesos-Trunk-Ubuntu-Build-Out-Of-Src-Disable-Java-Disable-Python-Disable-Webui/build/src/mesos-executor
>    \--- 5912 
> /x1/jenkins/jenkins-slave/workspace/Mesos-Trunk-Ubuntu-Build-Out-Of-Src-Disable-Java-Disable-Python-Disable-Webui/build/src/.libs/lt-mesos-executor
> [==========] 220 tests from 39 test cases ran. (69898 ms total)
> [  PASSED  ] 220 tests.
> [  FAILED  ] 0 tests, listed below:



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)

Reply via email to