[
https://issues.apache.org/jira/browse/MESOS-6664?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15727566#comment-15727566
]
Kevin Klues commented on MESOS-6664:
------------------------------------
{noformat}
commit ff553e86ade41a47121e6aaf8a5e5fdee1b4e9e6
Author: Kevin Klues <[email protected]>
Date: Tue Dec 6 15:21:45 2016 -0800
Added SIGTERM handler to gracefully shutdown IOSwitchboard server.
When receiving a SIGTERM, the io switchboard process will forcibly
unblock the server from waiting on a connection before attempting to
drain its `stdoutFromFd` and `stderrFromFd` file descriptors. Once
these fds are drained (or they become invalid), the server will shut
itself down as per the normal exit route.
Review: https://reviews.apache.org/r/54442/
{noformat}
{noformat}
commit 85f0c462ad807f2842d9debeb8ee8ce0cc0c7fe4
Author: Kevin Klues <[email protected]>
Date: Tue Dec 6 15:21:42 2016 -0800
Added function to unblock IOSwitchboard when waiting for connection.
Review: https://reviews.apache.org/r/54441/
{noformat}
> Force cleanup of IOSwitchboard server if it does not terminate after the
> container terminates.
> ----------------------------------------------------------------------------------------------
>
> Key: MESOS-6664
> URL: https://issues.apache.org/jira/browse/MESOS-6664
> Project: Mesos
> Issue Type: Task
> Reporter: Jie Yu
>
> In normal case, IOSwitchboard server will terminate after container
> terminates. However, we should be more defensive and always cleanup the
> IOSwitchboard server if it does not terminate within a reasonable grace
> period.
> The reason for the grace period is to allow the IOSwitchboard server to
> finish redirecting the stdout/stderr to the logger.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
