You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@mesos.apache.org by "Gavin (JIRA)" <ji...@apache.org> on 2019/04/29 09:27:28 UTC

[jira] [Issue Comment Deleted] (MESOS-4535) Logrotate ContainerLogger may not handle FD ownership correctly

     [ https://issues.apache.org/jira/browse/MESOS-4535?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Gavin updated MESOS-4535:
-------------------------
    Comment: was deleted

(was: www.rtat.net)

> Logrotate ContainerLogger may not handle FD ownership correctly
> ---------------------------------------------------------------
>
>                 Key: MESOS-4535
>                 URL: https://issues.apache.org/jira/browse/MESOS-4535
>             Project: Mesos
>          Issue Type: Bug
>          Components: modules
>            Reporter: Joseph Wu
>            Assignee: Joseph Wu
>            Priority: Blocker
>              Labels: logging, mesosphere
>             Fix For: 0.27.0
>
>
> One of the patches for [MESOS-4136] introduced the {{FDType::OWNED}} enum for {{Subprocess::IO::FD}}.
> The way the logrotate module uses this is slightly incorrect:
> # The module starts a subprocess with an output {{Subprocess::PIPE()}}.
> # That pipe's FD is passed into another subprocess via {{Subprocess::IO::FD(pipe, IO::OWNED)}}.
> # When the second subprocess starts, the pipe's FD is closed in the parent.
> # When the first subprocess terminates, the existing code will try to close the pipe again.  This effectively closes a random FD.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)