[
https://issues.apache.org/jira/browse/MESOS-7090?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15860161#comment-15860161
]
Joseph Wu commented on MESOS-7090:
----------------------------------
When running the Mesos agent in a Docker container, restarting the agent will
result in the death of all executors (except the ones started via Docker, given
the right flags) and companions processes. This includes any of the
non-trivial {{ContainerLoggers}}.
At the moment, we don't plan on fixing the ContainerLoggers to support this
case, because we highly recommend *not* running the Mesos agent in a Docker
container.
> On restart Mesos slave not recreating container loggers
> -------------------------------------------------------
>
> Key: MESOS-7090
> URL: https://issues.apache.org/jira/browse/MESOS-7090
> Project: Mesos
> Issue Type: Bug
> Components: agent
> Affects Versions: 1.0.1, 1.1.0
> Environment: CoreOS, Mesos Slave inside docker
> Reporter: Aleksandr Dorofeev
>
> # Mesos slave running inside docker container
> # MESOS_DOCKER_MESOS_IMAGE is set
> # Tried basic logger and LogrotateContainerLogger
> Flow:
> # Slave successfully create executor container
> # Task started
> # Both task /stdout and /stderr streams is works
> # Slave restarted
> # Task is reregistered on agent
> # /stdout and /stderr streams are stop update
> # Task and executor {{docker logs}} still ok
> When LogrotateContainerLogger is used. {{mesos-logrotate-logger}} is starts
> inside slave container and not started again when slave is restarted.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
