[
https://issues.apache.org/jira/browse/YARN-7973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16414336#comment-16414336
]
Shane Kumpf commented on YARN-7973:
-----------------------------------
{quote}Sorry, I am not clear about the design of container relaunch feature. In
what scenario is container relaunch used?
{quote}
Please see the existing {{ContainerRelaunch}} feature (YARN-3998) to better
understand the initial design. This JIRA is for properly handling that feature
with the Docker runtime. The {{ContainerRetryPolicy}} used by Native Services
results in the use of this feature.
{quote}what would happen if the intermediate state of the container is
preventing relaunch to run successfully?
{quote}
It is going to depend on your configuration. By default, Native Services
relaunches every 30 seconds until the app lifetime is exceeded. This is the
behavior with or without this patch. With a retry count set, the container will
fail after relaunching the specified number of times.
How relaunch is used, is up to the application/AM, so we can't just look at how
Native Services is using it, we need to fix relaunch for the Docker case.
As previously mentioned, IMO, we have two options:
1) The approach taken here to call "docker start" on the existing container.
2) Delete and launch a new Docker container with the same container ID name.
Given the design behind YARN-3998, #1 appears to be most appropriate. This may
allow some applications to recover existing data, which I believe to be
desirable.
> Support ContainerRelaunch for Docker containers
> -----------------------------------------------
>
> Key: YARN-7973
> URL: https://issues.apache.org/jira/browse/YARN-7973
> Project: Hadoop YARN
> Issue Type: Sub-task
> Reporter: Shane Kumpf
> Assignee: Shane Kumpf
> Priority: Major
> Attachments: YARN-7973.001.patch, YARN-7973.002.patch
>
>
> Prior to YARN-5366, {{container-executor}} would remove the Docker container
> when it exited. The removal is now handled by the
> {{DockerLinuxContainerRuntime}}. {{ContainerRelaunch}} is intended to reuse
> the workdir from the previous attempt, and does not call {{cleanupContainer}}
> prior to {{launchContainer}}. The container ID is reused as well. As a
> result, the previous Docker container still exists, resulting in an error
> from Docker indicating the a container by that name already exists.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
