[
https://issues.apache.org/jira/browse/YARN-1040?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15166462#comment-15166462
]
Arun Suresh commented on YARN-1040:
-----------------------------------
So that we are on the same page, If we were to separate what needs to be in
YARN vs what Slider etc. should handle, id say :
*YARN*
* Container Upgrade primitive:
** provide AM with APIs (via NMClient) to upgrade the Container.
** API takes 1) new {{ContainerLaunchContext}} and 2) a policy viz. *In-place*
(localize in parallel v2, start v2, stop v1) or *New+rollback* (stop v1,
localize v2, start v2) + (start v1 if start v2 fails) *or* list of primitive
composable commands if the above policies doesn't cover the use case.
** should negotiate Resource increase for in-place upgrade with RM prior to
upgrade via YARN-1197 (or perhaps use OPPORTUNISTIC containers to locally
negotiated at the NM for the resource spike needed for upgrade, once YARN-2877
is ready)
*Slider / or something similar*
* Application upgrade primitive
** Upgrade Orchestration Policy: Allow applications deployed via slider to
specify order in which tasks/roles are upgraded (or started)
** Allow applications to specify how containers of each role are upgraded
** Actually call the YARN container upgrade APIs (described above) to perform
upgrade of each container in the user specified order/policy
Makes sense ?
> De-link container life cycle from the process and add ability to execute
> multiple processes in the same long-lived container
> ----------------------------------------------------------------------------------------------------------------------------
>
> Key: YARN-1040
> URL: https://issues.apache.org/jira/browse/YARN-1040
> Project: Hadoop YARN
> Issue Type: Sub-task
> Components: nodemanager
> Affects Versions: 3.0.0
> Reporter: Steve Loughran
>
> The AM should be able to exec >1 process in a container, rather than have the
> NM automatically release the container when the single process exits.
> This would let an AM restart a process on the same container repeatedly,
> which for HBase would offer locality on a restarted region server.
> We may also want the ability to exec multiple processes in parallel, so that
> something could be run in the container while a long-lived process was
> already running. This can be useful in monitoring and reconfiguring the
> long-lived process, as well as shutting it down.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
