[
https://issues.apache.org/jira/browse/YARN-7440?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Chandni Singh updated YARN-7440:
--------------------------------
Attachment: YARN-7440.001.patch
Changes in patch 1:
1. During recovery if the service master receives a container whose entry is
not present in the registry, then it will trigger a _container recover_ event
corresponding to a pending component instance. This event will trigger the
_component instance_ to get the status of the container from NM and save it to
registry.
2. During recovery, for the containers whose entry is already present in the
registry, the status is not queried from NM again. Only the internal state of
_component instance_ is modified.
3. A test to verify the above behavior.
> Optimization to AM recovery when the service record doesn't exist for a
> container
> ---------------------------------------------------------------------------------
>
> Key: YARN-7440
> URL: https://issues.apache.org/jira/browse/YARN-7440
> Project: Hadoop YARN
> Issue Type: Sub-task
> Reporter: Chandni Singh
> Assignee: Chandni Singh
> Fix For: yarn-native-services
>
> Attachments: YARN-7440.001.patch
>
>
> When AM recovers, if the service record doesn’t exist for a container sent
> from RM, it can re-query the container status from NM, today it will release
> the container
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
