[
https://issues.apache.org/jira/browse/YARN-11475?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Sunil Kalva updated YARN-11475:
-------------------------------
Language: (was: java)
> Support for a new flag to decide whether to delete or retain the NM local
> state with a decommission command
> -----------------------------------------------------------------------------------------------------------
>
> Key: YARN-11475
> URL: https://issues.apache.org/jira/browse/YARN-11475
> Project: Hadoop YARN
> Issue Type: New Feature
> Components: client, nodemanager, resourcemanager
> Reporter: Sunil Kalva
> Assignee: Sunil Kalva
> Priority: Major
> Labels: pull-request-available
>
> We want to retain NM local state store while decommissioning the NM node in a
> few cases. For example, if there are any auxiliary services(for example
> external spark shuffle service) running and have created local data then when
> we decommission and recommission after upgrades, we want to retain the NM
> local state store so that these services keep serving the earlier data.
> This change is about passing a new flag from yarn admin CLI (-deleteState
> default true) with refreshNodes command to RM, and then sending the same to
> the NM along with noderefresh response to NM from RM.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
