[
https://issues.apache.org/jira/browse/YARN-3225?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14380102#comment-14380102
]
Junping Du commented on YARN-3225:
----------------------------------
Thanks for quickly response, [~devaraj.k]!
bq. If we make the refreshNodesForcefully as same as refreshNodes() API, there
could be a chance of becoming node state incosistent if the hosts file updated
with the newnodes or removal of existing nodes after the refreshNodesGracefully
and before the refreshNodesForcefully.
+1. That's good point.
bq. I feel we need to have difference that
refreshNodes()/refreshNodesGracefully() should read the hosts file and
refreshNodesForcefully() would not read the hosts file and only move the node
state from DECOMMISSIONING to DECOMMISSIONED.
Agree that this sounds better. However, as my suggestion above, we can still
keep the same Admin API. In AdminService side, we can handle request (not
boolean value now, could be a new enum :)) in three cases separately.
> New parameter or CLI for decommissioning node gracefully in RMAdmin CLI
> -----------------------------------------------------------------------
>
> Key: YARN-3225
> URL: https://issues.apache.org/jira/browse/YARN-3225
> Project: Hadoop YARN
> Issue Type: Sub-task
> Reporter: Junping Du
> Assignee: Devaraj K
> Attachments: YARN-3225-1.patch, YARN-3225.patch, YARN-914.patch
>
>
> New CLI (or existing CLI with parameters) should put each node on
> decommission list to decommissioning status and track timeout to terminate
> the nodes that haven't get finished.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
