[
https://issues.apache.org/jira/browse/HBASE-24900?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17274147#comment-17274147
]
Karthik Palanisamy commented on HBASE-24900:
--------------------------------------------
Thank you [~pankajkumar]. This is super important Jira.
I was checking one of my CDP cluster where the region never retained on every
graceful HBase stop and start. It looks like this problem after
https://issues.apache.org/jira/browse/HBASE-23035 change.
On every SCP, the call set to forceNewPlan=true
(ServerCrashProcedure#assignRegions) which will assign region to the new RS.
[https://github.com/apache/hbase/blob/master/hbase-server/src/main/java/org/apache/hadoop/hbase/master/procedure/ServerCrashProcedure.java#L536]
{code}
2021-01-26 05:13:02,455 INFO
org.apache.hadoop.hbase.master.assignment.TransitRegionStateProcedure: Starting
pid=9, ppid=2, state=RUNNABLE:REGION_STATE_TRANSITION_GET_ASSIGN_CANDIDATE,
locked=true; TransitRegionStateProcedure table=atlas_janus,
region=e4e72a9453107e1487f2fa79a03c8158, ASSIGN; rit=OPEN, location=null;
forceNewPlan=true, retain=false{code}
Users started reporting data locality problem due to this issue.
> Make retain assignment configurable during SCP
> ----------------------------------------------
>
> Key: HBASE-24900
> URL: https://issues.apache.org/jira/browse/HBASE-24900
> Project: HBase
> Issue Type: Sub-task
> Components: amv2
> Affects Versions: 3.0.0-alpha-1, 2.3.1, 2.1.9, 2.2.5
> Reporter: Pankaj Kumar
> Assignee: Pankaj Kumar
> Priority: Major
> Fix For: 3.0.0-alpha-1
>
>
> HBASE-23035 change the "retain" assignment to round-robin assignment during
> SCP which will make the failover faster and surely improve the availability,
> but this will impact the scan performance in non-cloud scenario.
> This jira will make this assignment plan configurable.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
