[
https://issues.apache.org/jira/browse/HBASE-18490?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16189316#comment-16189316
]
ramkrishna.s.vasudevan commented on HBASE-18490:
------------------------------------------------
[~huaxiang]
Thanks for the info. Yes you were right. I checked all cases with my test case
in branch-1 and branch-1.2. I was thinking that the regions though they are
assigned the META upate with the replica id for the server name, server seq
column etc are not happening. But they too seem to happen. One difference is
that in branch-1 and branch-2 is that in branch-1 we explicity call
retainAssignment() for the existing replica regions and for the new one we do
round robin. In branch-2 whether to force newPlan or not is determined
internally in the Assignmentprocedure.
But in either case is it guarenteed that when the replica is altered no two
replicas wil be in the same RS? Load balancer should take care of it ideally.
> Modifying a table descriptor to enable replicas does not create replica
> regions
> -------------------------------------------------------------------------------
>
> Key: HBASE-18490
> URL: https://issues.apache.org/jira/browse/HBASE-18490
> Project: HBase
> Issue Type: Bug
> Components: Region Assignment
> Affects Versions: 2.0.0-alpha-1
> Reporter: ramkrishna.s.vasudevan
> Assignee: ramkrishna.s.vasudevan
> Fix For: 2.0.0-beta-1
>
> Attachments: HBASE-18490_master_1.patch, HBASE-18490_master_2.patch,
> HBASE-18490_master.patch, TestRegionReplicasWithRestartScenarios.java
>
>
> After creating a table, if we try to modify the table to enable region
> replication, the new Htable Descriptor is not taken into account and the
> table is enabled again with default single region.
> Ping [~enis], [~tedyu], [~devaraj].
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
