[
https://issues.apache.org/jira/browse/HBASE-11059?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14027221#comment-14027221
]
Jimmy Xiang commented on HBASE-11059:
-------------------------------------
bq. During master starts I didn't see where we handle regions in PENDING_OPEN
state from META.
I see. We should send the RPC again for both PENDING_OPEN and PENDING_CLOSE at
master start in case failover.
bq. Before we offline region and then set FAILED_OPEN in ZK. This is likely a
new behavior.
I added the offline region part. I mean the race is an existing issue. With ZK,
RS can set FAILED_OPEN in ZK, before clean up the RS internal memory state, the
master could handle the FAILED_OPEN ZK event and try to open the region on the
same RS again and get back already opened. There seems not to have an easy fix.
Probably a little synchronization is needed. Perhaps we can address this one in
a separate issue.
> ZK-less region assignment
> -------------------------
>
> Key: HBASE-11059
> URL: https://issues.apache.org/jira/browse/HBASE-11059
> Project: HBase
> Issue Type: Improvement
> Components: master, Region Assignment
> Reporter: Jimmy Xiang
> Assignee: Jimmy Xiang
> Fix For: 0.99.0
>
> Attachments: hbase-11059.patch, hbase-11059_v2.1.patch,
> hbase-11059_v2.2.patch, hbase-11059_v2.patch, zk-less_am.pdf,
> zk-less_assignment.png
>
>
> It seems that most people don't like region assignment with ZK (HBASE-5487),
> which causes many uncertainties. This jira is to support ZK-less region
> assignment. We need to make sure this patch doesn't break backward
> compatibility/rolling upgrade.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
