[
https://issues.apache.org/jira/browse/HBASE-11059?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14028304#comment-14028304
]
Jeffrey Zhong commented on HBASE-11059:
---------------------------------------
{quote}
can we handle the pending_open/close issue in a separate jira? I will add some
tests to cover it.
{quote}
Fine with me.
There maybe a region double assignment issue when we turn the feature on(from
hbase.assignment.usezk=ON to hbase.assignment.usezk=OFF) during rolling
restart. Below are details:
1) Since we rolling-restart RS firstly, A RS will use the new
config(hbase.assignment.usezk=OFF)
2) Master is still using hbase.assignment.usezk=ON so AM creates offline znode
before sending region open. Though a region will be opened, the OFFLINE znode
will be left even after the region is opened.
3) When master restarts with new config(hbase.assignment.usezk=OFF), AM will
process RITs left over in ZK. During RIT processing on offline znode, the
region will be assigned somewhere while it's still open in another RS.
> 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)
