[
https://issues.apache.org/jira/browse/HBASE-22855?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16908073#comment-16908073
]
jianghua zhu edited comment on HBASE-22855 at 8/15/19 1:36 PM:
---------------------------------------------------------------
There is a possibility:
If there are HMaster, HRegion Server 1, HRegion Server 2, the following process
occurs.
1. HMaster sends requests to HRegionServer1 through RPC through assignment
mechanism, and loads the region information corresponding to hbase: meta.
2. For some reason, HRegion Server itself did not report its status to HMaster.
HMaster believed that HRegion Server had died at this time.
3. HMaster requests HRegionServer2 through assignment mechanism and loads the
region information corresponding to hbase: meta. And HRegionServer2 loads
quickly.
4. At this time, the location of hbase: meta is located at the HRegionServer2
node.
5. At this point, HRegionServer1 itself restores, reloads hbase: meta, and
completes. At this point, the location of hbase: meta changes.
was (Author: jianghuazhu):
There is one possibility:
Assuming that HMaster, HRegionServer1, and HRegionServer2 exist, the following
processes occur.
1.HMaster sends a request to HRegionServer1 via RPC via the assign mechanism to
load the Region information corresponding to hbase:meta.
2. The HRegionServer itself, for some reason, failed to report state to
HMaster, and HMaster assumed that the HRegionServer was dead.
3.HMaster makes a request to HRegionServer2 via the assign mechanism to load
the Region information corresponding to hbase:meta. And the HRegionServer2
loads quickly.
4. Retrieve the location of the hbase:meta at HRegionServer2
> Find the root cause why still using the old hbase:meta location after meta
> region moving.
> -----------------------------------------------------------------------------------------
>
> Key: HBASE-22855
> URL: https://issues.apache.org/jira/browse/HBASE-22855
> Project: HBase
> Issue Type: Bug
> Reporter: Zheng Hu
> Assignee: Zheng Hu
> Priority: Major
>
> In HBASE-214644, [~andrew.purt...@gmail.com] had a nice fix about the using
> the stale hbase:meta location, say update the hbase:meta location
> periodically. Also someone in user list also encountered this bug in
> HBASE1.4.9 [1].
> [~wenbang] from DiDi also have a question about the root cause under
> HBASE-214644 [2].
> I think we need to find out the root cause and fix it ?
> [1].
> https://mail-archives.apache.org/mod_mbox/hbase-user/201908.mbox/%3C114F31E6-B7F9-4209-AAE3-32A022729593%40gmail.com%3E
> [2].
> https://issues.apache.org/jira/browse/HBASE-21464?focusedCommentId=16902808&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-16902808
--
This message was sent by Atlassian JIRA
(v7.6.14#76016)
