[
https://issues.apache.org/jira/browse/HBASE-8291?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13633160#comment-13633160
]
Varun Sharma commented on HBASE-8291:
-------------------------------------
Nicholas, I was looking at the code and here is what I think:
processBatchCallback in HConnectionManager:
1) We call locateRegion at the beginning of every retry to locate the regions
2) When an exception happens, we call updateCachedLocations which:
a) Refreshes the entry on RegionMovedException
b) Keeps the entry on RegionOpeningException
c) Deletes the entry on any other exception (SocketTimeout, ConnectTimeout
and NotServingRegionException)
We could follow a similar pattern IMO:
1) connect() call always calls getRegionLocation without forcing reloads
(currently upon any retry, we refresh the region location)
2) When we have an exception, we call updateCachedLocations which handles all
kinds of exceptions and modifies the cache - this automatically takes care of
RegionMovedException, RegionOpeningException and NotServingRegionException
Since we rely on updateCachedLocations to do the right thing (refresh/delete
stale cached locations) - we no longer need to do a reload in the connect()
call in ServerCallable. Let me know what you think ?
Thanks
Varun
> ServerCallable should handle RegionMovedException
> -------------------------------------------------
>
> Key: HBASE-8291
> URL: https://issues.apache.org/jira/browse/HBASE-8291
> Project: HBase
> Issue Type: Bug
> Components: Client
> Affects Versions: 0.95.0
> Reporter: Nicolas Liochon
>
> See HBASE-8285
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
