[
https://issues.apache.org/jira/browse/HBASE-2104?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12800902#action_12800902
]
Lars George commented on HBASE-2104:
------------------------------------
Robert, see https://issues.apache.org/jira/browse/HBASE-2013 I have a unit test
there that I think simulates your scenario. It creates a case where during a MR
job a RegionServer is killed. You have a 50/50 chance that the regions are not
redeployed before the job passes the missing region.
That is sort of a race condition and partially a design feature. But I have I
guess similar concerns where I need to be able to rely on the MR to get all
rows. I better have the job fail then silently skip rows.
> No exception thrown while during scanning no connection can be made to the
> regionserver
> ---------------------------------------------------------------------------------------
>
> Key: HBASE-2104
> URL: https://issues.apache.org/jira/browse/HBASE-2104
> Project: Hadoop HBase
> Issue Type: Bug
> Components: client
> Affects Versions: 0.20.1
> Reporter: Robert Hofstra
>
> When a regionservers is stopped (shutdown or crash) and on the same moment a
> client performs a scan on that regionserver no exception is thrown at client
> side nor a reconnect toanother regionsserver is tried.
> The ResultScanner.Iterator.hasNext ()just returns false, so the client
> assumes that there are no records anymore.
> In the ScannerCallable.call I notice that the java.net.ConnectionException is
> catched and a empyt Result array is returned.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
