[
https://issues.apache.org/jira/browse/HBASE-816?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Andrew Purtell updated HBASE-816:
---------------------------------
Attachment: 816.v2.patch
v2 is a less conservative approach that skips the presumably already mapped row
for the key used to restart scanning. Otherwise maybe the work that caused the
lease to expire is performed again, which causes the lease to expire, setting
up an infinite loop.
> TableMap should survive USE
> ---------------------------
>
> Key: HBASE-816
> URL: https://issues.apache.org/jira/browse/HBASE-816
> Project: Hadoop HBase
> Issue Type: Improvement
> Components: mapred
> Affects Versions: 0.2.0
> Reporter: Andrew Purtell
> Assignee: Andrew Purtell
> Priority: Minor
> Attachments: 816.patch, 816.v2.patch
>
>
> If work in a TableMap task exceeds the client lease time, the task will fail
> with an UnknownScannerException. TableInputFormatBase should handle this by
> restarting with a new scanner from the last key seen.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
