hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "stack (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-11815) Flush and compaction could just close the tmp writer if there is an exception
Date Tue, 26 Aug 2014 15:24:58 GMT

    [ https://issues.apache.org/jira/browse/HBASE-11815?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14110823#comment-14110823
] 

stack commented on HBASE-11815:
-------------------------------

Patch lgtm.  We failing when we try to add the metadata after getting an IOE?  That would
make sense.  Should we do this on every IOE?  Will we have files that are half-written [~ram_krish]
w/o their metadata?  Do they get cleaned up on IOE?  If so, +1.

> Flush and compaction could just close the tmp writer if there is an exception
> -----------------------------------------------------------------------------
>
>                 Key: HBASE-11815
>                 URL: https://issues.apache.org/jira/browse/HBASE-11815
>             Project: HBase
>          Issue Type: Bug
>            Reporter: ramkrishna.s.vasudevan
>            Assignee: ramkrishna.s.vasudevan
>            Priority: Trivial
>             Fix For: 0.99.0, 2.0.0, 0.98.6
>
>         Attachments: HBASE-11815.patch, HBASE-11815_1.patch
>
>
> A minor change.
> {code}
>         try {
>           flushed = performFlush(scanner, writer, smallestReadPoint);
>         } finally {
>           finalizeWriter(writer, cacheFlushId, status);
>         }
> {code}
> Whenever there is a failure during flush we should close the writer but adding the meta
data and setting the status would not be needed.
> {code}
>     status.setStatus("Flushing " + store + ": appending metadata");
>     writer.appendMetadata(cacheFlushSeqNum, false);
>     status.setStatus("Flushing " + store + ": closing flushed file");
>     writer.close();
> {code}



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Mime
View raw message