hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Enis Soztutar (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-8036) ProtobufUtil.multi behavior is inconsistent in case of errors
Date Tue, 12 Mar 2013 23:31:13 GMT

    [ https://issues.apache.org/jira/browse/HBASE-8036?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13600616#comment-13600616

Enis Soztutar commented on HBASE-8036:

TestMultiParallel.testActiveThreadsCount() failed again. That test checks the HTable pool
size to see whether there we create one thread per region server. 
The error log indicates that the pool's behavior is correct, but the META thinks that there
are still 3 servers lying around (a side effect of the previous test method testFlushCommitsWithAbort()).
Will investigate more.
> ProtobufUtil.multi behavior is inconsistent in case of errors
> -------------------------------------------------------------
>                 Key: HBASE-8036
>                 URL: https://issues.apache.org/jira/browse/HBASE-8036
>             Project: HBase
>          Issue Type: Bug
>    Affects Versions: 0.95.0
>            Reporter: Sergey Shelukhin
>            Assignee: Enis Soztutar
>            Priority: Blocker
>             Fix For: 0.95.0
>         Attachments: hbase-8036_v1.patch, hbase-8036_v2.patch, hbase-8036_v3.patch
> ProtobufUtil splits operations by regions and performs multiple client.multi calls. In
case if there are certain errors inside RS, HRegionServer adds the corresponding exceptions
to MultiResponse, PU continues the multi request for other regions, and returns partial failure.

> In case of other errors (for example, region not served exception), the entire multi
operation stops executing, and previous successes and partial results are disregarded.
> ProtobufUtil should probably catch ServiceException separately for each client.multi
call, make it a partial-failure exception for all actions for this region, and also continue
the batch, to make the behavior consistent.
> Alternatively, if we want to avoid continuing the batch in case of some server-wide errors/connection
problems/etc., server should do that for region-specific errors (add exception to results
for each action).

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

View raw message