hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Ashish Singhi (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-15446) Do not abort RS when WAL append for bulk load event marker is failed
Date Fri, 11 Mar 2016 09:56:53 GMT

    [ https://issues.apache.org/jira/browse/HBASE-15446?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15190736#comment-15190736
] 

Ashish Singhi commented on HBASE-15446:
---------------------------------------

bq. When u say this, u mean fail op to user app which tried to bulk load the file? But the
file is actually loaded here in this cluster.
Yes, but I don't see any issue if its loaded again, data is same but then we will have another
hfile with different seq id but I don't see any major impact with it.
Also this will happen only when the input data is not in same FS. As in the case of same FS
(most common use case I believe) the hfiles will be actually moved so on next retry we will
not find this file to load and will get FileNotFoundException which I think we can catch that
exception and ignore it.

> Do not abort RS when WAL append for bulk load event marker is failed
> --------------------------------------------------------------------
>
>                 Key: HBASE-15446
>                 URL: https://issues.apache.org/jira/browse/HBASE-15446
>             Project: HBase
>          Issue Type: Bug
>    Affects Versions: 1.3.0
>            Reporter: Ashish Singhi
>            Assignee: Ashish Singhi
>
> During bulk load process when the RS fails to append the bulk load event marker in WAL
we abort that RS which is actually not really required. Instead just throw back the exception
to the client and let the client retry. This will be helpful in case of replication of bulk
loaded data.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message