[jira] [Commented] (HBASE-16685) Revisit execution of SnapshotCopy in MapReduceBackupCopyService

["Vladimir Rodionov (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HBASE-16685?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15514740#comment-15514740 ] 

Vladimir Rodionov commented on HBASE-16685:
-------------------------------------------

One handler blocked for hours is OK as long as we have 100 of them. I do not see any issue here. I do not mind changing code if it is not a big one though.  Master does not have direct dependencies on M/R I think - only on backup package.  

> Revisit execution of SnapshotCopy in MapReduceBackupCopyService
> ---------------------------------------------------------------
>
>                 Key: HBASE-16685
>                 URL: https://issues.apache.org/jira/browse/HBASE-16685
>             Project: HBase
>          Issue Type: Improvement
>            Reporter: Ted Yu
>              Labels: backup
>
> During code review of backup / restore, Matteo raised comment on the following code:
> {code}
>         res = snapshotCp.run(options);
> {code}
> I think this will be a first time where HBase server has a direct dependency on a MR job.
> also we need to revisit this code to avoid having one handler blocked for N hours while we are doing a copy.
> This issue is to address the above comment.



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