hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "bharath v (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-8963) Add configuration option to skip HFile archiving
Date Thu, 20 Mar 2014 09:59:45 GMT

    [ https://issues.apache.org/jira/browse/HBASE-8963?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13941572#comment-13941572

bharath v commented on HBASE-8963:


The main motivation for this option are the following

1) Sometimes customers want to retain hfiles in the archive directory for a day or so after
deletes, as a backup. So they configure the cleaner to run every 24hours to cleanup the files.
This obviously takes up space. However not all tables require this feature and some can be
deleted from disk as soon as the table is deleted / compactions are done. So this patch helps
such situations where they can mark skip_archive for such test tables and not let it pass
through archive so that they can reclaim the space immediately after deletion. 

2) Avoid unnecessary NN operations for not-so-useful tables. If we move to the archive, instead
of N file delete calls, we make N file moves and then do N file delete calls. Depending on
the table size N can be very huge and we can avoid putting pressure on NN. 

> Add configuration option to skip HFile archiving
> ------------------------------------------------
>                 Key: HBASE-8963
>                 URL: https://issues.apache.org/jira/browse/HBASE-8963
>             Project: HBase
>          Issue Type: Improvement
>            Reporter: Ted Yu
>            Assignee: bharath v
>             Fix For: 0.99.0
>         Attachments: HBASE-8963.trunk.v1.patch, HBASE-8963.trunk.v2.patch, HBASE-8963.trunk.v3.patch,
HBASE-8963.trunk.v4.patch, HBASE-8963.trunk.v5.patch, HBASE-8963.trunk.v6.patch, HBASE-8963.trunk.v7.patch
> Currently HFileArchiver is always called when a table is dropped.
> A configuration option (either global or per table) should be provided so that archiving
can be skipped when table is deleted.

This message was sent by Atlassian JIRA

View raw message