hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "huaxiang sun (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-7880) HFile Recovery/Rewrite Tool
Date Fri, 24 Mar 2017 17:35:42 GMT

    [ https://issues.apache.org/jira/browse/HBASE-7880?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15940789#comment-15940789

huaxiang sun commented on HBASE-7880:

We recently run into one case with hfile truncated. User want to recover as much data as possible.
I hacked some quick code based on [~mbertozzi]'s code and it seems work. Maybe we can add
this as an option of hfile tool.

For the case we run into,  the file is truncated, all data blocks are good until the last
one in the trancated hfile, so cells can be recovered until the last block.
For the other cases such as there are corrupted data blockw in the middle, we can skip these
blocks and continue with the next one.

For the corrupted data block, maybe we can recover some data. Not sure about this part yet,
some homework needs to be done. But this can be enhanced a bit later. Any thoughts? Thanks.

> HFile Recovery/Rewrite Tool
> ---------------------------
>                 Key: HBASE-7880
>                 URL: https://issues.apache.org/jira/browse/HBASE-7880
>             Project: HBase
>          Issue Type: New Feature
>          Components: HFile
>    Affects Versions: 0.95.2
>            Reporter: Matteo Bertozzi
>            Assignee: Matteo Bertozzi
>            Priority: Minor
>         Attachments: HBASE-7880-v0.patch
> Sometimes is useful to have a tool to migrate files from a new version to an old version
(e.g. convert a new XYZ encoded/compressed file to an old "uncompressed" format)
> also it will be useful to been able to recover an hfile from a corrupted state. (e.g.
trailer missing/broken, ...) 
> The "user" can provide the information about the file (compression & co) and  try
to recover as much as possible from the file by reading data blocks.

This message was sent by Atlassian JIRA

View raw message