[
https://issues.apache.org/jira/browse/HBASE-23634?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17148246#comment-17148246
]
Michael Stack commented on HBASE-23634:
---------------------------------------
Another argument for split to hfile rather than recovered.edit apart from
improved MTTR is that on the occasion when hbase manufactures an 'orphan
region', a region that it has lost accounting on, our 'fix' is for the operator
to run bulk load of the Region back up into the table to paper over any
possibility of data loss. This works if the Region has hfiles only. We do not
yet have a bulk load that will hoist hfiles AND any recovered.edits that may be
hanging out in the Region back up into the live cluster.
e.g. here was an orphan region... only had recovered.edits files in it.
{code}
5.1 If Orphan Region is nought but recovered edits files, how to load up these?
$ sudo -u hdfs hdfs dfs -ls -R /hbase/t1/55bf1d59d31199658f61679e9c8a8793
drwxr-xr-x - hbase hadoop 0 2019-11-21 01:42
/hbase/t1/55bf1d59d31199658f61679e9c8a8793/recovered.edit
-rw-r--r-- 3 hbase hadoop 387922 2019-11-21 01:42
/hbase/t1/55bf1d59d31199658f61679e9c8a8793/recovered.edits/0000000000000771292
-rw-r--r-- 3 hbase hadoop 739746 2019-11-21 01:42
/hbase/t1/55bf1d59d31199658f61679e9c8a8793/recovered.edits/0000000000000771295
-rw-r--r-- 3 hbase hadoop 417407 2019-11-21 01:41
/hbase/t1/55bf1d59d31199658f61679e9c8a8793/recovered.edits/0000000000000771298
{code}
> Enable "Split WAL to HFile" by default
> --------------------------------------
>
> Key: HBASE-23634
> URL: https://issues.apache.org/jira/browse/HBASE-23634
> Project: HBase
> Issue Type: Task
> Affects Versions: 3.0.0-alpha-1, 2.3.0
> Reporter: Guanghao Zhang
> Priority: Blocker
> Fix For: 2.4.0
>
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
