[
https://issues.apache.org/jira/browse/HBASE-18049?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16011767#comment-16011767
]
Jingcheng Du commented on HBASE-18049:
--------------------------------------
Thanks [~yuzhih...@gmail.com] and [~anoop.hbase] for the comments!
bq. So here in new change as well, we need getCause() check?
Right, we should add this.
bq. A solution to restore the state and data is to do a MOB major compaction
right?
Right, this is a simple solution which can probably fix the issue. But I think
users need to investigate the cases a little before running such commands.
Given such info in logs might mislead users to directly run the compaction? I
think we'd better not add this for now?
> It is not necessary to re-open the region when MOB files cannot be found
> ------------------------------------------------------------------------
>
> Key: HBASE-18049
> URL: https://issues.apache.org/jira/browse/HBASE-18049
> Project: HBase
> Issue Type: Bug
> Components: Scanners
> Reporter: Jingcheng Du
> Assignee: Jingcheng Du
> Fix For: 2.0.0
>
> Attachments: HBASE-18049.patch
>
>
> In HBASE-17712, we try to re-open the region when store files cannot be
> found. This is useful for store files in a region, but is not necessary when
> the MOB files cannot be found, because the store files in a region only
> contain the references to the MOB files and a re-open of a region doesn't
> help the lost MOB files.
> In this JIRA, we will directly throw DNRIOE only when the MOB files are not
> found in {{MobStoreScanner}} and {{ReversedMobStoreScanner}}. Other logics
> keep the same.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
