[jira] [Commented] (HBASE-22843) [HBCK2] Fix HBCK2 after HBASE-22777 & HBASE-22758

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

    [ https://issues.apache.org/jira/browse/HBASE-22843?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16906409#comment-16906409 ] 

stack commented on HBASE-22843:
-------------------------------

Good on you Sakthi. 

On #1, it was this change that changed the access (by me). It is not in a release yet so we could change it back if we do it quick?

HBASE-22777 Add a multi-region merge (for fixing overlaps)  Makes MergeTableRegionsProcedure do more than just two regions at a time. Compatible as MTRP was done considering one day it'd do more than two at a time.  Changes hardcoded assumption that merge parent regions are named mergeA and mergeB in a column on the resultant region. Instead can have N columns on the merged region, one for each parent merged. Column qualifiers all being with 'merge'.  Most of code below is undoing the assumption that there are two parents on a merge only.

Same for #2.

On #3, it was changed by openinx in HBASE-22758 Remove the unneccesary info cf deletion in DeleteTableProcedure#deleteFromMeta (#424)

This has not made a release yet.

Trouble with depending on hbase for these methods is that then HBCK2 is tied tightly to particular patch releases. I'd say in future we should copy the methods local. For now I think we should change hbase itself (quickly -- smile).







> [HBCK2] Fix HBCK2 after HBASE-22777 & HBASE-22758
> -------------------------------------------------
>
>                 Key: HBASE-22843
>                 URL: https://issues.apache.org/jira/browse/HBASE-22843
>             Project: HBase
>          Issue Type: Sub-task
>          Components: hbase-operator-tools
>            Reporter: Sakthi
>            Assignee: Sakthi
>            Priority: Major
>
> There were few changes made in HBASE-22777 & HBASE-22758, which would require their propogation to HBCK2 as well to prevent compilation issues. These are the ones:
> 1. MetaTableAccessor#makePutFromTableState was changed from public to private. HBCK2 uses it in HBaseFsck#generatePuts() for rebuildMeta(). We either need to revert the change to public or make similar adjustments in HBCK2.
> 2.
> {code:java}
>  -PairOfSameType<RegionInfo> getMergeRegions(Result data) 
>  +List<RegionInfo> getMergeRegions(Cell[] cells){code}
> Due to this change, HBaseFsck was fixed accordingly but not HBCK2. Let's do it there as well.
> 3. MetaTableAccessor#deleteRegion() was renamed to MetaTableAccessor#deleteRegionInfo(). Even this should be integrated with HBCK2.



--
This message was sent by Atlassian JIRA
(v7.6.14#76016)