[HBASE-22976] [HBCK2] Add RecoveredEditsPlayer

• Type: Sub-task
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 3.0.0-alpha-1, 2.3.3, 2.4.0, 2.2.7
• Component/s: hbck2, walplayer
• Hadoop Flags: Reviewed

Release Note

WALPlayer can replay the content of recovered.edits directories.

Side-effect is that WAL filename timestamp is now factored when setting start/end times for WALInputFormat; i.e. wal.start.time and wal.end.time values on a job context. Previous we looked at wal.end.time only. Now we consider wal.start.time too. If a file has a name outside of wal.start.time<->wal.end.time, it'll be by-passed. This change-in-behavior will make it easier on operator crafting timestamp filters processing WALs.

• Assignee: Michael Stack
• Reporter: Michael Stack
• Created: 05/Sep/19 18:31
• Updated: 10/Oct/20 16:20
• Resolved: 09/Oct/20 16:21

Description

We need a recovered edits player. Messing w/ the 'adoption service' – tooling to adopt orphan regions and hfiles – I've been manufacturing damaged clusters by moving stuff around under the running cluster. No reason to think that an hbase couldn't lose accounting of a whole region if a cataclysm. If so, region will have stuff like the '.regioninfo', dirs per column family w/ store files but it could too have a 'recovered_edits' directory with content in it. We have a WALPlayer for errant WALs. We have the FSHLog tool which can read recovered_edits content for debugging data loss. Missing is a RecoveredEditsPlayer.

I took a look at extending the WALPlayer since it has a bunch of nice options and it can run at bulk. Ideally, it would just digest recovered edits content if passed an option or recovered edits directories. On first glance, it didn't seem like an easy integration.... Would be worth taking a look again. Would be good if we could avoid making a new, distinct tool, just for Recovered Edits.

The bulkload tool expects hfiles in column family directories. Recovered edits files are not hfiles and the files are x-columnfamily so this is not the way to go though a bulkload-like tool that moved the recovered edits files under the appropriate region dir and asked the region reopen would be a possibility (Would need the bulk load complete trick of splitting input if the region boundaries in the live cluster do not align w/ those of the errant recovered edits files).

Issue Links

• relates to: HBASE-25159 [hbck2] Add an 'adoption service' for 'Orphaned Regions'
• links to: GitHub Pull Request #2504