Dieter De Paepe created HBASE-28715:
---------------------------------------
Summary: Backup delete incorrectly handles bulk loaded entries
Key: HBASE-28715
URL: https://issues.apache.org/jira/browse/HBASE-28715
Project: HBase
Issue Type: Bug
Components: backup&restore
Affects Versions: 2.6.0, 3.0.0, 4.0.0-alpha-1
Reporter: Dieter De Paepe
To delete a backup, `BackupAdminImpl#deleteBackup` calls
`BackupSystemTable#deleteBulkLoadedRows` on all rows returned by
`BackupSystemTable#readBulkLoadedFiles(backupId)`.
This last method comes down to a scan over the bulk load table for all rows
matching "bulk:BACKUPID:".
However, current code never writes rows of that format. The only methods that
do so are `BackupSystemTable#writeBulkLoadedFiles` and
`BackupSystemTable#createPutForBulkLoadedFile`, which are only used in test
code.
Note that the bulkload rows that do get written (see `BackupObserver`) only get
deleted when creating incremental backups, but suffer from HBASE-28706.
Consequences: I suspect that if you do some bulk loads, create a full backup,
and delete that backup before doing any incremental backup, those entries in
the bulk load table will never get deleted, meaning those HFiles can never get
deleted (because `BackupLogCleaner` blocks their deletion).
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
