[HBASE-15442] HBase Backup Phase 2: Potential data loss and or data duplication in incremental backup

• Type: Bug
• Status: Closed
• Priority: Critical
• Resolution: Fixed
• Labels: backup
• Assignee: Vladimir Rodionov
• Reporter: Vladimir Rodionov
• Created: 10/Mar/16 22:07
• Updated: 01/Jul/22 20:02
• Resolved: 16/Mar/16 01:42

Description

Suppose we have two tables T1 and T2

1. Create full backup T1 with backup id = B1
2. Create full backup T2 backupId = B2
3. New data arrived into file WAL1
4. Create incremental backup of T1 with backupId = B3
5. Create incremental backup of T2 with backupid = B4

The directory structure for backup site after this steps

BACKUP_ROOT/WALs/B3
BACKUP_ROOT/WALs/B4
BACKUP_ROOT/T1/B1
BACKUP_ROOT/T2/B2

File WAL1 may end up either in BACKUP_ROOT/WALs/B3 or in both:
BACKUP_ROOT/WALs/B3 and BACKUP_ROOT/WALs/B4 location. Both are bad: in first case we lose data for backup B4 in second case we have duplicate copies of WAL1

Issue Links

• is part of: HBASE-14123 HBase Backup/Restore Phase 2