[HBASE-22784] OldWALs not cleared in a replication slave cluster (cyclic replication bw 2 clusters)

• Type: Bug
• Status: Resolved
• Priority: Blocker
• Resolution: Fixed
• Affects Version/s: 1.4.9, 1.4.10
• Fix Version/s: 1.5.0, 1.4.11
• Component/s: regionserver, Replication
• Hadoop Flags: Reviewed
• Assignee: Wellington Chevreuil
• Reporter: Solvannan R M
• Created: 02/Aug/19 16:15
• Updated: 23/Oct/19 12:03
• Resolved: 14/Oct/19 03:38

Description

When a cluster is passive (receiving edits only via replication) in a cyclic replication setup of 2 clusters, OldWALs size keeps on growing. On analysing, we observed the following behaviour.

1. New entry is added to WAL (Edit replicated from other cluster).
2. ReplicationSourceWALReaderThread(RSWALRT) reads and applies the configured filters (due to cyclic replication setup, ClusterMarkingEntryFilter discards new entry from other cluster).
3. Entry is null, RSWALRT neither updates the batch stats (WALEntryBatch.lastWalPosition) nor puts it in the entryBatchQueue.
4. ReplicationSource thread is blocked in entryBachQueue.take().
5. So ReplicationSource#updateLogPosition has never invoked and WAL file is never cleared from ReplicationQueue.
6. Hence LogCleaner on the master, doesn't deletes the oldWAL files from hadoop.

NOTE: When a new edit is added via hbase-client, ReplicationSource thread process and clears the oldWAL files from replication queues and hence master cleans up the WALs

Please provide us a solution

 

Issue Links

• is related to: HBASE-23169 Random region server aborts while clearing Old Wals
• relates to: HBASE-23205 Correctly update the position of WALs currently being replicated.