[CASSANDRA-18044] Make CommitLogReplayer replay all tables in a keyspace when only keyspace is specified in cassandra.replayList property

• Type: Improvement
• Status: Resolved
• Priority: Normal
• Resolution: Fixed
• Fix Version/s: 5.0-alpha1, 5.0
• Component/s: Local/Commit Log
• Change Category: Operability
• Complexity: Normal
• Platform: All

Source Control Link

https://github.com/apache/cassandra/commit/3775c8f11dc998ba9c74769a14b9443272c750c6

Test and Documentation Plan

junit

• Assignee: Stefan Miklosovic
• Reporter: Stefan Miklosovic

Authors

Stefan Miklosovic

Reviewers

Branimir Lambov

• Created: 14/Nov/22 15:46
• Updated: 12/Sep/23 13:02
• Resolved: 24/Nov/22 18:24

Description

CommitLogReplayer reacts on a property "cassandra.replayList" (1) which replays only some tables if needed upon commit log replay.

However, imagine you have a keyspace of 50 tables and other keyspace of 20 tables and you want to replay only all tables of the first keyspace.

Currently, an operator has to enumerate all 50 tables which is not desirable and it is error prone.

I suggest to change the logic in such a way that if there is only keyspace listed, we would retrieve all its tables dynamically in runtime and replayed only these.

The solution would be receptive to both keyspace.table as well as only "keyspace" format.

Valid examples of such property:

ks1,ks2,ks3.tb1 // it will retrieve all tables of ks1 and ks2 and appends ks3.tb1
ks1.tb1,ks2.tb2
ks1,ks2,ks3 // it will retrieve tables of all these keyspaces

https://github.com/apache/cassandra/blob/trunk/src/java/org/apache/cassandra/db/commitlog/CommitLogReplayer.java#L379-L387