[CASSANDRA-17168] Don't block gossip when clearing snapshots for failing repairs

• Type: Bug
• Status: Resolved
• Priority: Normal
• Resolution: Fixed
• Fix Version/s: 4.0.2, 4.1-alpha1, 4.1
• Component/s: Consistency/Repair
• Bug Category: Availability - Unavailable
• Severity: Normal
• Complexity: Normal
• Discovered By: Adhoc Test
• Platform: All
• Since Version: 4.0

Source Control Link

https://github.com/apache/cassandra/commit/98e798f567368f826fc3a57ddb6cdc464e741fe3

Test and Documentation Plan

tests

• Assignee: Marcus Eriksson
• Reporter: Marcus Eriksson

Authors

Marcus Eriksson

Reviewers

David Capwell

• Created: 24/Nov/21 08:08
• Updated: 27/May/22 19:25
• Resolved: 17/Jan/22 08:21

Description

We clear snapshots in the GossipTasks thread when a repair session fails due to a replica shutting down. If there are many tables/repair sessions ongoing this can take a long time. With enough tables being repaired at the same time even checking if the snapshots exists can take long enough to mark nodes down.

We should clear snapshots in a separate thread and add a flag to tell us whether this repair session can have snapshots to avoid checking if the directory exists.