[CASSANDRA-15123] Avoid keeping sstables marked compacting forever when user defined compaction gets interrupted

• Type: Bug
• Status: Resolved
• Priority: Normal
• Resolution: Fixed
• Fix Version/s: 3.11.5, 4.0
• Component/s: Local/Compaction
• Bug Category: Degradation - Resource Management
• Severity: Normal
• Complexity: Low Hanging Fruit
• Discovered By: User Report
• Platform: All
• Since Version: 3.11.0

Source Control Link

https://github.com/apache/cassandra/commit/b773bc7ac51fedc07145017edaefa919fac25696

Test and Documentation Plan

adds unit test

• Assignee: Marcus Eriksson
• Reporter: Marcus Eriksson

Authors

Marcus Eriksson

Reviewers

Blake Eggleston

• Created: 10/May/19 12:14
• Updated: 07/Aug/19 08:28
• Resolved: 07/Aug/19 08:28

Description

When we have both repaired + unrepaired data on a node, we create multiple compaction tasks and run them serially. If one of those tasks gets interrupted or throws exception we will keep sstables in the other tasks as compacting forever.