[CASSANDRA-15817] Prevent repair from overrunning compaction

• Type: Bug
• Status: Resolved
• Priority: Low
• Resolution: Fixed
• Fix Version/s: 4.0, 4.0-beta2
• Component/s: Consistency/Repair
• Bug Category: Degradation - Resource Management
• Severity: Low
• Complexity: Low Hanging Fruit
• Discovered By: User Report
• Platform: All
• Since Version: 0.3

Source Control Link

https://github.com/apache/cassandra/commit/dfd0aebf73faa8f910e68ff5d50f91fae4f2669e

Test and Documentation Plan

Tested by hand; disabled by default (via very large value) and not included in cassandra.yaml or NEWS.txt

• Assignee: Jeff Jirsa
• Reporter: Jeff Jirsa

Authors

Jeff Jirsa

Reviewers

David Capwell, Marcus Eriksson

• Created: 16/May/20 03:33
• Updated: 21/Dec/20 09:31
• Resolved: 24/Aug/20 07:13

Description

Repair can run and stream faster than a host can compact. At some point, if a host is sufficiently out of sync, or compaction is especially expensive, it makes sense to intentionally block repair so that compaction can catch up