[CASSANDRA-17731] Clean up ScheduledExecutors, CommitLog, and MessagingService shutdown for in-JVM dtests

• Type: Bug
• Status: Resolved
• Priority: Normal
• Resolution: Fixed
• Fix Version/s: 4.0.6, 4.1-beta1, 5.0-alpha1, 5.0
• Component/s: Test/dtest/java
• Bug Category: Correctness - Test Failure
• Severity: Low
• Complexity: Low Hanging Fruit
• Discovered By: Code Inspection
• Platform: All
• Since Version: 4.0

Source Control Link

https://github.com/apache/cassandra/commit/d51f90201b2d127166612384a75b4a49da9776d1

Test and Documentation Plan

n/a (the patch itself is test fixes)

• Assignee: Caleb Rackliffe
• Reporter: Caleb Rackliffe

Authors

Caleb Rackliffe

Reviewers

David Capwell

• Created: 05/Jul/22 20:01
• Updated: 12/Sep/23 13:02
• Resolved: 14/Jul/22 16:56

Description

There appear to be two problems w/ the way we shut down ScheduledExecutors in Instance in 4.0+:

1.) We do it twice, Ince as part of a larger batch of shutdown activity, and then again in its own parallelRun() block.
2.) It happens before MessagingService shuts down, but some messaging-related threads (see StreamSession#closeSession()) can submit tasks to nonPeriodicTasks.

We should do it once, and do it after the MessagingService has properly shut down.

Issue Links

• fixes: CASSANDRA-17139 readWriteDuringBootstrapTest - org.apache.cassandra.distributed.test.ring.BootstrapTest
• is related to: CASSANDRA-16778 Jvm-dtest race on closing and log messages
• relates to: CASSANDRA-17139 readWriteDuringBootstrapTest - org.apache.cassandra.distributed.test.ring.BootstrapTest