[CASSANDRA-19972] Nodetool cms reconfigure reports success when the operation fails due to streaming issues

• Type: Bug
• Status: Resolved
• Priority: Normal
• Resolution: Fixed
• Fix Version/s: 5.x
• Component/s: Transactional Cluster Metadata
• Bug Category: Correctness - Transient Incorrect Response
• Severity: Normal
• Complexity: Normal
• Discovered By: DTest
• Platform: All
• Since Version: NA

Source Control Link

https://github.com/apache/cassandra/commit/a27e09930dea92574edd48a0c7a6098af4e7c081

Test and Documentation Plan

New in-jvm dtest. CI to follow

• Assignee: Sam Tunnicliffe
• Reporter: Sam Tunnicliffe

Authors

Sam Tunnicliffe

Reviewers

Marcus Eriksson

• Created: 01/Oct/24 15:07
• Updated: 01/Nov/24 18:42
• Resolved: 01/Nov/24 18:42

Description

If streaming fails during CMS reconfiguration, the operation enters a blocked state. The operation will still be pending and resumable by an operator & the KPI that indicates the CMS is pending/due reconfiguration will correctly indicate the state. However, the nodetool command itself exits with a success status which is especially problematic in tests. 
This is easy to repro in an in-jvm dtest by attempting to a CMS reconfiguration without enabling the NETWORK feature on the cluster.

Issue Links

• links to: Trunk PR