[CASSANDRA-19719] CEP-15: (Accord) When nodes are removed from a cluster, need to update topology tracking to avoid being blocked

• Type: Bug
• Status: Resolved
• Priority: Normal
• Resolution: Fixed
• Fix Version/s: NA
• Component/s: Accord
• Bug Category: Correctness - Unrecoverable Corruption / Loss
• Severity: Critical
• Complexity: Low Hanging Fruit
• Discovered By: Fuzz Test
• Platform: All
• Since Version: NA

Source Control Link

https://github.com/apache/cassandra/commit/09dd3d69af6755d6f68b6e0d019238f793f3246a

Test and Documentation Plan

new tests

• Assignee: David Capwell
• Reporter: David Capwell

Authors

David Capwell

Reviewers

Blake Eggleston

• Created: 18/Jun/24 20:54
• Updated: 01/Jul/24 20:18
• Resolved: 01/Jul/24 20:18

Description

When doing a host replacement or decom nodes will leave the cluster and won't ever come back, this can put accord's topology sync logic into a bad state as its going to wait forever for those nodes to acknowledge the epoch or parts of the ranges.

Issue Links

• links to: GH: accord trunk
• links to: GH: cep-15-accord