[CASSANDRA-19709] Always repair the full range when repairing system_cluster_metadata

• Type: Improvement
• Status: Open
• Priority: Normal
• Resolution: Unresolved
• Fix Version/s: 5.1-alpha1
• Component/s: Transactional Cluster Metadata
• Change Category: Operability
• Complexity: Low Hanging Fruit
• Platform: All
• Assignee: Sam Tunnicliffe
• Reporter: Marcus Eriksson

Authors

Sam Tunnicliffe

Reviewers

Marcus Eriksson

• Created: 17/Jun/24 12:54
• Updated: 17/Jun/24 15:10

Description

Since system_cluster_metadata uses a custom partitioner we need to repair the full range whenever an operator uses -st and -et tokens to avoid breaking existing repair automations. This is fine since the amount of data in this keyspace should be small.

https://github.com/krummas/cassandra/commit/6ab8fcd1126f67b0117e1ee3c1fd1d4b40ac2362