[CASSANDRA-18345] Enable streaming SAI components as part of repair

• Type: Improvement
• Status: Resolved
• Priority: Normal
• Resolution: Fixed
• Fix Version/s: 5.x
• Component/s: Feature/2i Index
• Labels: SAI
• Epic Link: CEP-7 Storage Attached Index (SAI)
• Change Category: Operability
• Complexity: Normal
• Platform: All

Source Control Link

https://github.com/apache/cassandra/commit/5d3f257477cba2d7f33f842dba4582d0660f5738

Test and Documentation Plan

New DTest included to check if SAI components are streamed properly and if querying the index still works after rebuilding the node. Two variants of the test: zeroCopy and nonZeroCopy.

• Assignee: Piotr Kolaczkowski
• Reporter: Mike Adamson

Authors

Piotr Kolaczkowski

Reviewers

Andres de la Peña, Caleb Rackliffe, Mike Adamson

• Created: 20/Mar/23 17:40
• Updated: 29/Jun/23 20:49
• Resolved: 29/Jun/23 20:49

Description

SAI registers it's components with the SSTable descriptor expecting them to form part of the SSTable lifecycle, including streaming. This is not the case.

The current SSTable format in Cassandra uses a fixed set of components (Components.STREAMING_COMPONENTS) when streaming SSTables. This needs to change to use the set of components that are registered with a specific SSTable by calls to SSTable.registerComponents.

Issue Links

• links to: GitHub Pull Request #2420