[SPARK-48586] Remove lock acquisition in doMaintenance() by making a deep copy of file mappings in RocksDBFileManager in load()

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.4.3
• Fix Version/s: 4.0.0, 3.5.2
• Component/s: Structured Streaming
• Labels: pull-request-available
• Assignee: Riya Verma
• Reporter: Riya Verma
• Created: 11/Jun/24 17:39
• Updated: 28/Jun/24 08:00
• Resolved: 27/Jun/24 08:53

Description

Currently the lock of the RocksDB state store is acquired when uploading the snapshot inside maintenance tasks when change log checkpointing is enabled, which causes lock contention between query processing tasks and state maintenance thread. We want to eliminate lock contention to decrease latency of streaming queries so lock acquisition inside maintenance tasks should be avoided. This can introduce race conditions between task and maintenance threads. By making a deep copy of versionToRocksDBFiles and localFilesToDfsFiles in RocksDBFileManager, we can ensure that the file manager state is not updated by task thread when background snapshot uploading tasks attempt to upload a snapshot.

Issue Links

• links to: GitHub Pull Request #46942