gardnervickers commented on pull request #7929: URL: https://github.com/apache/kafka/pull/7929#issuecomment-640691697
In 6152847, I changed the approach taken in this PR to move away from managing the producer state snapshot file as part of the segment lifecycle. This was done for two reasons, first it turned out to be a bit difficult to reason about producer state snapshot files which did not have a corresponding segment file (such as the snapshot taken during shutdown). Second, moving the management of deletion to the LogSegment ended up splitting ownership of the producer state snapshot file between the `LogSegment` and the `ProducerStateManager`, which seemed to complicate when things like truncation could happen. Instead, a call is made to the `ProducerStateManager` to delete any matching producer snapshot files when `LogSegment`'s are deleted. This occurs during retention and as a result of log cleaning (compaction). For producer state snapshot files which do not have a corresponding segment, they will be cleaned up on restart (unless they are the latest snapshot file). I think this ends up being simpler while more closely matching the current set of responsibilities, where the `ProducerStateManager` "owns" creating/reading/deleting snapshot files, but I'd be interested in hearing other opinions 😄. cc/ @junrao ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org
