1996fanrui opened a new pull request, #28662: URL: https://github.com/apache/flink/pull/28662
This PR depends on https://github.com/apache/flink/pull/28651, https://github.com/apache/flink/pull/28652, https://github.com/apache/flink/pull/28659, https://github.com/apache/flink/pull/28660 and https://github.com/apache/flink/pull/28661 Please only review the last 4 commits (prefixed [FLINK-40080]). ## What is the purpose of the change [FLINK-40080] Drain spilled state into channels; replay spilled slices into checkpoints. It's part 6 of FLINK-38544 (checkpointing during recovery of unaligned checkpoint state) — the disk-side counterpart of the recovery-checkpoint coordination. The trigger contract is widened so checkpoint start atomically snapshots the not-yet-drained spill slice, the channel-state writer learns to replay such a slice into a checkpoint (`addInputDataFromSpill` / `writeInputFromSpill`), the barrier-handler dispatch gains the reader-handover step, and the incremental `FetchedChannelStateDrainer` lands. Runtime behavior is unchanged — no production path constructs a drainer, and the handler factory does not select the spilling handlers. ## Brief change log - [FLINK-40080][checkpoint] Widen RecoveryCheckpointTrigger#snapshotAndInsertBarriers to return a state reader - [FLINK-40080][checkpoint] ChannelStateWriter#addInputDataFromSpill: replay spilled segments into the checkpoint - [FLINK-40080][checkpoint] ChannelState step 3: hand the snapshot reader to the writer - [FLINK-40080][checkpoint] FetchedChannelStateDrainer: incremental drain with atomic snapshot-and-insert ## Verifying this change - New unit tests: FetchedChannelStateDrainerTest, FetchedChannelStateDrainerConcurrencyTest, ChannelStateWriterImplAddInputDataFromSpillTest, ChannelStateDispatcherTest (extended to the full 3-step dispatch) ## Does this pull request potentially affect one of the following parts: - Dependencies (does it add or upgrade a dependency): no - The public API, i.e., is any changed class annotated with `@Public(Evolving)`: no - The serializers: no - The runtime per-record code paths (performance sensitive): no - Anything that affects deployment or recovery: JobManager (and its components), Checkpointing, Kubernetes/Yarn, ZooKeeper: no - The S3 file system connector: no ## Documentation - Does this pull request introduce a new feature? no -- 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. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
