wellkilo commented on issue #939: URL: https://github.com/apache/flink-agents/issues/939#issuecomment-5868378993
I'd like to take this on and implement the underlying runtime fix, if nobody is actively working on it. I checked current `main` and the merged #1104 safeguard. The intended scope is: 1. Detect the actual batch keyed-state backend inside `ActionExecutionOperator` at runtime, so explicit `BATCH` and `AUTOMATIC` modes are handled according to the backend Flink actually created. 2. When that backend is active, finish the current input's action chain before `processElement` returns. This keeps all keyed state for one key alive until the batch backend's key-complete contract is satisfied. 3. Preserve the existing cross-key mailbox concurrency for streaming/general-purpose keyed backends. 4. Replace the graph-construction rejection from #1104 with a real multi-key BATCH regression test using the default batch state backend, plus a control that pins streaming concurrency. The public `CompileUtils.connectToAgent` signatures and event contracts would remain unchanged. The externally visible change is that the currently rejected configuration completes normally without dropping records. @weiqingy, I noticed your earlier "glad to take this on" comment. Please let me know if you already have an active implementation; otherwise I'll proceed with the above focused scope. -- 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]
