n-o-t-mugen commented on issue #39693: URL: https://github.com/apache/airflow/issues/39693#issuecomment-7014889083
I'd like to pick this up and get it over the line, building on @yssxr's analysis in #71452 (auto-closed as stale). I'll keep the credit. @Zim95, happy to coordinate if you've already started. Still reproduces on `main` (4f459eab70): `PodManager.await_pod_completion` only stops waiting early on base-container completion when `istio_enabled or do_xcom_push`, so any other long-running sidecar keeps the pod in `Running` and the task hangs. Proposed approach: - Stop waiting once the base container has terminated, whatever the sidecar is, and let the existing `on_finish_action` / cleanup path handle the pod, so `KEEP_POD` is still respected. - Make sure native Kubernetes sidecars (init containers with `restartPolicy: Always`) keep working, since they don't block pod completion. - Unit tests covering a generic sidecar, Istio, the XCom sidecar and a native sidecar. @jedcunningham @amoghrajesh, does this direction match the earlier review feedback on Istio / native sidecars? I'll open a PR shortly. -- 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]
