potiuk commented on PR #68788: URL: https://github.com/apache/airflow/pull/68788#issuecomment-5154210142
Fixing #39184 is worthwhile — tasks hanging indefinitely because no sidecar is injected to read `/airflow/xcom/return.json` is a bad failure mode, since there is nothing in the logs pointing at the cause. Two practical notes before I review the substance. This branch is 1047 commits behind `main` and currently conflicting, so it needs a rebase. More importantly, it overlaps with #68786 from you: both edit `xcom_sidecar.py` and `tests/.../test_xcom_sidecar.py`. That PR turns out to contain a genuine caller-mutation fix — `add_xcom_sidecar` was rebinding `pod_cp.spec.volumes` to the *original* pod's list, so the `insert` mutated the caller's pod despite the deepcopy — and it is a one-line change with tests around it. My suggestion is to land #68786 first and rebase this one on top. It is the smaller and independently correct change, and this PR's `PodDefaults` restructuring (`VOLUME_MOUNT_NAME`, `XCOM_SIDECAR_COMMAND`, the `ApiClient` import) is much easier to rebase onto that one line than the reverse. Doing it the other way round means resolving the same test file twice. Once the ordering is settled and this is rebased, I will review the SparkKubernetesOperator side properly. --- Drafted-by: Claude Code (Opus 5); reviewed by @potiuk before posting -- 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]
