amoghrajesh commented on code in PR #67715:
URL: https://github.com/apache/airflow/pull/67715#discussion_r3345853542
##########
providers/apache/spark/src/airflow/providers/apache/spark/hooks/spark_submit.py:
##########
@@ -802,6 +862,95 @@ def _start_driver_status_tracking(self) -> None:
f"returncode = {returncode}"
)
+ def _poll_k8s_driver_via_api(self) -> None:
+ """Poll the K8s driver pod phase until it reaches a terminal state."""
+ pod_name = self._kubernetes_driver_pod
+ namespace = self._connection["namespace"]
+ app_id = self._kubernetes_application_id or pod_name
+
+ if not pod_name:
+ raise ValueError("K8s driver pod name not set; cannot poll
status.")
+
+ client = kube_client.get_kube_client()
+ poll_interval = max(self._status_poll_interval, 20)
+ if poll_interval != self._status_poll_interval:
+ self.log.info(
+ "status_poll_interval=%ds is below the 20s minimum for K8s API
polling; using 20s.",
+ self._status_poll_interval,
+ )
+ # Mirror `missed_job_status_reports` / `max_missed_job_status_reports`
from
+ # `_start_driver_status_tracking`: tolerate transient failures before
giving up.
+ consecutive_unknown = 0
+ max_consecutive_unknown = 3
+ consecutive_api_errors = 0
+ max_consecutive_api_errors = 3
+ consecutive_pending = 0
+ pending_warn_threshold = 10
+
+ try:
+ while True:
+ try:
+ pod = client.read_namespaced_pod(pod_name, namespace)
+ consecutive_api_errors = 0
+ except kube_client.ApiException as e:
+ consecutive_api_errors += 1
+ self.log.warning(
+ "ApiException polling pod %s (%d/%d): %s",
+ pod_name,
+ consecutive_api_errors,
+ max_consecutive_api_errors,
+ e,
+ )
+ if consecutive_api_errors >= max_consecutive_api_errors:
+ raise RuntimeError(
+ f"K8s API unreachable after
{consecutive_api_errors} consecutive errors "
+ f"while polling {app_id}; giving up."
+ ) from e
+ time.sleep(poll_interval)
+ continue
+
+ phase = pod.status.phase or "Initializing"
Review Comment:
Good point. Added a note to the requirements section calling this out and
pointing to `execution_timeout` as the mitigation. I agree that matching by
container name can be a follow up
--
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]
