passionworkeer commented on issue #59626:
URL: https://github.com/apache/airflow/issues/59626#issuecomment-4038905695
This is a well-documented issue. The proposed solution of adding internal
retry logic in PodManager.read_pod() is the right approach.
For the implementation, consider:
1. **In pod_manager.py**, modify ead_pod() to catch ApiException with
status 404 and retry:
`python
import time
from kubernetes.client.rest import ApiException
def read_pod(self, name, namespace, retries=3):
for attempt in range(retries):
try:
return self.core_v1.read_namespaced_pod(name, namespace)
except ApiException as e:
if e.status == 404 and attempt < retries - 1:
time.sleep(2 ** attempt) # exponential backoff
continue
raise
`
2. **Detect preemption** - check for eason=Preempted in pod events before
failing
3. **Use pod.metadata.resource_version** to ensure stale reads don't cause
issues
This way the operator handles transient preemption automatically without
requiring users to set high task-level retries.
--
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]
