Airflow times out on Long Running Kubernetes Jobs with logging gaps #26482
Replies: 6 comments 3 replies
-
Happy to raise this as a PR, I.e. changing |
Beta Was this translation helpful? Give feedback.
-
(note my version exp issue was 1.1.14 |
Beta Was this translation helpful? Give feedback.
-
Note, I tried this fix and it didn't work. The best thread I have found is this one: |
Beta Was this translation helpful? Give feedback.
-
Does this happens with latest Airflow version? |
Beta Was this translation helpful? Give feedback.
-
Converted it to discussion. We have revamped K8S integration since quite heavily and the log streaming part was specifically rewritten using more stable and newer k8s library so this issue is not relevant any more most likely (we can always reopen if it is in the latest version - so please upgrade and double-check @alittlesliceoftom |
Beta Was this translation helpful? Give feedback.
-
Hi @potiuk , still happening with Airflow |
Beta Was this translation helpful? Give feedback.
-
This is listed too in other locations:
https://issues.apache.org/jira/browse/AIRFLOW-3534
https://issues.apache.org/jira/browse/AIRFLOW-4526
Only known fix is to disable log stream:
When disabling the
get_logsparameter everything works because the log stream is skipped.
Suggested fix - up the retry count on the below line, or allow this to be configured.
airflow/airflow/kubernetes/pod_launcher.py
Line 198 in b23fc13
Beta Was this translation helpful? Give feedback.
All reactions