good first issuekind:feature
仓库指标
- 星标
- (44,809 个星标)
- PR 合并指标
- (PR 指标待抓取)
描述
Description
Current liveness check probes use the 'airflow jobs' command which directly queries the backend DB as opposed to actually querying an endpoint or checking the status of the process itself.
e.g. Triggerer liveness probe
exec [sh -c CONNECTION_CHECK_MAX_COUNT=0 AIRFLOW__LOGGING__LOGGING_LEVEL=ERROR exec /entrypoint \
airflow jobs check --job-type TriggererJob --hostname $(hostname)] delay=10s timeout=20s period=60s #success=1 #failure=5
This command only checks the backend DB to see if there are any jobs. Additionally, the exit code is always 0 regardless of how many jobs there are. Ideally, the liveness check is done by querying some endpoint on the triggerer to see if it's still running.
Use case/motivation
Would like a liveness check that is more aware of the process rather than the stored state
Related issues
No response
Are you willing to submit a PR?
- Yes I am willing to submit a PR!
Code of Conduct
- I agree to follow this project's Code of Conduct