Alex created AIRFLOW-6013:
-----------------------------
Summary: Last heartbeat check is too precise
Key: AIRFLOW-6013
URL: https://issues.apache.org/jira/browse/AIRFLOW-6013
Project: Apache Airflow
Issue Type: Bug
Components: worker
Affects Versions: 1.10.6
Reporter: Alex
Assuming the last hearbeat is not older than the heartbeat_time_limit, this
warning will Always fire:
[https://github.com/apache/airflow/blob/1.10.6/airflow/jobs/local_task_job.py#L120]
There's a few commands between:
[https://github.com/apache/airflow/blob/1.10.6/airflow/jobs/base_job.py#L195]
and
[https://github.com/apache/airflow/blob/1.10.6/airflow/jobs/local_task_job.py#L111]
so _(timezone.utcnow() - self.latest_heartbeat).total_seconds()_ will always be
some small but non-0 number.
We get many log warnings in our task-logs similar to:
{code:java}
WARNING - Time since last heartbeat(0.01 s) < heartrate(5.0 s), sleeping for
4.991735 s{code}
Does local_task_job need the extra check on last_heartbeat?
[https://github.com/apache/airflow/blob/1.10.6/airflow/jobs/local_task_job.py#L121]
Since base_job is already making sure to sleep through the gap:
[https://github.com/apache/airflow/blob/1.10.6/airflow/jobs/base_job.py#L187]
?
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
