kagent monitors the services running on local host by periodically executing 'systemct status SERVICE'. There are certain services such as 'airflow-scheduler' which by design restarts. kagent reports to Hopsworks that is dead. As a consequence, it pollutes the logs but most importantly it fills up the 'alerts' table in hopsworks database with false-positives.
Instead of white-listing a service, which can be dangerous, kagent should identify a service dead N times before sending an alert to Hopsworks.