We are running Airflow on Kubernetes.

   1. What is the advantage of setting the replicas property to a value
   greater than 1 for the web server, scheduler and triggerrer?
   2. If this value is set to > 1, then multiple scheduler pods get
   launched. Is this done for more system throughput? How do the schedulers
   coordinate with each other as to which dag to schedule, etc?
   3. Same question for triggerer.
   4. If running in Kubernetes and the replicas are each set to 1, then K8
   will restart the process whenever any pod goes down for any reason. Thus a
   value > 1 is *not *strictly needed for availability. Please confirm.

Thanks.

Reply via email to