Hi team,
because most of the workload in spark 1.6 cluster has been moved to dataproc, we plan to reduce the number of nodes in both regular(scheduled) and backfill cluster. We'll also scale down the instance type for airflow worker since calling dataproc client should consume less resource than running spark-submit locally.
You can see the details in these tasks:
We plan to do this today at 17.00+ Jakarta time.
Let me know if you have any concern.