Autoscaling Coordinator/historical node

[soumyajose0784]

We have deployed druid cluster in kubernetes using druid operator with autoscaling configured for broker & middlemanager. Would like to know if we have any directions for autoscaling other service pods(coordinator& historical)

[AdheipSingh]

here are few things to be noted for scaling druid nodes

1. MM - Make sure you are aware of the consequences of using MM with HPA
   KubernetesTaskRunner for running druid tasks as kubernetes jobs apache/druid#8801 (comment)
   KubernetesTaskRunner for running druid tasks as kubernetes jobs apache/druid#8801 (comment)

2. For brokers - https://druid.apache.org/docs/latest/operations/basic-cluster-tuning.html#number-of-brokers
   HPA can work for brokers in case you want to scale on mem/cpu or external metrics.

3. Coordinator - IMHO not needed.

4. Historicals - Since historicals holds segments. adding historicals means coordinator reshuffles segments, you can also scale historicals vertically. operator should support that once Fix76: Scale STS / PVC vertically to support resizing of druid nodes #97 this finalized