Any reason to use a load balancer for coordinators and overlords? It is typically recommended to use a load balancer on the brokers and to configure zookeeper for HA.
I don’t know if this solution would work, but that may.
You could do the trick with your probes (maybe the readiness one) and a custom script.
This script would check the status of the pod (leader,follower) and return true only in case of leader.
This way, no matter how you expose your pods, the real target will be determined by the deployment (it uses deployment, right ?) And no traffic should go to the follower (as it is “not ready” in k8s sense of meaning)
Hey! Sorry I’m late to the party… could you not use the Druid Router?
The Router can be configured to forward requests to the active Coordinator or Overlord process.
Since which Druid version is this router stuff available ?
I’m still in 0.15.1 and either it is not available on this one or I really missed something ^^
Aha! Yeah I think it was there…
Never tried this… but does it not work if you set it as a management proxy and then use the prefixes?