Druid In GKE

Hello All

I have put Druid Co coordinator and overlord in Google Kubernetes Engine using internal load balancer. When I up scale Mater node. I get response as below

Unknown exception / Unexpected response status [307] description [Temporary Redirect] from request url [http://xxxxxxxxx:8081/druid/coordinator/v1/metadata/segments?includeOvershadowedStatus] /

org.apache.druid.java.util.common.RE

where xxxxxxxx is the load balancer ip address.

Is there any way to avoid redirection ?.

Thanks,

Ashish

Hi,

I don’t think you can avoid it, as coordinator (like overlord) in HA behave this way (one master, many followers that redirect to master)

Here is a similar issue I opened on overlord behaviour : https://github.com/apache/incubator-druid/issues/8630

Le mer. 27 nov. 2019 à 21:52, Andriod Guru ashishn87@gmail.com a écrit :

Any work around?

Thanks

Ashish

Ashish,

Load balancers(round robin, etc.) and Leader elections (ZK based) have different behaviors. It is not necessary to use load balancers for the master services of Druid as it uses ZK to manage the state of leaders and followers. Putting a load balancer in front of them all only introduce erratic behavior and the cluster will no longer be in a consistent state.

Rommel Garcia

Thanks. I am going now with statefulsets and vertical load balancing to see if this is viable option.

Thanks

Ashish