This message was deleted.

You can checkout https://github.com/datainfrahq/druid-operator

Thanks @Adheip Singh I did check earlier the repo which you have shared then only I thought of asking this question here... Could you please help with 1st and 2nd question if possible?

• it is advisable and preferred •

outside of k8s cluster

- can you elaborate a bit here.

IMO : Depends on the load that you're wanting to deal with. We're currently using the Druid helm chart and terraform for a K8s deployment. But we're also hosting our services differently and looking to explore auto scale for increased load. But if you dont have a very significant load I think deployment via K8s definitely adds significant complexity, which aren't really there with single server deployments. If your memory requirements are telling you that you need to host services separately/you think you'll have to scale pretty quickly K8s would be a best bet. For a POC or a small load service I think k8s would be overkill.

@Adheip Singh Outside K8s means like independent server in clustered mode, without k8s, just like it's mentioned in Druid official doc here

for onprem deployments we deploy druid using ansible to automate installation, patching, and upgrades. On GCP we use GKE and the druid operator above to deploy druid clusters at multi-petabyte scale. Both work, though id advise if you are going to use a k8s deployment for deploying druid onprem at a pretty large scale, k8s/Storage Arrays/CNI's etc.... can become a bottleneck with regards to performance and stability if the underlying infrastructure isnt designed well. Hence why we only use the k8s operator on managed cloud infra (GKE/EKS)