Question Regarding Cluster API

[itamar-marom]

There is a problem with CRDs:

• They are very big and may get to a point where Kubernetes will not allow them (will reach limits like annotation limit)
• It is very hard to maintain such an object

In my opinion, we need to use Kubernetes' power and capabilities in this:

• Using ConfigMap and Secret objects for configuration (and adding a Ref to them in the CRD)
• Creating CRD for a Node and letting the cluster hold only Cluster Level configuration and state

For example in Druid, the DruidCluster should hold the external, the deploymentOrder.
We will have a DruidNode CRD that will hold refs to Configmaps for configuration (like appConfigGroup) and also the Kubernetes relevant spec.

We will need to watch more resources like ConfigMap but this is much more Kubernetes native