K8s deployment: Number of Pods


When deploying on K8s using the autonomous operator, 2 extra Pods will get created apart from pods for the cluster nodes. Those 2 pods will be: 1 for the operator and 1 for the admission controller. Is the understanding correct?

What are the resources required for these 2 additional pods?

Does the documentation include these details?

If monitoring is configured, then a separate namespace will get created. How many pods will be created in the monitoring namespace?

Many thanks in advance.


Yes, there will be 2 pods (1 for the operator and 1 for the admission controller) when deploying on k8s using the operator. You can find the details on resources needed here: https://docs.couchbase.com/operator/current/concept-operator.html

If you enable monitoring, one monitoring pod will be created per node in the couchbase cluster deployed. You can refer to https://blog.couchbase.com/step-by-step-guide-for-running-couchbase-autonomous-operator-2-0-with-prometheus-part-2/ for more details


Thanks Arun

I had checked the first link and I could not find the resources required for the CAO & DAC pod in terms of CPU/memory. Am I missing something?

On monitoring front I figured out, that the Couchbase exporter will be a side-car container (not a separate container), 1 for each server node. So for each server node there will be 1 pod which will have 2 containers: 1 for the server couchbase node & other for the prometheus exporter.


The resources required are negligible, you can probably get away with 100m CPUs for both of them. Memory again is minimal for the DAC I’d say try with 128Mi. The Operator is slightly different in that it caches a lot of the resource it manages to avoid overloading the Kubernetes API, again I’d try 128Mi to start with and watch for restarts. Due to the caching the more resources you have in a namespace, the more memory will be used, so it’s hard to give a single answer.

Prometheus is indeed implemented with the sidecar model of deployment until Couchbase server supports Prometheus monitoring natively.