Create a horizontal autoscaling

[Rub21]

cc. @batpad

[batpad]

@Rub21 can you clarify if you mean this should be for the cluster itself? i.e. whether we should make it so that the nodes of the cluster auto-scale, or just the pods inside the cluster auto-scaling? I think we probably do want both, just want to clarify the intention of this ticket.

[Rub21]

@batpad I think to scale horizontally the node is not possible, because the /var/lib/postgresql/data should be read by only one engine, but I really am not sure if that is right...

On the other side we could not scale the db-statefulset pod, statefulset is a pod which is goin to play only one in the cluster. I think the web pod is the only one to scale.

[batpad]

@Rub21 that is correct - we only ever want 1 instance of the db-statefulset pod.

It would be nice to have some sort of auto-scaling for the web pods. My question, thought, was around if we need to figure out auto-scaling for the Kubernetes cluster itself? Like we may start off with 2 t2-medium machines, but as the project / usage grows, want to grow to say 3 c3.large machines (or equivalent in GCP) - I'm not sure the best way to automate that based on resource usage.

Probably not a huge concern for now, but something to look at for the future.

[spwoodcock]

This is pretty easy to achieve using either default Horizontal Pod Autoscaler, or Keda:

• HPA (https://kubernetes.io/docs/tasks/run-application/horizontal-pod-autoscale) if we only need to scale based on cpu/memory usage.
• KEDA (https://keda.sh) for scaling based on any other event, like network traffic etc.