We are deploying Pega Platform 8.3 + Pega Marketing 8.3 by using Pega Kubernetes Helm Charts and Pega Container Images within a GKE cluster.
The helm charts use a value.yaml file to fill up various parameters among those are the replicas number per application layer (web, backend, stream, search, cassandra).
Our concern is after deployment should we decideto increase/descrease the number of replicas does it have an impact to the application or it can be done without issues?
I guess stateless deployments such as web, backend, treafik are not expected to have any impact but what about the stateful ones (cassandra, stream, search)? Can those have their replicas increased anytime or this is something that should applied only during initial deployment.
The Pega values.yaml gives you a starting point of your deployment. You can scale up/down as needed per your performance load tests - that is the big advantage of k8s deployment over traditional deployment where you would have to procure servers/VMs beforehand. This assumes of course you have enough resources in the pool - you are running on GKE, so that should not be any issues. Some deployments (e.g., pega-batch/pega-web) comes with horizontal autoscaler (although fairly primitive, only use cpu/memory as metrics) where k8s will scale up/down automatically based on the autoscaler metrics.