We currently have three Zero nodes and three Alpha groups, each group consisting of three Alpha nodes. In our development environment, we want to scale down one of the Alpha groups from 3 nodes to 1.
However, simply reducing the replicas count in the StatefulSet doesn’t work as expected, since Kubernetes is unaware of Dgraph’s group concept. The nodes removed are chosen randomly, which doesn’t allow us to control which Alpha nodes are removed from which group.
Do you have any recommended approach or best practices for scaling down Alpha groups in this scenario?