Currently when I deploy database services in k8s I want to control the zones that masters and tservers run in.
I divide my statefulsets in separately zones where each stateful set has a zone label, a scheduling restriction for the zone, and a storage class restricted to that zone.
Despite having 3+ statefulsets, they all use the same master nodes.
What might be needed to support this is to have the cluster CRD allow tserver and master node configurations to be specified multiple times, once for each zone, with slightly different options (e.g. storage class, labels, names, location command line parameter) and the operator would still provide the appropriate parameters for the nodes to discover one another.
Currently when I deploy database services in k8s I want to control the zones that masters and tservers run in.
I divide my statefulsets in separately zones where each stateful set has a zone label, a scheduling restriction for the zone, and a storage class restricted to that zone.
Despite having 3+ statefulsets, they all use the same master nodes.
What might be needed to support this is to have the cluster CRD allow tserver and master node configurations to be specified multiple times, once for each zone, with slightly different options (e.g. storage class, labels, names, location command line parameter) and the operator would still provide the appropriate parameters for the nodes to discover one another.