I’m continuing to hear rumblings about applications requiring their own “dedicated Kubernetes clusters” and thinking to myself, “Didn’t we go through this with Java JVM and Eclipse ‘runtime’ tech?” If an application requires it’s own cluster then what is the point of Kubernetes in the first place? Vault suggests exactly this. I wonder if this is because of the scope and criticality of Vault. If some other controller goes bad then all the corporate secrets in Vault would be up for grabs, but still. The big take away here is that if Kubernetes is truly going to continue to take off (as it has) that multi-cluster is definitely the future. Getting clusters to talk to each other and considering them more like individual “servers” in the old days with “processes” on them is very much the preferred paradigm for IT architecture. There’s going to be a ton of cluster-to-cluster communication stuff that will emerge in the next few years.
#k8s #architecture #multicluster