Seems to be the only way to really run stuff on kube (and not on GCE), until kube has support for sticky node persistence. It's no coincidence that all of the kube examples use redis as a datastore
I'm really looking forward to what's coming next in the 2015 roadmap for both Mesos and Kubernetes. In particular, I'm excited about the resource model implementation, which will provide a richer context for scheduling pods.