What happens if we use an external etcd and have random names for the sensu servers. Do they need to talk to each other? Is there some autocleanup to keep the sensu server list clean from old servers?
1 Like
Hey @runningman84 
sorry we missed this one. We would expect this to work just fine. In K8s I have generally used a StatefulSet for the sensu-backend (with ordinal hostnames like sensu-backend-0, sensu-backend-1, and sensu-backend-2), which I frequently scale up and down. I haven’t tried using a Deployment or other controller (or the equivalent concepts in an IaaS cloud), but I would expect it should work. The benefit of separating etcd clustering from sensu-backends (by disabling Sensu’s embedded etcd and operating etcd externally) is that we would expect the cluster membership and leader election to work the same with less predictable sensu-backend hostnames – and you’re not limited to cluster member sizing based on quorum (e.g. you don’t need to have exactly 3-nodes, 5-nodes, or 7-nodes of sensu-backend like you do with etcd).
I hope this helps! Let us know how things go as you give this a try. In fact, I’d consider any experience and/or configuration details you could share to be a worthy contribution you could submit here for some free SWAG: Sensu | Contribute to Sensu. Get rewarded. 
1 Like