It's a cash strapped start-up. What are the worst consequences of running only three nodes, three schedulable masters?
Are there some ways can one mitigate against the risks?
DevOps Stack Exchange is a question and answer site for software engineers working on automated testing, continuous delivery, service integration and monitoring, and building SDLC infrastructure. It only takes a minute to sign up.Sign up to join this community
Best practices for replicating masters for HA clusters
Try to place master replicas in different zones. During a zone failure, all masters placed inside the zone will fail. To survive zone failure, also place nodes in multiple zones (see multiple-zones for details).
Do not use a cluster with two master replicas. Consensus on a two-replica cluster requires both replicas running when changing persistent state. As a result, both replicas are needed and a failure of any replica turns cluster into majority failure state. A two-replica cluster is thus inferior, in terms of HA, to a single replica cluster.
When you add a master replica, cluster state (etcd) is copied to a new instance. If the cluster is large, it may take a long time to duplicate its state. This operation may be sped up by migrating etcd data directory, as described here (we are considering adding support for etcd data dir migration in future).
I would like to emphasize the following quote
A two-replica cluster is thus inferior, in terms of HA, to a single replica cluster.
If less then 3 nodes will be used then it is basically more safe to run 1 master, but if one goes down there will be an issue as well. So it is better to run at least three masters.