Ansible has been better than Chef and Puppet for small environments. I looked at cdist but it wasn't faster for my use case. Also, Ansible executes rules ordered unlike Chef and Puppet which helps reduce your state space. If you incrementally maintain servers, by definition, your only tested configuration is the one you executed. The way to improve reliability is to start from the same (container) state, and now the only maintenance ought to be changes (OS upgrades has been it for me). Ansible across 1 server and 2 desktops with no changes takes ~3m17s and I wish it was way faster. As of know I manage with tagging things and only running a subset of things. Consider standardizing on a single distro (I use Debian and it's served me extremely well over the years). +1 on centralizing, too, till you have an use case that requires more servers. servers <= containers. Simplify. Kubernetes is complicated.