Recommendations for production setups ¶
The getting started-documentation is a fast way of spinning up a Kubernetes cluster, but there are some aspects of kOps that require extra consideration. This document will highlight the most important things you should know about before deploying your production workload.
High availability ¶
Running only a single master can be error prone and disruptive.
Read through the high availability documentation to learn how to set up a cluster with redundant control plane.
The default networking of kOps, kubenet, is not recommended for production. Most importantly, it does not support network policies, nor does it support internal networking.
Read through the networking page and choose a stable CNI.
Private topology ¶
By default kOps will create clusters using public topology, where all nodes and the Kubernetes API are exposed on public Internet.
Cluster spec ¶
kops command allows you to configure some aspects of your cluster, but for almost any production cluster, you will want to change settings that are not accessible through the CLI. The cluster spec can be exported as a yaml file and checked into version control.
Read through the cluster spec page and familiarize yourself with the key options that kOps offers.
If your cluster contains multiple Instance Groups, or if you manage multiple clusters, you want to use generate the cluster spec using templates.
Read through the templating documentation to learn how to make use of templates.