We strongly discourage using KIND in production because it isn't designed for or supported as such.
For example:
- There's no supported way to maintain patching and upgrades
- The technology fundamentally lacks things like functioning OOM metrics / various things related to resource restriction at the moment
- As a upstream minimal distro, no additional opinionated security configuration is implemented
KIND clusters are designed to be cheap (relatively), ephemeral, and relatively "standard".
They are not built to span multiple physical machines or as permanent long-lived clusters.
If you'd like to bootstrap a production-grade cluster, please see:
Which are all part of the Kubernetes project under SIG Cluster Lifecycle.