There is no limitation while you're provisioning Kubernetes cluster(s) for your development purposes. We're lucky enough to have such a diversified community where offers all kinds of cool tools and well-written tutorials.
So this is really subject to your preference, requirements and resource constraints.
For me, I have set some objectives while thinking the "best way" to have a local Kubernetes cluster for learning, development right in my laptop.
In short, the Kubernetes cluster provisioned:
- MUST be multiple "nodes", e.g. 1 Masters + 3 Workers;
- MUST be easily accessible to each component for necessary tweaking;
- MUST consume bare minimum resources -- I do care about my laptop's battery life;
- MUST be embedded with best practices, as more as possible.
The potential ways might keep growing/evolving along the days.
But as of now, below are what I've tried and considered "great".
| Ranking | Approach | Docs |
|---|---|---|
| 1 | k3s + k3d | Refer to README-K3D-K3S.md |
| 2 | Kubernetes in Docker, aka kind | Refer to README-KIND.md |
| 3 | Vagrant + VirtualBox | Refer to README-VAGRANT.md |
If you like the OKD/OpenShift approach, instead of the upstream K8s, you may try these:
- OKD in Docker, refer to README-OKD-IN-DOCKER.md
- Or MiniShift, which you can find tutorial everywhere.
Note: all the experiments were on Mac, but working in a Linux env should be very similar.
Enjoy!