Current situation
We currently have a kola test for flannel. It starts a cluster of three machines using flannel, and verifies that etcd is healthy and that the machines can communicate with each other.

Impact
As we don't test the other CNIs, we rely on users reporting issues, which has meant breaking their production clusters more than once.

Ideal future situation

We should create similar test cases for other common CNIs:

• Calico
• Cilium
• AWS CNI

This issue shall track the work for all of them.