I'm attempting to setup an experimental kubernetes cluster to see whether it is worth it to migrate our current mesos cluster to kubernetes. Since we are providing our own CoreOS servers I'm attempting a bare metal intall. The problem is that the containers that kubernetes starts do not have access to the network (example), while containers started manually do indeed have working network access (example).
For the network I chose calico and followed the instructions here. I'm pretty confident that the calico setup is correct since I am able to ping containers across hosts, just not when they were started by Kubernetes.
The pause container of the pod gets an address assigned, as can be seen here, however from inside the container the network is not reachable and the other containers in the same pod show now network configuration doing docker inspect (which I assume is correct since they share the pause-containers network).
Does anybody have an idea on what the issue might be, or how to debug this further?
0 Answers