In combination with the cloud-controller-manager that handles the cloudy stuff, the networking should be a plain CNI bridge; set by a DaemonSet that drops the following file in /etc/cni/net.d on every node.
Since the kubenet->plain CNI stuff is proceeding, I want to test out how close we are using CNI-only (without the kubenet bells and whistles) on a cloud env like GCE.
I think the CNI spec will look something like this:
In combination with the cloud-controller-manager that handles the cloudy stuff, the networking should be a plain CNI bridge; set by a DaemonSet that drops the following file in
/etc/cni/net.d
on every node.Since the kubenet->plain CNI stuff is proceeding, I want to test out how close we are using CNI-only (without the kubenet bells and whistles) on a cloud env like GCE.
I think the CNI spec will look something like this:
(shamelessly inspired from @jbeda's project here: https://github.com/jbeda/kubeadm-gce-tf/blob/master/tf-scripts/prereq.sh#L46 :smile:)