Kubernetes refuses to work with rkt, containers not starting, wrong systemd error

I have hit a wall trying to setup kubernetes with rkt on coreos when deploying the master node.

This is the main problem I see from the kubelet logs...some systemd error which is obviously wrong because coreos latest stable uses 225...

I have followed the Deploy Kubernetes Master Node(s) and the Running Kubernetes on rkt guides.

I have tried all the hyperkube images from quay.io and the rkt versions 1.0.0,1.2.1, 1.5.1 making sure to switch the correct fly aci image in the kubelet-wrapper...no changes whatsoever.

for who missed the link to the log: http://pastebin.com/raw/thLd0r2R (it was the link on "this" which I bolded now)

Anyway seems like the kubelet is still not ready to run containerized, so I managed to bootstrap a setup with the kubelet ran from binary, still using rkt for scheduler apiserverand controller, and using a second wrapper for the proxy because the proxy needs stage1-fly.

So it is composed of:

• systemd unit for rkt-api /usr/bin/rkt run apiserver ...
• systemd unit for the proxy /usr/bin/rkt run coreos/hyperkube proxy...--stage1-path=/usr/share/rkt/stage1-fly.aci ...
• systemd unit for the kubelet /opt/bin/kubelet ...
• k8s manifests for apiserver, scheduler, controller

Sadly there is the bug that the hyperkube does not catch the kubectl flags so I also have to use the kubectl binary besides the kubelet binary and the containerized hyperkube bin. Eventually the hyperkube will be all that is needed with only 1 systemd unit and the manifests.