I am deploying Kubernetes 1.4 on Ubuntu 16 on Raspberry Pi 3 following the instructions at http://kubernetes.io/docs/getting-started-guides/kubeadm/. The master starts and the minion joins no problem but when I add weave kubedns won't start. Here's the pods:
k8s@k8s-master:~$ kubectl get pods --all-namespaces
NAMESPACE NAME READY STATUS RESTARTS AGE
kube-system etcd-k8s-master 1/1 Running 1 23h
kube-system kube-apiserver-k8s-master 1/1 Running 3 23h
kube-system kube-controller-manager-k8s-master 1/1 Running 1 23h
kube-system kube-discovery-1943570393-ci2m9 1/1 Running 1 23h
kube-system kube-dns-4291873140-ia4y8 0/3 ContainerCreating 0 23h
kube-system kube-proxy-arm-nfvvy 1/1 Running 0 1h
kube-system kube-proxy-arm-tcnta 1/1 Running 1 23h
kube-system kube-scheduler-k8s-master 1/1 Running 1 23h
kube-system weave-net-4gqd1 0/2 CrashLoopBackOff 54 1h
kube-system weave-net-l758i 0/2 CrashLoopBackOff 44 1h
The events log doesn't show anything. getting logs for kube-dns doesn't get anything either.
What can I do to debug?
UPDATE: As Bryan pointed out, Flannel is not the only overlay network anymore.
Note this two hints in the kubeadm install documentation:
Flannel is the only network overlay support for arm
If you are on another architecture than amd64, you should use the flannel overlay network as described in the multi-platform section
When using Flannel, you need to make a kubectl init --por-network-cidr=10.244.0.0/16
Note: this will autodetect the network interface to advertise the
master on as the interface with the default gateway. If you want to
use a different interface, specify
--api-advertise-addresses= argument to kubeadm init. If you want to use flannel as the pod network, specify
--pod-network-cidr=10.244.0.0/16 if you’re using the daemonset manifest below. However, please note that this is not required for any
other networks besides Flannel.
You may also would like to check my automated step-by-step installation for Raspberry Pi 3 with Ansible, since there is no issue with DNS and probably will work with Ubuntu 16 as well:
NAMESPACE NAME READY STATUS RESTARTS AGE
default busybox-894550917-7vj3z 1/1 Running 0 15h
default busybox-894550917-p9vnl 1/1 Running 1 3d
default gogs-3464422143-cf5wb 1/1 Running 0 16h
kube-system dummy-2501624643-pxmgz 1/1 Running 2 3d
kube-system etcd-master.cluster.local 1/1 Running 2 3d
kube-system kube-apiserver-master.cluster.local 1/1 Running 2 3d
kube-system kube-controller-manager-master.cluster.local 1/1 Running 2 3d
kube-system kube-discovery-1659614412-vrhv4 1/1 Running 2 3d
kube-system kube-dns-4211557627-kpsj4 4/4 Running 8 3d
kube-system kube-flannel-ds-d1bgg 2/2 Running 6 3d
kube-system kube-flannel-ds-fcp4b 2/2 Running 6 3d
kube-system kube-flannel-ds-n7p3m 2/2 Running 6 3d
kube-system kube-flannel-ds-tn7nd 2/2 Running 6 3d
kube-system kube-flannel-ds-vpk4w 2/2 Running 6 3d
kube-system kube-proxy-5nmtn 1/1 Running 2 3d
kube-system kube-proxy-gq7bz 1/1 Running 2 3d
kube-system kube-proxy-lkkgm 1/1 Running 2 3d
kube-system kube-proxy-mlh3v 1/1 Running 1 3d
kube-system kube-proxy-sg8n8 1/1 Running 2 3d
kube-system kube-scheduler-master.cluster.local 1/1 Running 2 3d
kube-system kubernetes-dashboard-3507263287-h9q33 1/1 Running 2 3d
kube-dns won't start until the network is up.
Look in the kubelet logs on each machine for more information about the crash that is causing the CrashLoopBackoff.
How did you get ARM images for Weave Net? The weaveworks/weave-kube image on DockerHub is only built for x64.
Edit: as @pidster says Weave Net now supports ARM
