cilium: Observing ImagePullBackOff and CrashLoopBackoff in Kubernetes example with cluster built with kops running in AWS

When following the Getting Started docs for installing Cilium on Kubernetes, the pods created end up in either an ImagePullBackOff state or a CrashLoopBackoff state as follows:

$ kubectl get pods --namespace kube-system
NAME                              READY     STATUS             RESTARTS   AGE
cilium-gbz8v                      0/1       Error              0          1m
cilium-pxfz7                      0/1       CrashLoopBackOff   3          1m
cilium-z2cnm                      0/1       ImagePullBackOff   0          1m
...

General Information

• Cilium version: Docker Image cilium/cilium:stable
• Kernel version: 4.4.111-k8s
• Orchestration system version in use: Kubernetes 1.8.6
• Link to relevant artifacts: https://github.com/cilium/cilium/blob/master/examples/kubernetes/cilium.yaml

How to reproduce the issue

Create a k8s cluster using:

kops create cluster \
  --cloud=aws \
  --name=<REDACTED> \
  --state=s3://<REDACTED> \
  --zones=us-west-2b \
  --node-count=2 \
  --node-size=t2.medium \
  --master-size=m4.large \
  --dns=private \
  --dns-zone=<REDACTED> \
  --dns-zone=<REDACTED> \
  --yes

After about 15 minutes the nodes will be up and DNS records will have been propagated. Execute comand:

$ kubectl create -f https://raw.githubusercontent.com/cilium/cilium/HEAD/examples/kubernetes/cilium.yaml

and observe state of pods over a few minutes.