I am trying to set up a single node kubernetes cluster for demo and testing purposes, and I want it to behave like a 'full blown' k8s cluster (like google container engine). My client has their own k8s installation, which for this discussion we can assume acts pretty much like google container engine's k8s installation.

Getting the Ingress IP on Full Blown K8s

I am creating a wordpress pod and exposing it as a service, as described in this tutorial: https://cloud.google.com/container-engine/docs/tutorials/hello-wordpress

If you want to replicate the issue, just can just copy paste the commands below, which I lifted from the tutorial: (This assumes you have a project called 'stellar-access-117903'.. if not please set to name of your Google Container Engine project.)

# set up the cluster  (this will take a while to provision)
#
 gcloud config set project stellar-access-117903
 gcloud config set compute/zone us-central1-b
 gcloud container clusters create hello-world \
     --num-nodes 1 \
     --machine-type g1-small

# Create the pod, and expose it as a service
#
kubectl run wordpress --image=tutum/wordpress --port=80
kubectl expose rc wordpress --type=LoadBalancer

# Describe the service
kubectl describe services wordpress

The output of the describe command contains a line 'LoadBalancer Ingress: {some-ip-address}' which is exactly what I'd expect. Now, when I do the same thing with the single node cluster setup i don't get that line. I am able to hit the wordpress service at the IP that appears in the output of the 'describe service' command.. But in 'single node' mode, the IP that is printed out is the >cluster IP< of the service, which typically (as I understand it) is not publicly accessible. For some reason it is publicly accessible in single node mode. We can replicate this with the following steps.

NOT Getting the Ingress IP on Single Node K8s

First setup single node k8s, as described in this tutorial: https://github.com/kubernetes/kubernetes/blob/master/docs/getting-started-guides/docker.md

For easy reproducibility, I have included all the commands below, so you can just copy/paste:

K8S_VERSION=1.1.1

sudo docker run --net=host -d gcr.io/google_containers/etcd:2.0.12 /usr/local/bin/etcd --addr=127.0.0.1:4001 --bind-addr=0.0.0.0:4001 --data-dir=/var/etcd/data


sudo docker run \
    --volume=/:/rootfs:ro \
    --volume=/sys:/sys:ro \
    --volume=/dev:/dev \
    --volume=/var/lib/docker/:/var/lib/docker:ro \
    --volume=/var/lib/kubelet/:/var/lib/kubelet:rw \
    --volume=/var/run:/var/run:rw \
    --net=host \
    --pid=host \
    --privileged=true \
    -d \
    gcr.io/google_containers/hyperkube:v${K8S_VERSION} \
    /hyperkube kubelet --containerized --hostname-override="127.0.0.1" --address="0.0.0.0" --api-servers=http://localhost:8080 --config=/etc/kubernetes/manifests

sudo docker run -d --net=host --privileged gcr.io/google_containers/hyperkube:v${K8S_VERSION} /hyperkube proxy --master=http://127.0.0.1:8080 --v=2


# set your context to use the locally running k8s API server 
#
kubectl config set-cluster dev --server=http://localhost:8080
kubectl config set-context dev --cluster=dev  --namespace=$NS
kubectl config use-context dev

Now, execute the very same commands that you performed against Google Container Engine's k8s

# Create the pod, and expose it as a service
#
kubectl run wordpress --image=tutum/wordpress --port=80
kubectl expose rc wordpress --type=LoadBalancer

# Describe the service
kubectl describe services wordpress

The output of the last command (which you will see has no 'Ingress' information) is:

Name:           wordpress
Namespace:      default
Labels:         run=wordpress
Selector:       run=wordpress
Type:           LoadBalancer
IP:         10.0.0.61
Port:           <unnamed>   80/TCP
NodePort:       <unnamed>   31795/TCP
Endpoints:      172.17.0.30:80
Session Affinity:   None
No events.

In google container engine's k8s, I see events like ' Creating load balancer ', ' Load balancer created '. But nothing like that happens in the single node instance.

I am wondering ... is there some configuration I need to do to get them to work identically ? It is very important that they work identically... differing only in their scalability, because we want to run tests against the single node version, and it will be very confusing if it behaves differently.

Thanks in advance for your help -chris