#PaddleSurf

###Description:

Simple setup of a Ruby app with Sinatra/Thin connecting to a Redis backend using Docker and Kubernetes on GCE. Loadbalancing is done with k8s Service, discovery with kube-dns.

###Usage

#####Docker Images

• install Docker
• build the Docker images using the command in the comment of each Dockerfile

#####Google Cloud infra

• Go to Google Cloud website and create a new project

• retag our Docker images to publish them to the Google Cloud registry

$ sudo docker tag app eu.gcr.io/paddlesurf-XXXX/app:v1
$ sudo docker tag redis eu.gcr.io/paddlesurf-XXXX/redis:v1

• install Google Cloud SDK and kubectl

$ ./google-cloud-sdk/install.sh
$ gcloud init
$ gcloud components install kubectl

• We can then create our container infrastructure using the SDK: It will create a kubernetes cluster (Debian hosts, Google images)

$ gcloud container clusters create paddlesurf
$ gcloud container clusters get-credentials paddlesurf \
  --zone europe-west1-d --project <projectid>

After creating the cluster, you should get the output of: gcloud container clusters list

#####Terraform variant : This is more complex and only useful if we need to manage more services than just the container cluster...

• install terraform
• create infra using terraform:
• export your Google Cloud credentials (json file): $ export GOOGLE_CREDENTIALS=<string>
• modify your project name in terraform/paddlesurf_k8_gce.tf
• launch terraform: $ terraform plan $ terraform apply
• get the endpoint url: $ terraform show We can now test the URL with curl -kv https://1.2.3.4

#####Push docker images

$ gcloud docker push eu.gcr.io/paddlesurf-XXXX/app:v1
$ gcloud docker push eu.gcr.io/paddlesurf-XXXX/redis:v1

#####Create the kubernetes services: we have one yaml config for each service (k/v store database and app). Each yaml contain a Service definition, either:

• internal port-based: Services map a port on each cluster node to ports on one or more pods.
• loadbalancer: Clients connect to the external IP address, which forwards to app pod.

$ kubectl create -f k8s/redis.yaml
$ kubectl create -f k8s/app.yaml
$ kubectl get services
NAME         CLUSTER-IP    EXTERNAL-IP   PORT(S)    AGE
app          10.3.255.86   1.2.3.4       4567/TCP   1m
kubernetes   10.3.240.1    <none>        443/TCP    6h
redis        10.3.242.37   <nodes>       6379/TCP   1m
$ kubectl get pods
$ kubectl get deployments
$ kubectl get ep app
NAME      ENDPOINTS                     AGE
app       10.0.0.4:4567,10.0.1.6:4567   1m

You can use kubectl logs <pod> to see the output or kubectl describe pods Or connect to your kubernetes cluster webUI directly:

$ gcloud container clusters get-credentials paddlesurf \
  --zone europe-west1-d --project paddlesurf-XXXX
$ kubectl proxy
$ open http://localhost:8001/ui

The application connects to redis using redis host populated with the REDIS_HOST env variable, using kube-dns.

###Access our application

• Get the external IP address and port of our loadbalanced application:

$ kubectl get services app
NAME         CLUSTER-IP    EXTERNAL-IP   PORT(S)    AGE
app          10.3.255.86   1.2.3.4       4567/TCP   1m

and connect using curl http://1.2.3.4:4567.