Task overview

How to deploy application into Kubernetes using werf.

werf uses Helm with some additions to deploy applications into Kubernetes. In this article we will create a simple web application, build all needed images, write helm templates and run it on your Kubernetes cluster.


  • Working Kubernetes cluster. It may be Minikube or regular Kubernetes installation. Read the article about Minikube setup to set up local minikube instance with Docker registry.

  • Working Docker registry.

    • Accessible from a host machine to push images to the Docker registry and read info about images being deployed.

    • Accessible from Kubernetes nodes to pull images from the Docker registry.

  • Installed werf dependencies on the host system.

  • Installed multiwerf on the host system.

  • Installed kubectl on a host machine configured to access your Kubernetes cluster (https://kubernetes.io/docs/tasks/tools/install-kubectl/).

NOTICE In the following steps we will use :minikube as REPO argument of werf commands. If you are using own Kubernetes and Docker registry installation, specify your own REPO address instead of :minikube.

Select werf version

This command should be run prior running any werf command in your shell session:

. $(multiwerf use 1.1 stable --as-file)

The application

Our example application is a simple web application. To run this application we only need a web server.

Desirable Kubernetes layout:

 | backend (Deployment) |
  | frontend (Ingress) |

Where backend is a web server, frontend is a proxy for our application and also it is an entry point server to access Kubernetes cluster from outside.

Application files

werf expects that all files needed to build and deploy are residing in the same directory with application source files itself (if any).

So let’s create empty application directory on host machine:

mkdir myapp
cd myapp

Prepare an image

We need to prepare main application image with a web server. Create the following werf.yaml in the root of the application directory:

project: myapp
configVersion: 1

image: ~
from: python:alpine
  - file:
      path: /app
      state: directory
      mode: 0755
  - name: Prepare main page
        <!DOCTYPE html>
            <img src="https://flant.com/images/logo_en.png" style="max-height:100%;" height="76">
      dest: /app/index.html

Our web application consists of a single static web page which created right in the config. This page is served by python HTTP server.

Build and push an image with the following command:

werf build-and-publish --stages-storage :local --tag-custom myapp --images-repo :minikube

The image name consists of REPO and TAG. We have specified :minikube as a REPO — this is a shortcut for werf-registry.kube-system.svc.cluster.local:5000/myapp. As we have specified myapp as a tag, for our example werf will push into the Docker registry image with the name werf-registry.kube-system.svc.cluster.local:5000/myapp:myapp.

Prepare deploy configuration

werf uses helm under the hood to apply Kubernetes configuration. To describe kubernetes configuration werf also use helm configuration files (templates, values) with some extensions, such as secret files and secret values, additional Helm Go templates to generate image names and some more.

Backend configuration

Place a file .helm/templates/010-backend.yaml with configuration of backend and then we will see what’s going on in detail:

apiVersion: apps/v1
kind: Deployment
  name: {{ .Chart.Name }}-backend
    service: {{ .Chart.Name }}-backend
  replicas: 4
      service: {{ .Chart.Name }}-backend
        service: {{ .Chart.Name }}-backend
      - name: backend
        workingDir: /app
        command: [ "python3", "-m", "http.server", "8080" ]
{{ werf_container_image . | indent 8 }}
            path: /
            port: 8080
            scheme: HTTP
            path: /
            port: 8080
            scheme: HTTP
        - containerPort: 8080
          name: http
          protocol: TCP
{{ werf_container_env . | indent 8 }}
apiVersion: v1
kind: Service
  name: {{ .Chart.Name }}-backend
  clusterIP: None
    service: {{ .Chart.Name }}-backend
  - name: http
    port: 8080
    protocol: TCP

In this configuration Deployment with name myapp-backend (after template {{ .Chart.Name }}-backend expansion) with several replicas specified.

Construction {{ werf_container_image . | indent 8 }} is an addition of werf to helm which:

  • always generates right image name (werf-registry.kube-system.svc.cluster.local:5000/myapp:latest in our case), and

  • may generate other related fields (such as imagePullPolicy) based on some external conditions.

Go template function werf_container_image is the valid way to specify image from config in Kubernetes resource configuration. There may be multiple images described in config see the reference for details.

Construction {{ werf_container_env . | indent 8 }} is another addition of werf to helm which may generate environment variables section for the Kubernetes resource. It is needed for Kubernetes to shut down and restart deployment pods only when docker image has been changed, see the reference for details.

Finally, in this configuration Service myapp-backend specified to access Pods of Deployment myapp-backend.

Frontend configuration

To describe frontend configuration place file .helm/templates/090-frontend.yaml with the following content:

apiVersion: networking.k8s.io/v1beta1
kind: Ingress
  name: myapp-frontend
    kubernetes.io/ingress.class: "nginx"
    nginx.ingress.kubernetes.io/ssl-redirect: "false"
  - host: myapp.local
      - path: /
          serviceName: myapp-backend
          servicePort: 8080

This Ingress configuration set up nginx proxy server for host myapp.local to our backend web server myapp-backend.

Run deploy

If you use minikube then enable ingress addon before running deploy:

minikube addons enable ingress

Run deploy with werf:

werf deploy --stages-storage :local --images-repo :minikube --tag-custom myapp --env dev

With this command werf will create all Kubernetes resources using helm and watch until myapp-backend Deployment is ready (when all replicas Pods are up and running).

Environment --env is a required param needed to generate helm release name and kubernetes namespace.

Helm release with name myapp-dev will be created. This name consists of project name myapp (which you’ve placed in the werf.yaml) and specified environment dev. Check docs for details about helm release name generation.

Kubernetes namespace myapp-dev will also be used. This name also consists of project name myapp and specified environment dev. Check docs for details about Kubernetes namespace generation.

Check your application

Now it is time to know the IP address of your Kubernetes cluster. If you use minikube get it with (in the most cases the IP address will be

minikube ip

Make sure that host name myapp.local is resolving to this IP address on your machine. For example append this record to the /etc/hosts file: myapp.local

Then you can check application by url: http://myapp.local.

Delete application from Kubernetes

To completely remove deployed application run this dismiss werf command:

werf dismiss --env dev --with-namespace

See also

For all werf deploy features such as secrets take a look at reference.