Files mentioned in the chapter

  • Dockerfile
  • werf.yaml

Dockerfile is a classic method to build images. Probably, your applications are already using this mechanism for the assembly. That is why we will start with it and then learn how to use it with werf. In the next chapters, we will speed up the build using an alternative syntax for describing the assembly process, but in this chapter, we will focus on getting the result quickly.

Preparing the workplace

We assume that you have already installed werf and Docker.

Create a directory on your computer and follow these steps:

git clone
cp -r werf-guides/examples/nodejs/000_app ./
cd 000_app 
git init
git add .
git commit -m "initial commit"

This way you will copy the code of the Node.js application to a local directory and initialize a Git repository in it.

Note that werf follows the principles of giterminism: it fully relies on the state described in the Git repository. This means that files not committed to the Git repository will be ignored by default. Thereby, if you have the source code, then you can turn an application to the specific operating condition at any time.

Dockerfile-based build process

The build configuration of our application consists of the following steps:

  • pull the base Node.js image (e.g., node:14-stretch);
  • add the application code to it;
  • install npm dependencies.

Now let’s insert all these steps into a Dockerfile:

FROM node:14-stretch
COPY . .

RUN apt update
RUN apt install -y tzdata locales
RUN npm ci

CMD ["node","/app/app.js"]
FROM node:14-stretch WORKDIR /app COPY . . RUN apt update RUN apt install -y tzdata locales RUN npm ci CMD ["node","/app/app.js"]

Integrating Dockerfile into werf

Let’s add our Dockerfile to werf. To do this, create a werf.yaml file in the repository root (it describes the build process of the entire project):

project: werf-guided-nodejs
configVersion: 1
image: basicapp
dockerfile: Dockerfile
project: werf-guided-nodejs configVersion: 1 --- image: basicapp dockerfile: Dockerfile
What do all those lines mean?

werf.yaml starts with a mandatory meta-information section:

project: werf-guided-nodejs
configVersion: 1
project: werf-guided-nodejs configVersion: 1


  • project is a field containing the unique name of the application project. The default project name is used when generating the Helm release name and the namespace in Kubernetes. Changing the name of an active project is a tedious task that requires a number of manual actions (more information about possible consequences is available here);
  • configVersion specifies the syntax version used in the werf.yaml (currently, only version 1 is supported).

The next section, called the image config section, defines the build process.

image: basicapp
dockerfile: Dockerfile
--- image: basicapp dockerfile: Dockerfile

The image: basicapp line sets the image ID that will be used in the rollout configuration as well as for invoking werf commands for a specific image described in werf.yaml (for example, werf build basicapp, werf run basicapp, etc.).

The dockerfile: Dockerfile line specifies that the build configuration is defined in the existing file located at the Dockerfile path.

You can also make use of other directives (their description is available here).

The single werf.yaml file can contain the definitions of an arbitrary number of images.


Now that you have successfully added Dockerfile and werf.yaml described above, it is necessary to commit changes to Git:

git add .
git commit -m "work in progress"

The build command starts the assembly process:

werf build

The sub-chapter “Speeding up the build” contains instructions on how to adapt the Dockerfile-based build process to an alternative werf syntax called Stapel and gain access to some advanced features, such as Git history-based incremental rebuilds, the usage of Ansible and inter-assembly cache, convenient diagnostic tools, and much more.

But even now, you may notice that werf outputs the build logs in the extended format:

┌ ⛵ image basicapp
│ ┌ Building stage basicapp/dockerfile
│ │ basicapp/dockerfile  Sending build context to Docker daemon  69.63kB
│ │ basicapp/dockerfile  Step 1/16 : FROM node:14-stretch
│ │ basicapp/dockerfile   ---> b90fa0d7cbd1
│ │ basicapp/dockerfile  Step 2/16 : WORKDIR /app
│ │ basicapp/dockerfile   ---> Using cache
│ │ basicapp/dockerfile  Successfully built 02a0a425890a
│ │ basicapp/dockerfile  Successfully tagged a1cbf6dc-343f-4a77-b846-d0f12a700cb7:latest
│ ├ Info
│ │       name: werf-guided-nodejs:a473b87e1ad65f102fa90f8c6647b03056e5ae95ff1ef3c5e7fd2c31-1605597979927
│ │       size: 953.1 MiB
│ └ Building stage basicapp/dockerfile (21.94 seconds)
└ ⛵ image basicapp (22.04 seconds)

Running time 22.07 seconds


Let’s run the built image using the werf run command:

werf run --docker-options="--rm -p 3000:3000" -- node /app/app.js

Note that we set the docker parameters via --docker-options, while the startup command is preceded by two hyphens.

You may also notice that werf run also performs the build. In other words, it is not really necessary to run the build separately.

Now you can access the application locally on port 3000:

Making changes

As you might guess, we are going to continually update our application. Let’s see how to do this in the right way by making some arbitrary changes to the application code:

app.get('/labels', function (req, res) {
  halt_if_errors(global_errors, res);
  res.send('Our changes');
app.get('/labels', function (req, res) { halt_if_errors(global_errors, res); res.send('Our changes'); });
  1. Stop the running werf run (by pressing Ctrl+C in the console where it is running).
  2. Start it again:
     werf run --docker-options="--rm -p 3000:3000" -- node /app/app.js
  3. Watch as the application is being rebuilt and restarted, and then connect to the API:
  4. You probably expect to see the Our changes message, but it isn’t there. Everything is the same… but why?

The thing is we forgot to commit changes to Git prior to step 1 in the scenario above.

What is the correct way, and why go through all those troubles?
  1. Make changes to the code.
  2. Commit them:
    git add .
    git commit -m "wip"
  3. Restart werf run:
     werf run --docker-options="--rm -p 3000:3000" -- node /app/app.js
  4. View the result in the browser.

Werf follows the principles of giterminism as we mentioned in the beginning of the article. A strict binding to Git ensures the reproducibility of each specific solution. More details about giterminism mechanics and developers mode with --dev flag are available in the “What you need to know” chapter. Until then, we’ll focus on building and delivering an application to the cluster.