Adding images

To build images with werf, you have to add a description of each image to the werf.yaml file of the project. The image description starts with the image directive specifying the image name:

project: example
configVersion: 1
---
image: frontend
# ...
---
image: backend
# ...
---
image: database
# ...

The image name is a unique internal image identifier that allows it to be referenced during configuration and when invoking werf commands.

Next, for each image in werf.yaml, you have to define the build instructions using Dockerfile or Stapel.

Dockerfile

Writing Dockerfile instructions

The format of the image build instructions matches the Dockerfile format. Refer to the following resources for more information:

Using Dockerfiles

The typical configuration of a Dockerfile-based build looks as follows:

# Dockerfile
FROM node
WORKDIR /app
COPY package*.json /app/
RUN npm ci
COPY . .
CMD ["node", "server.js"]

# werf.yaml
project: example
configVersion: 1
---
image: backend
dockerfile: Dockerfile

Using a specific Dockerfile stage

You can also define multiple target images using various stages of the same Dockerfile:

# Dockerfile
FROM node as backend
WORKDIR /app
COPY package*.json /app/
RUN npm ci
COPY . .
CMD ["node", "server.js"]

FROM python as frontend
WORKDIR /app
COPY requirements.txt /app/
RUN pip install -r requirements.txt
COPY . .
CMD ["gunicorn", "app:app", "-b", "0.0.0.0:80", "--log-file", "-"]

# werf.yaml
project: example
configVersion: 1
---
image: backend
dockerfile: Dockerfile
target: backend
---
image: frontend
dockerfile: Dockerfile
target: frontend

… as well as define images based on different Dockerfiles:

# werf.yaml
project: example
configVersion: 1
---
image: backend
dockerfile: dockerfiles/Dockerfile.backend
---
image: frontend
dockerfile: dockerfiles/Dockerfile.frontend

Selecting the build context directory

The context directive sets the build context. Note: In this case, the path to the Dockerfile must be specified relative to the context directory:

project: example
configVersion: 1
---
image: docs
context: docs
dockerfile: Dockerfile
---
image: service
context: service
dockerfile: Dockerfile

In the example above, werf will use the Dockerfile at docs/Dockerfile to build the docs image and the Dockerfile at service/Dockerfile to build the service image.

Using build secrets

A build secret is any sensitive information, such as a password or API token, required during the build process of your application.

Build arguments and environment variables are not suitable for passing secrets to the build process, as they are retained in the final image.

Instead, you can define secrets in the werf.yaml file to use them during the build process.

# werf.yaml
project: example
configVersion: 1
---
image: backend
dockerfile: Dockerfile
secrets:
  - env: AWS_ACCESS_KEY_ID
  - id: aws_secret_key
    env: AWS_SECRET_ACCESS_KEY
  - src: "~/.aws/credentials"
  - id: plainSecret
    value: plainSecretValue

To access a secret during the build, use the --mount=type=secret flag in the Dockerfile’s RUN instructions.

When a secret is mounted, it is available as a file by default. The default mount path for the secret file inside the build container is /run/secrets/<id>. If an id is not specified for a secret in werf.yaml, the default value is assigned automatically:

  • For env, the id defaults to the name of the environment variable.
  • For src, the id defaults to the file name (e.g., for /path/to/file, the id will be file).

For value — the id field is mandatory.

# Dockerfile
FROM alpine:3.18

# Example of using a secret from an environment variable
RUN --mount=type=secret,id=AWS_ACCESS_KEY_ID \
    export WERF_BUILD_SECRET="$(cat /run/secrets/AWS_ACCESS_KEY_ID)"

# Example of using a secret from a secrets file
RUN --mount=type=secret,id=credentials \
    AWS_SHARED_CREDENTIALS_FILE=/run/secrets/credentials \
    aws s3 cp ...

# Example of mounting a secret as an environment variable
RUN --mount=type=secret,id=AWS_ACCESS_KEY_ID,env=AWS_ACCESS_KEY_ID \
    --mount=type=secret,id=aws-secret-key,env=AWS_SECRET_ACCESS_KEY \
    aws s3 cp ...

# Example of mounting a secret as a file with a different name
RUN --mount=type=secret,id=credentials,target=/root/.aws/credentials \
    aws s3 cp ...

# Example of using a raw value that will not be stored in the final image
RUN --mount=type=secret,id=plainSecret \
    export WERF_BUILD_SECRET="$(cat /run/secrets/plainSecret)"

Using the SSH agent

You can provide access to the SSH agent socket or SSH keys during the build process. This is particularly useful if your Dockerfile contains commands that require SSH authentication, such as cloning a private repository.

To enable this, use the --mount=type=ssh flag with RUN commands:

FROM alpine
RUN apk add --no-cache openssh-client
RUN mkdir -p -m 0700 ~/.ssh && ssh-keyscan gitlab.com >> ~/.ssh/known_hosts
RUN --mount=type=ssh ssh -q -T git@gitlab.com 2>&1 | tee /hello

You can find detailed information about using the SSH agent in werf here.

Adding arbitrary files to the build context

By default, the build context of a Dockerfile image only includes files from the current project repository commit. Files not added to Git or non-committed changes are not included in the build context. This logic follows the giterminism configuration default.

You can add files that are not stored in Git to the build context through the contextAddFiles directive in the werf.yaml file. You also have to enable the contextAddFiles directive in werf-giterminism.yaml (more about giterminism):

# werf.yaml
project: example
configVersion: 1
---
image: app
dockerfile: Dockerfile
context: app
contextAddFiles:
- file1
- dir1/
- dir2/file2.out

# werf-giterminism.yaml
giterminismConfigVersion: 1
config:
  dockerfile:
    allowContextAddFiles:
    - app/file1
    - app/dir1/
    - app/dir2/file2.out

In the above configuration, the build context will include the following files:

  • app/**/* of the current project repository commit;
  • app/file1, app/dir2/file2.out files and the dir1 directory in the project directory.

Multiplatform Build

werf supports multiplatform and cross-platform builds, allowing you to create images for various architectures and operating systems (for more details, see the relevant section of the documentation).

Cross-compilation

If your project requires cross-compilation, you can use multi-stage builds to create artifacts for target platforms using the build platform. The following build arguments are available for this purpose:

  • TARGETPLATFORM: the platform for the build result (e.g., linux/amd64, linux/arm/v7, windows/amd64).
  • TARGETOS: the OS of the target platform.
  • TARGETARCH: the architecture of the target platform.
  • TARGETVARIANT: the variant of the target platform.
  • BUILDPLATFORM: the platform of the node performing the build.
  • BUILDOS: the OS of the build platform.
  • BUILDARCH: the architecture of the build platform.
  • BUILDVARIANT: the variant of the build platform.

Example:

FROM --platform=$BUILDPLATFORM golang:alpine AS build
ARG TARGETPLATFORM
ARG BUILDPLATFORM
RUN echo "I am running on $BUILDPLATFORM, building for $TARGETPLATFORM" > /log

FROM alpine
COPY --from=build /log /log

In this example, the FROM instruction is pinned to the build platform of the builder using the --platform=$BUILDPLATFORM option to prevent emulation. The $BUILDPLATFORM and $TARGETPLATFORM arguments are then used in the RUN instruction.

Stapel

Stapel is a built-in alternative syntax for describing build instructions. Detailed documentation on Stapel syntax is available in the corresponding documentation section.

Below is an example of a minimal configuration of the Stapel image in werf.yaml:

project: example
configVersion: 1
---
image: app
from: ubuntu:22.04

And here’s how you can add source files from Git to an image:

project: example
configVersion: 1
---
image: app
from: ubuntu:22.04
git:
- add: /
  to: /app

There are 4 stages for describing arbitrary shell instructions, as well as a git.stageDependencies directive to set up triggers to rebuild these stages whenever the corresponding stages change (learn more):

project: example
configVersion: 1
---
image: app
from: ubuntu:22.04
git:
- add: /
  to: /app
  stageDependencies:
    install:
    - package-lock.json
    - Gemfile.lock
    beforeSetup:
    - app/webpack/
    - app/assets/
    setup:
    - config/templates/
shell:
  beforeInstall:
  - apt update -q
  - apt install -y libmysqlclient-dev mysql-client g++
  install:
  - bundle install
  - npm install
  beforeSetup:
  - bundle exec rails assets:precompile
  setup:
  - rake generate:configs

Auxiliary images and Golang templating are also supported. You can use the former to import files into the target image (similar to COPY-from=STAGE in multi-stage Dockerfile):

{{ $_ := set . "BaseImage" "ubuntu:22.04" }}

{{ define "package:build-tools" }}
  - apt update -q
  - apt install -y gcc g++ build-essential make
{{ end }}

project: example
configVersion: 1
---
image: builder
from: {{ .BaseImage }}
shell:
  beforeInstall:
{{ include "package:build-tools" }}
  install:
  - cd /app
  - make build
---
image: app
from: alpine:latest
import:
- image: builder
  add: /app/build/app
  to: /usr/local/bin/app
  after: install

For more info on how to write Stapel instructions refer to the documentation.

Linking images

Inheritance and importing files

A multi-stage mechanism allows you to define a separate image stage in a Dockerfile and use it as a basis for another image or copy individual files from it.

werf allows to do this not only within a single Dockerfile, but also for arbitrary images defined in werf.yaml, including those built from different Dockerfiles, or built by the Stapel builder. werf will take care of all orchestration and dependency management and then build everything in one step (as part of the werf build command).

Below is an example of using a base.Dockerfile image as the basis for a Dockerfile image:

# base.Dockerfile
FROM ubuntu:22.04
RUN apt update -q && apt install -y gcc g++ build-essential make curl python3

# Dockerfile
ARG BASE_IMAGE
FROM ${BASE_IMAGE}
WORKDIR /app
COPY . .
CMD [ "/app/server", "start" ]

# werf.yaml
project: example
configVersion: 1
---
image: base
dockerfile: base.Dockerfile
---
image: app
dockerfile: Dockerfile
dependencies:
- image: base
  imports:
  - type: ImageName
    targetBuildArg: BASE_IMAGE

The next example deals with importing files from a Stapel image into a Dockerfile image:

# werf.yaml
project: example
configVersion: 1
---
image: builder
from: golang
git:
- add: /
  to: /app
shell:
  install:
  - cd /app
  - go build -o /app/bin/server
---
image: app
dockerfile: Dockerfile
dependencies:
- image: builder
  imports:
  - type: ImageName
    targetBuildArg: BUILDER_IMAGE

# Dockerfile
ARG BUILDER_IMAGE
FROM ${BUILDER_IMAGE} AS builder

FROM alpine
COPY --from=builder /app/bin /app/bin
CMD [ "/app/bin/server", "server" ]

Passing information about the built image to another image

werf allows you to get information about the built image when building another image. Suppose, the build instructions of the app image require the names and digests of the auth and controlplane images published in the container registry. The configuration in this case would look like this:

# modules/auth/Dockerfile
FROM alpine
WORKDIR /app
COPY . .
RUN ./build.sh

# modules/controlplane/Dockerfile
FROM alpine
WORKDIR /app
COPY . .
RUN ./build.sh

# Dockerfile
FROM alpine
WORKDIR /app
COPY . .

ARG AUTH_IMAGE_NAME
ARG AUTH_IMAGE_DIGEST
ARG CONTROLPLANE_IMAGE_NAME
ARG CONTROLPLANE_IMAGE_DIGEST

RUN echo AUTH_IMAGE_NAME=${AUTH_IMAGE_NAME}                     >> modules_images.env
RUN echo AUTH_IMAGE_DIGEST=${AUTH_IMAGE_DIGEST}                 >> modules_images.env
RUN echo CONTROLPLANE_IMAGE_NAME=${CONTROLPLANE_IMAGE_NAME}     >> modules_images.env
RUN echo CONTROLPLANE_IMAGE_DIGEST=${CONTROLPLANE_IMAGE_DIGEST} >> modules_images.env

# werf.yaml
project: example
configVersion: 1
---
image: auth
dockerfile: Dockerfile
context: modules/auth/
---
image: controlplane
dockerfile: Dockerfile
context: modules/controlplane/
---
image: app
dockerfile: Dockerfile
dependencies:
- image: auth
  imports:
  - type: ImageName
    targetBuildArg: AUTH_IMAGE_NAME
  - type: ImageDigest
    targetBuildArg: AUTH_IMAGE_DIGEST
- image: controlplane
  imports:
  - type: ImageName
    targetBuildArg: CONTROLPLANE_IMAGE_NAME
  - type: ImageDigest
    targetBuildArg: CONTROLPLANE_IMAGE_DIGEST

During the build, werf will automatically insert the appropriate names and identifiers into the referenced build-arguments. werf will take care of all orchestration and dependency mapping and then build everything in one step (as part of the werf build command).

Using intermediate and final images

By default, all images are final, allowing the user to operate with them using their names as arguments for most werf commands and in Helm chart templates. The final directive can be used to regulate this property of an image.

Intermediate images (final: false), unlike final images:

Example of using the final directive:

project: example
configVersion: 1
---
image: builder
final: false
dockerfile: Dockerfile.builder
---
image: app
dockerfile: Dockerfile.app
dependencies:
- image: builder
  imports:
  - type: ImageName
    targetBuildArg: BUILDER_IMAGE_NAME