Working with Docker Registries

How to push, pull and tag Docker images in Codefresh pipelines

Codefresh contains first class Docker registry support. This means that you don’t need to manually write docker login and docker pull/push commands inside pipelines. You use instead declarative YAML and all credential configuration is configured centrally once.

Viewing Docker images

To see all images currently from all your connected Registries, select Artifacts -> Images from the left sidebar and you will see a sorted list of all images.

For each image you get some basic details such as the git branch, commit message and hash that created it, date of creation as well as all tags. You can click on any image and look at its individual metadata.

On the top left of the screen you can find several filters that allow you to search for a specific subset of Docker images:

Filters include:

• Tagged/untagged images
• Base image name
• Git branch
• Tag
• Pipeline volumes

You can add multiple filters and they will work in an AND manner.

On the right side of the screen you also have a list of buttons for actions on each Docker image. These are:

• Launching a Docker image as a test environment
• Promoting a Docker image (explained in the following sections)
• Looking at the docker commands that allow you to pull the image locally on your workstation
• Re-running the pipeline that created this image

Pulling Docker images

Pulling Docker images in Codefresh is completely automatic. You only need to mention a Docker image by name and Codefresh will automatically pull it for you and use it in a pipeline.

Pulling public images

To pull a public image from Dockerhub or other public registry you simply mention the name of the image and tag that you want to use. For example:

CollectAllMyDeps:
  title: Install dependencies
  image: python:3.6.4-alpine3.6
  commands:
    - pip install .

This freestyle step will pull from Dockerhub the image python:3.6.4-alpine3.6 and then run the command pip install . inside it. You can see the image get pulled in the Codefresh pipeline log.

The image will also be cached in the image cache without any other configuration.

Codefresh will also automatically pull for you any images mentioned in Dockerfiles (i.e. the FROM directive) as well as service containers.

Pulling private images

To pull a private image from one of your connected registries, again you mention the image by name and tag. In order for Codefresh to understand that you are talking about a private image you need to prepend the appropriate prefix of the registry domain.

For example in the case of ACR:

registry-name.azurecr.io/my-docker-repo/my-image-name:tag

You can find the full name of any docker image by visiting the image dashboard and looking at the URL field of any tag:

The exact format of the full image name will depend on the type of registry you use. Codefresh will use the domain prefix of each image to understand which integration it will use. It will then take care of all docker login and docker pull commands on its own behind the scenes.

For example if you have connected Azure, AWS and Google registries, you can pull 3 images for each in a pipeline like this:

codefresh.yml

version: '1.0'
steps:
  my_go_unit_tests:
    title: Running Go Unit tests
    image: 'us.gcr.io/project-k8s-sample-123454/my-golang-app:prod'
    commands:
      - go test -v
  my_mvn_unit_tests:
    title: Running Maven Unit tests
    image: '123456789012.dkr.ecr.us-west-2.amazonaws.com/my-java-app:latest'
    commands:
      - mvn test
  my_python_unit_tests:
    title: Running Python Unit tests
    image: 'my-azure-registry.azurecr.io/kostis-codefresh/my-python-app:master'
    commands:
      - python setup.py test        

Codefresh will automatically login to each registry using the credentials you have defined centrally and pull all the images. The same thing will happen with Dockerfiles that mention any valid docker image in their FROM directive.

Pulling images that were just built in the same pipeline

Codefresh allows you to create a docker image on demand and use it in the same pipeline that created it. In several scenarios (such as unit tests) it is very common to use a Docker image right after it was built.

In that case, as a shortcut Codefresh allows you to simply mention the name of the respective build step.

codefresh.yml

version: '1.0'
steps:
  main_clone:
    title: Cloning main repository...
    type: git-clone
    repo: 'codefresh-contrib/python-flask-sample-app'
    revision: 'master'
    git: github
  MyAppDockerImage:
    title: Building Docker Image
    type: build
    image_name: my-app-image
    working_directory: ./
    tag: 'master'
    dockerfile: Dockerfile
  MyUnitTests:
    title: Running Unit tests
    image: '${{MyAppDockerImage}}'
    commands:
      - python setup.py test
  

In this pipeline Codefresh:

1. Checks out source code with the git-clone step
2. Builds a docker image that gets named my-app-image:master. Notice the lack of docker push commands
3. In the next step automatically uses that image and runs python setup.py test inside it. Again notice the lack of docker pull commands.

The important line here is the following:

    image: ${{MyAppDockerImage}}
    

This says to Codefresh “in this step please use the Docker image that was built in the step named MyAppDockerImage”.

You can see the automatic pull inside the Codefresh logs.

The image will still be pushed in your default Docker registry. If you don’t want this behavior you can simply add the disable_push property in the build step.

Pushing Docker images

Pushing to your default Docker registry is completely automatic. All successful build steps automatically push to the default Docker registry of your Codefresh account without any extra configuration.

To push to another registry you only need to know how this registry is linked into Codefresh and more specifically what is unique name of the integration. You can see that name by visiting your integrations screen or asking your Codefresh administrator.

Once you know the registry identifier you can use it an push step or build step by mentioning the registry with that name:

codefresh.yml

  build_image:
    title: Building my app image
    type: build
    image_name: my-app-image
    dockerfile: Dockerfile
    tag: 'master'
  push_to_registry:
    title: Pushing to Docker Registry 
    type: push
    #Name of the build step that is building the image
    candidate: '${{build_image}}'
    tag: '1.2.3'
    # Unique registry name
    registry: azure-demo       

Notice that

• the candidate field of the push step mentions the name of the build step (build_image) that will be used for the image to be pushed
• The registry is only identified by name (i.e. azure-demo). The domain and credentials are not part of the pipeline (they are already known to Codefresh by the Docker registry integration)

You can also override the name of the image with any custom name. This way the push step can choose any image name regardless of what was used in the build step.

codefresh.yml

  build_image:
    title: Building my app image
    type: build
    image_name: my-app-image
    dockerfile: Dockerfile
    tag: 'master'
  push_to_registry:
    title: Pushing to Docker Registry 
    type: push
    #Name of the build step that is building the image
    candidate: '${{build_image}}'
    tag: '1.2.3'
    # Unique registry name
    registry: azure-demo
    image_name: my-company/web-app       

Here the build step is creating an image named my-app-image:master but the push step will actually push it as my-company/web-app:1.2.3

For more examples such as using multiple tags, or pushing in parallel see the push examples

Pushing images with an optional prefix

There are some Registry providers that require a specific prefix for all your Docker images. This is often the name of an organization, account or other top level construct defined by the Registry.

codefresh.yml

  build:
    title: "Building Docker image"
    type: "build"
    image_name: "acme-company/trivial-go-web"
    working_directory: "${{clone}}"
    tag: "latest"
    dockerfile: "Dockerfile.multistage"
    stage: "build"
    registry: azure

The example above will push the final Docker image as kostisazureregistry.azurecr.io/acme-company/trivial-go-web:latest. However, you can also setup the prefix globally once in the Docker integration screen. This way you can simplify your pipelines and have them mention only the final image name.

The example above will automatically prefix all your Docker images with acme-company.

Now you can simplify your build/push step as below:

codefresh.yml

  build:
    title: "Building Docker image"
    type: "build"
    image_name: "trivial-go-web"
    working_directory: "${{clone}}"
    tag: "latest"
    dockerfile: "Dockerfile.multistage"
    stage: "build"
    registry: azure

The final Docker image will still be kostisazureregistry.azurecr.io/acme-company/trivial-go-web:latest.

Working with multiple registries with the same domain

With Codefresh you can connect multiple registries on a global level. This allows you to create pipelines that push/pull images to different registries without having to deal with Docker credentials inside the pipeline itself.

However, there are several times where you have multiple registries that have the same domain. For example you might have two Dockerhub accounts connected to Codefresh (so both of them can resolve images for the docker.io domain)

This means that when you reference an image by domain name (e.g. in a freestyle step), Codefresh might not know which Docker registry account you want to use for the pull action.

Notice, that this is not a Codefresh limitation, but a Docker one. Even with vanilla Docker you cannot login to multiple registries at the same time if they share the same domain.

To solve this problem, Codefresh will automatically detect connected registries that have the same domain and allow you to designate the primary one, that will be used for image resolution when pulling Docker images.

In the example above, even though two Dockerhub integrations are connected to Codefresh, only the primary one will be used for pulling images from the docker.io domain (You can still use the second one in push/build steps using the registry property).

You can override the default behavior in each pipeline, by adding the optional registry_context property to instruct Codefresh on how to use a specific registry for pulling Docker images (if you have more than one for the same domain).

You can use the registry_context property in build, push, freestyle and composition steps.

The registry_context property takes as value the name of an external connected registry. Build and composition steps accept an array of values as registry_context. In all cases, by using this optional property you instruct Codefresh to use a specific registry for pulling images.

Notice that the optional registry_context and registry_context properties only affect the pulling of Docker images. The registry used for pushing images is still declared explicitly in build and push pipeline steps.

The syntax for the freestyle step is the following:

  test:
    title: "Running test"
    type: "freestyle" 
    image: "gcr.io/my-google-project/my-image:latest"
    registry_context: my-second-gcr-registry # define what registry will be used for pulling the image
    working_directory: "${{clone}}" 
    commands:
      - "ls"

The syntax for the build step is the following:

  build:
    title: "Building Docker image"
    type: "build"
    image_name: "trivial-go-web"
    working_directory: "${{clone}}"
    tag: "latest"
    dockerfile: "Dockerfile.multistage"
    stage: "build"
    registry_contexts: # define what registries will be used for pulling images
      - second-dockerhub
      - production-azure
    registry: azure

The syntax for the push step is the following:

  push:
    title: "Pushing 1st Docker image"
    type: push
    image_name: "kostiscodefresh/trivial-go-web"
    tag: "latest"
    stage: "push" 
    registry: dockerhub # Docker registry to push to 
    registry_context: second-dockerhub # Docker registry to pull images from
    candidate: ${{build}}

The syntax for the composition step is the following:

  my-composition:
    title: Running Composition
    type: composition
    registry_contexts: 
      - first-gcr
      - second-gcr
    arguments:
      composition:
        version: '2'
        services:
          db:
            image: postgres
      composition_candidates:
        test_service:
          image: 'alpine:3.9'
          command: printenv
          working_dir: /app
          environment:
            - key=value

Let’s look at an example. We assume that we have two GCR integrations:

The first integration is the “production” one and the second one is the “staging” one. The production one is designated as primary. This means that by default even though both integrations work for the gcr.io domain, only the primary one will be used in Codefresh pipeline for pulling images.

Let’s say however that you want to build a Docker image that has a FROM statement from an image that exists in the staging registry. The image should be pushed to the production registry. You can use the registry_context property as shown below:

  build:
    title: "Building Docker image"
    type: "build"
    image_name: "gcr.io/production-project/my-image"
    working_directory: "${{clone}}"
    tag: "latest"
    dockerfile: "Dockerfile"
    stage: "build"
    registry: production-gcr
    registry_contexts: # define what registries will be used for pulling images
      - staging-gcr

Behind the scenes Codefresh will

1. First login to the “staging” Docker registry using the “staging” credentials
2. Build the docker image, by resolving the FROM statements with “staging” images, pulling them as needed using the staging credentials
3. Tag the Docker image
4. Login to the “production” Docker registry
5. Push the final Docker image to the “production” registry.

If your primary Docker registry is also the one that is used in your pipelines, you don’t need the registry_context property at all, as this is the default behavior. When searching for an image to pull Codefresh will look at all your Docker registries (if they manage only a single domain), plus your “primary” Docker registries in case you have multiple Docker registries for the same domain.

Promoting Docker images

Apart from building and pushing a brand new docker image, you can also “promote” a docker image by copying it from one registry to another. You can perform this action either from the Codefresh UI or automatically from pipelines.

Promoting images via the Codefresh UI

You have the capability to “promote” any image of your choosing and push it to an external Registry that you have integrated into Codefresh (such as Azure, Google, Bintray etc)

Visit the Docker image dashboard and click the Promote button for the image you wish to promote:

You will get a list of your connected registries. Choose the target Registry and define the tag that you want to push. Then click the Promote button to “copy” this image from its existing registry to the target Registry.

Promoting images in pipelines

You can also copy images from one registry to the other inside a pipeline. This is accomplished by specifying an existing image in the candidate field of the push step.

For example:

codefresh.yml

  promote_to_production_registry:
    title: Promoting to Azure registry 
    type: push
    candidate: us.gcr.io/project-k8s-sample-123454/my-golang-app
    tag: '1.2.3'
    # Unique registry name
    registry: azure-demo      

In the example above we promote an image from GCR to ACR (which is already setup as azure-demo).

You can even “promote” docker images within the same registry by simply creating new tags. For example:

codefresh.yml

  promote_to_production:
    title: Marking image with prod tag
    type: push
    candidate: my-azure-registry.azurecr.io/kostis-codefresh/my-python-app:1.2.3
    tag: 'production'
    # Unique registry name
    registry: azure-demo      

In the example above the image my-azure-registry.azurecr.io/kostis-codefresh/my-python-app:1.2.3 is re-tagged as my-azure-registry.azurecr.io/kostis-codefresh/my-python-app:prod. A very common pattern is to mark images with a special tag such as prod after the image has been deployed successfully.

What to read next

• Push pipeline step
• External Docker Registries
• Accessing a Docker registry from your Kubernetes cluster
• Build and Push an image example