Build and Push to GAR

Use a CI pipeline to build and push an image to GAR.

This topic explains how to configure the Build and Push to GAR step in a Harness CI pipeline. This step is used to build and push to Google Artifact Registry (GAR).

You need:

• Access to GAR and a GAR repo.

• A Harness CI pipeline with a Build stage.

• A GCP connector.

OIDC-based authentication is supported in the Build and Push to GAR step only when using BuildX. When using Kaniko for building a Docker image, OIDC is not supported.

Kubernetes cluster build infrastructures require root access

With Kubernetes cluster build infrastructures, Build and Push steps use Chainguard's maintained Kaniko fork. Other build infrastructures use drone-docker. Kaniko can run as non-root in many cases. Root access is only required when your Dockerfile performs privileged operations (e.g., installing system packages, modifying root-owned directories).

If your build runs as non-root (runAsNonRoot: true), and you want to run the Build and Push step as root, you can set Run as User to 0 on the Build and Push step to use the root user for that individual step only.

If your security policy doesn't allow running as root, go to Build and push with non-root users.

KANIKO ROOT REQUIREMENTS

With Kubernetes infrastructure, Build and Push steps use Kaniko which requires root user access. Set runAsUser: 0 on the step only when your Dockerfile performs privileged operations like installing system packages or modifying system files. For enhanced security, keep your stage running as non-root and elevate only the Build and Push step when necessary.

Add a Build and Push to GAR step

In your pipeline's Build stage, add a Build and Push to GAR step and configure the settings accordingly.

Here is a YAML example of a Build and Push to GAR step.

- step:
    type: BuildAndPushGAR
    name: BuildAndPushGAR_1
    identifier: BuildAndPushGAR_1
    spec:
      connectorRef: YOUR_GCP_CONNECTOR_ID
      host: LOCATION-docker.pkg.dev
      projectID: GOOGLE_CLOUD_CONSOLE_PROJECT_ID
      imageName: REPO_NAME/IMAGE_NAME
      tags:
        - <+pipeline.sequenceId>

When you run a pipeline, you can observe the step logs on the build details page. If the Build and Push to GAR step succeeds, you can find the uploaded image on GAR.

You can also:

• Build images without pushing

• Build multi-architecture images

Build and Push to GAR step settings

The Build and Push to GAR step has the following settings. Depending on the stage's build infrastructure, some settings might be unavailable or optional. Settings specific to containers, such as Set Container Resources, are not applicable when using a VM or Harness Cloud build infrastructure.

Name

Enter a name summarizing the step's purpose. Harness automatically assigns an Id (Entity Identifier) based on the Name. You can change the Id.

GCP Connector

The Harness GCP connector to use to connect to GAR. The GCP account associated with the GCP connector needs specific roles. For more information, go to the Google Cloud Platform (GCP) connector settings reference.

This step supports GCP connectors that use access key authentication. It doesn't support GCP connectors that inherit delegate credentials.

If you are using this step with Harness Cloud build infrastructure, you can also leverage the OIDC connectivity mode in your GCP connector.

Host

The Google Artifact Registry hostname, for example LOCATION-docker.pkg.dev. For more information, go to the GAR documentation on Repository and image names.

Project ID

The Google Cloud Console Project ID. For more information, go to the GAR documentation on Repository and image names.

Image Name

The name of the repository where you want to push the artifact and the name you want to give the image, such as REPO_NAME/IMAGE_NAME. For more information, go to the GAR documentation on Repository and image names.

The target repository must be a standard repository.

Tags

Add Docker build tags. This is equivalent to the -t flag. For more information, go to the GAR documentation on Tagging images.

Add each tag separately.

When you push an image to a repo, you tag the image so you can identify it later. For example, in one pipeline stage, you push the image, and, in a later stage, you use the image name and tag to pull it and run integration tests on it.

Harness expressions are a useful way to define tags. For example, you can use the expression <+pipeline.sequenceId> as a tag. This expression represents the incremental build identifier, such as 9. By using a variable expression, rather than a fixed value, you don't have to use the same image name every time.

For example, if you use <+pipeline.sequenceId> as a tag, after the pipeline runs, you can see the Build Id in the output.

And this same number is applied as the tag for your image in GAR.

Later in the pipeline, you can use the same expression to pull the tagged image, such as REPO_NAME/IMAGE_NAME:<+pipeline.sequenceId>.

Base Image Connector

Select an authenticated connector to download base images from a Docker-compliant registry. If you do not specify a Base Image Connector, the step downloads base images without authentication. Specifying a Base Image Connector is recommended because unauthenticated downloads generally have a lower rate limit than authenticated downloads.

When using Base Image Connector, pushing to or pulling from multiple Docker registries with the same URL prefix (e.g., https://index.docker.io) is not supported. This limitation occurs because the second registry's credentials overwrite the first in the Docker config file. This issue doesn't affect registries with completely unique URLs, such as separate JFrog instances. This limitation does not apply to following build and push steps only on K8 - ACR, GAR, ECR.

Optimize

With Kubernetes cluster build infrastructures, select this option to enable --snapshotMode=redo. This setting causes file metadata to be considered when creating snapshots, and it can reduce the time it takes to create snapshots. For more information, go to the Kaniko documentation for the snapshotMode flag.

For information about setting other kaniko runtime flags, go to Environment variables.

Dockerfile

The name of the Dockerfile. If you don't provide a name, Harness assumes that the Dockerfile is in the root folder of the codebase.

Context

Enter a path to a directory containing files that make up the build's context. When the pipeline runs, the build process can refer to any files found in the context. For example, a Dockerfile can use a COPY instruction to reference a file in the context.

Labels

Specify Docker object labels to add metadata to the Docker image.

Build Arguments

The Docker build-time variables. This is equivalent to the --build-arg flag.

Target

The Docker target build stage, equivalent to the --target flag, such as build-env.

Docker layer caching and Remote cache image

There are two ways in which you can leverage Docker Layer Caching: Enable Docker layer caching ('caching' property) or Remote cache image ('remoteCacheRepo' property). Refer to Enable Docker layer caching for your build to learn more.

Environment Variables (plugin runtime flags)

Run as User

With Kubernetes cluster build infrastructures, you can specify the user ID to use to run all processes in the pod if running in containers. For more information, go to Set the security context for a pod.

This step requires root access. You can use the Run as User setting if your build runs as non-root (runAsNonRoot: true), and you can run the Build and Push step as root. To do this, set Run as User to 0 to use the root user for this individual step only.

If your security policy doesn't allow running as root, go to Build and push with non-root users.

Set container resources

Set maximum resource limits for the resources used by the container at runtime:

• Limit Memory: The maximum memory that the container can use. You can express memory as a plain integer or as a fixed-point number using the suffixes G or M. You can also use the power-of-two equivalents Gi and Mi. The default is 500Mi.

• Limit CPU: The maximum number of cores that the container can use. CPU limits are measured in CPU units. Fractional requests are allowed; for example, you can specify one hundred millicpu as 0.1 or 100m. The default is 400m. For more information, go to Resource units in Kubernetes.

Timeout

Set the timeout limit for the step. Once the timeout limit is reached, the step fails and pipeline execution continues. To set skip conditions or failure handling for steps, go to:

• Step Skip Condition settings

• Step Failure Strategy settings

Conditions, looping, and failure strategies

You can find the following settings on the Advanced tab in the step settings pane:

• Conditional Execution: Set conditions to determine when/if the step should run.

• Failure Strategy: Control what happens to your pipeline when a step fails.

• Use looping strategies: Define a matrix, repeat, or parallelism strategy for an individual step.

Troubleshoot Build and Push steps

Go to the CI Knowledge Base for questions and issues related to building and pushing images, such as:

• What drives the Build and Push steps? What is kaniko?

• Does a kaniko build use images cached locally on the node? Can I enable caching for kaniko?

• Can I run Build and Push steps as root if my build infrastructure runs as non-root? What if my security policy doesn't allow running as root?

• Can I set kaniko and drone-docker runtime flags, such as skip-tls-verify or custom-dns?

• Can I push without building?

• Can I build without pushing?

• Is remote caching supported in Build and Push steps?

• Why doesn't the Build and Push step include the content of VOLUMES from my Dockerfile in the final image?

• Can I use a specific version of kaniko or drone-docker?

• How do I fix this kaniko container runtime error: kaniko should only be run inside of a container?

• Can I push and pull from two different docker registries that have same prefix for registry URL ?

• Why does the parallel execution of build and push steps fail when using Buildx on Kubernetes?

• Why do Build and Push steps fail with "Error while loading buildkit image: exit status 1" when /var/lib/docker is included in shared paths during DIND execution?