Skip to main content

How to Build and push with non-root users

Introduction

Building with non-root users may be necessary for an organization due to security or compliance requirements. This page describes how to build and push Docker images as a non-root user, and outlines Harness options and best practices for Kubernetes-based builds.


Kaniko Builds

By default, Harness uses Kaniko to build and push images in Kubernetes cluster build infrastructure.

note

We will support a maintained version of Kaniko as our default. More information to come.

  • Kaniko does not require privileged mode (privileged: false).
  • Kaniko always runs as root (rootless mode is not supported).
  • If your policy requires non-root for builds: Kaniko will not meet that requirement. See Buildah below.
  • Kaniko is recommended for most security-conscious teams as it minimizes risk by not requiring privileged pods.

BuildX Builds

You may also use BuildX by enabling the CI_USE_BUILDX_ON_K8 flag via Harness Support.

  • BuildX requires both root user and privileged mode in Kubernetes.
  • This tool provides the richest feature set, but is the least restrictive in terms of privileges, and is not recommended where least-privilege is a hard requirement.

Drone Docker

You can also use drone-docker for image builds as a plugin step.

  • This method is generally used for backwards compatibility or in specific CI workflows.
  • Privilege and user requirements will be similar to Docker/BuildX (privileged and root).

Individual Step Root / Buildah Plugin

If your Kubernetes cluster build infrastructure is configured to run as non-root (runAsNonRoot: true and runAsUser: 1000), you have two options for building images:

  1. Enable root access for individual steps (if allowed by policy):
    If your build infrastructure is set up for non-root execution (runAsNonRoot: true), but your security policy allows Build and Push to run as root, set Run as User to 0 for that step only. This enables only that step to run as root.

    If your security policy forbids root user entirely for all steps, use the Buildah plugin running as a non-root user, with privileged mode enabled.

  2. Use the Buildah plugin:
    If your policy does not allow root for any step, use the Buildah plugin in a Plugin step, and specify a non-root UID (runAsUser: 1000). Buildah can run as a non-root user (e.g., UID 1000). However, it still requires privileged mode in Kubernetes (except for some OpenShift SCC scenarios). This is due to kernel-level operations required by Buildah, even in rootless mode.

    tip

    Buildah requires special configuration for OpenShift clusters (use the anyuid SCC and privileged: false).
    For standard Kubernetes clusters, privileged: true is required even for non-root user.

    To use the Buildah plugin, your pipeline must meet these requirements:


Add a Plugin step

At the point in your pipeline where you want to build and upload an image, add a Plugin step that uses the buildah plugin, for example:

                - step:
type: Plugin
name: buildah-docker
identifier: buildahdocker
spec:
connectorRef: YOUR_IMAGE_REGISTRY_CONNECTOR
image: plugins/buildah-docker:1.1.0-linux-amd64
privileged: true # false for OpenShift with anyuid SCC
settings:
repo: myDockerHub1234/test
tags: buildahtest
registry: https://index.docker.io/v2/
dockerfile: Dockerfile
username: <+secrets.getValue("DOCKER_HUB_USER")>
password: <+secrets.getValue("DOCKER_HUB_SECRET")>
runAsUser: "1000"

This step requires the following specifications:

  • connectorRef: Provide the ID of a connector corresponding to your push destination.
    • Docker Hub: Docker connector
    • ACR: Azure connector
    • ECR: AWS connector
    • GAR/GCR: GCP connector
  • image: Specify the plugin image and tag to use, such as plugins/buildah-docker:1.1.0-linux-amd64. If you don't specify a tag, the latest tag is used by default. Go to an image's Docker Hub page to browse available tags:
  • privileged: Set to false for OpenShift clusters. Set to true for non-OpenShift clusters.
  • runAsUser: Specify the ID of the non-root user to use for this step, such as 1000.
  • settings: Add the following settings as key-value pairs.
    • repo: The name of the repository where you want to store the image, for example, <hub-user>/<repo-name>. For private registries, specify a fully qualified repo name.
    • tags: Specify tags for your image.
    • registry: Specify the registry index, such as https://index.docker.io/v2/. The registry format for ECR is aws_account_id.dkr.ecr.region.amazonaws.com.
    • dockerfile: Specify the Dockerfile to use for the build.
    • username: Provide the username to access the push destination, either as plaintext or an expression referencing a Harness secret or pipeline variable, such as <+pipeline.variables.DOCKER_HUB_USER>.
    • password: An expression referencing a Harness secret or pipeline variable containing the password to access the push destination, such as <+pipeline.variables.DOCKER_HUB_SECRET>.
    • For more information and additional settings, including AWS S3 settings, go to Buildah plugin settings.

Buildah plugin settings

For information about Buildah plugin settings, go to the Buildah README, the Buildah Drone Plugins Marketplace page, and the the main.go file for each destination:

Many Buildah plugin settings correspond with settings for the built-in Build and Push steps. If you're encountering an error with the buildah plugin configuration, you can reference the settings definitions for the built-in steps for guidance on the expected value for the equivalent Buildah settings. However, keep in mind that the configuration for Build and Push steps (such as field names and location in the YAML) is not an exact match to the Plugin step configuration.

DestinationBuildah imageBuildah main.goEquivalent Build and Push step
Docker Hubbuildah-dockerDocker main.goBuild and Push to Docker Registry
ACRbuildah-acrACR main.goBuild and Push to ACR
ECRbuildah-ecrECR main.goBuild and Push to ECR
GAR/GCRbuildah-grcGCR main.goBuild and Push to GAR/Build and Push to GCR

Stage YAML example

This YAML example shows a Build (CI) stage with a Kubernetes cluster build infrastructure running as non-root (runAsNonRoot: true and runAsUser: "1000") and a Plugin step running the Buildah plugin.

    - stage:
identifier: stage1
type: CI
name: stage1
spec:
cloneCodebase: true
infrastructure:
type: KubernetesDirect
spec:
connectorRef: YOUR_KUBERNETES_CLUSTER_CONNECTOR
namespace: YOUR_NAMESPACE
automountServiceAccountToken: true
nodeSelector: {}
containerSecurityContext:
runAsNonRoot: true
runAsUser: "1000"
os: Linux
execution:
steps:
- step:
identifier: buildah plugin
type: Plugin
name: buildah_plugin
spec:
connectorRef: YOUR_DOCKER_CONNECTOR
image: plugins/buildah-docker:1.1.0-linux-amd64
privileged: true
settings:
repo: myhub/test-repo
tags: builadhrootless
registry: https://index.docker.io/v2/
dockerfile: Dockerfile2
username: <+pipeline.variables.DOCKER_HUB_USER>
password: <+secrets.getValue("MyDockerPAT")>
runAsUser: "1000"

Build an image without pushing

You can use your CI pipeline to test a Dockerfile used in your codebase and verify that the resulting image is correct before you push it to your Docker repository.

The following configuration is valid with the Docker Buildah plugin (plugins/buildah-docker) only. For other configurations, go to Build without pushing.

  1. In your CI pipeline, go to the Build stage that includes the Plugin step with the Docker Buildah plugin.
  2. In the Build stage's Overview tab, expand the Advanced section.
  3. Select Add Variable and enter the following:
    • Name: PLUGIN_DRY_RUN
    • Type: String
    • Value: true
  4. Save and run the pipeline.