Skip to main content

Harness CI pipeline execution errors with the delegate in an EKS cluster and the AWS connector using IRSA

Conditions​

  • Module: Harness CI
  • Environment:
    • Infrastructure: Kubernetes
    • OS: Linux
  • Delegate location: EKS cluster
  • AWS connector configuration: Uses IRSA

Problem​

The pipeline can pull images from private ECR repositories, but when you try to push an application image with the Build and Push to ECR step, you get the following error:

+ /kaniko/executor --dockerfile=/harness/docker/Dockerfile --context=dir:///harness/ \
--destination=<aws-account-id>.dkr.ecr.sa-east-1.amazonaws.com/hermod:latest --snapshotMode=redo \
--digest-file=/kaniko/digest-file
error checking push permissions -- make sure you entered the correct tag name, \
and that you are authenticated correctly, and try again: \
checking push permission for "<aws-account-id>.dkr.ecr.sa-east-1.amazonaws.com/hermod:latest": \
Post "https://<aws-account-id>.dkr.ecr.sa-east-1.amazonaws.com/v2/hermod/blobs/uploads/": EOF
exit status 1

Solution​

  1. Configure the Service account in the Kubernetes cluster build infrastructure settings. For example:

    infrastructure:
      type: KubernetesDirect
      spec:
        connectorRef: test-connector
        namespace: harness-delegate-ng
        serviceAccountName: **harness-delegate-sa**
        automountServiceAccountToken: true
        nodeSelector: {}
        os: Linux

  2. Before your Build and Push to ECR step, add a Run step that runs the following command:

    aws ecr get-login-password --region <aws_region> | docker login --username AWS --password-stdin <aws_account_id>.dkr.ecr.<aws_region>.amazonaws.com