Configuring a ROSA cluster to pull images from AWS Elastic Container Registry (ECR)

Prerequisites

• AWS CLI
• Openshift CLI 4.8+
• Docker

Background

Quick Introduction by Ryan Niksch & Charlotte Fung on YouTube.

There are two options to use to authenticate wth Amazon ECR to pull images.

The traditional method is to create a pull secret for ecr.

Example:

oc create secret docker-registry ecr-pull-secret  \
  --docker-server=<registry id>.dkr.ecr.<region>.amazonaws.com  \
  --docker-username=AWS --docker-password=$(aws ecr get-login-password) \
  --namespace=hello-world

However Amazon ECR tokens expire every 12 hours which will mean you will need to re-authenticate every 12 hours either through scripting or do so manually.

A second, and preferred method, is to attach an ECR Policy to your cluster’s worker machine profiles which this guide will walk you through.

Attach ECR Policy Role

You can attach an ECR policy to your cluster giving the cluster permissions to pull images from your registries. ROSA worker machine instances comes with pre-defined IAM roles, named differently depending on whether its a STS cluster or a non-STS cluster.

STS Cluster Role

ManagedOpenShift-Worker-Role is the IAM role attached to ROSA STS compute instances.

non-STS Cluster Role

<cluster name>-<identifier>-worker-role is the IAM role attached to ROSA non-STS compute instances.

Tip: To find the non-STS cluster role run the following command with your cluster name:

aws iam list-roles | grep <cluster_name>

ECR Policies

ECR has several pre-defined policies that give permissions to interact with the service. In the case of ROSA, we will be pulling images from ECR and will only need to add the AmazonEC2ContainerRegistryReadOnly policy.

1. Add the AmazonEC2ContainerRegistryReadOnly policy to the ManagedOpenShift-Worker-Role for STS clusters or the <cluster name>-<identifier>-worker-role for non-STS clusters.

   STS Example:

    aws iam attach-role-policy \
     --role-name ManagedOpenShift-Worker-Role \
     --policy-arn "arn:aws:iam::aws:policy/AmazonEC2ContainerRegistryReadOnly"
   

Test it Out

1. Log into ECR

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

2. Create a repository

   aws ecr create-repository \
    --repository-name hello-world \
    --image-scanning-configuration scanOnPush=true \
    --region region
   

3. Pull an image

   docker pull openshift/hello-openshift
   

4. Tag the image for ecr

   docker tag openshift/hello-openshift:latest <registry id>.dkr.ecr.<region>.amazonaws.com/hello-world:latest
   

   note: you can find the registry id and URI with the following command

   aws ecr describe-repositories
   

   
   
   

5. Push the image to ECR

   docker push <registry id>.dkr.ecr.<region>.amazonaws.com/hello-world:latest
   

6. Create a new project

   oc new project hello-world
   

7. Create a new app using the image on ECR

   oc new-app --name hello-world --image <registry id>.dkr.ecr.<region>.amazonaws.com/hello-world:latest
   

8. View a list of pods in the namespace you created:

   oc get pods
   

   Expected output:

   

   If you see the hello-world pod running … congratulations! You can now pull images from your ECR repository.
   
   

Clean up

1. Simply delete the project you created to test pulling images:

   oc delete project hello-world
   

2. You may also want to remove the arn:aws:iam::aws:policy/AmazonEC2ContainerRegistryReadOnly policy from the worker nodes if you do no want them to continue to have access to the ECR.