Configuring a ROSA cluster to pull images from AWS Elastic Container Registry (ECR)
Prerequisites
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.
Add the
AmazonEC2ContainerRegistryReadOnly
policy to theManagedOpenShift-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
Log into ECR
aws ecr get-login-password --region region | docker login --username AWS \ --password-stdin aws_account_id.dkr.ecr.region.amazonaws.com
Create a repository
aws ecr create-repository \ --repository-name hello-world \ --image-scanning-configuration scanOnPush=true \ --region region
Pull an image
docker pull openshift/hello-openshift
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
Push the image to ECR
docker push <registry id>.dkr.ecr.<region>.amazonaws.com/hello-world:latest
Create a new project
oc new project hello-world
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
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
Simply delete the project you created to test pulling images:
oc delete project hello-world
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.