This repository contains tools to maintain a static container registry using nerdctl, a Docker-compatible CLI for containerd.

It targets an AWS-compatible S3 storage account and represents the path of least resistance on my setup – running MacOS on ARM M1 with containerd on my development system and using Kubernetes on DigitalOcean to run the container workloads.

To learn why/when you would use this, see my post container registry on a budget.

Jérôme Petazzoni's registrish uses skopeo to extract image data and supports additional targets. I don't use Docker and therefore couldn't use the docker-daemon scheme with skopeo to work with local images.

Install the AWS CLI command-line tools. On MacOS you can use the awscli Homebrew formula.

Clone this repository or copy the main branch into your current working directory using:

curl -L https://github.com/malthe/static-container-registry/tarball/main | \
    tar --strip-components=1 -xz

The starting point is to build or pull down an image locally.

For example, we can pull down Google's hello world example. For convenience and because the upload script relies on them, we'll define environment variables IMAGE and TAG first:

$ export IMAGE=gcr.io/google-samples/node-hello TAG=1.0
$ nerdctl pull $IMAGE:$TAG

Save (export) this image to a temporary directory:

$ mkdir tmp
$ nerdctl save $IMAGE:$TAG | tar xv -C tmp

Define the following environment variables to prepare for the upload.

$ export AWS_ACCESS_KEY_ID=<access-key-id>
$ export AWS_SECRET_ACCESS_KEY=<secret-access-key>

If using a non-AWS service such as DigitalOcean Spaces, additionally set the region and endpoint:

$ export AWS_DEFAULT_REGION=<region>
$ export ENDPOINT=--endpoint-url https://$AWS_DEFAULT_REGION.digitaloceanspaces.com

That's it. We are ready to upload:

$ ( cd tmp ; ../upload.sh )

You can confirm that the registry was prepared correctly by pulling down the image. The exact image name will depend on which S3 service you're using.

$ nerdctl pull $BUCKET.$AWS_DEFAULT_REGION.digitaloceanspaces.com/$IMAGE:$TAG

You may need to specify your platform using --platform <name> if the image was built for a different platform than your local machine.