Wave allows provisioning container images on-demand, removing the need to build and upload them manually to a container registry.

Containers provisioned by Wave can be both disposable, i.e. ephemeral containers only accessible for a short period of time, and regular long-term registry-persisted container images.

• Authenticate the access to remote container registries;
• Augment container images i.e. dynamically add one or more container layers to existing images;
• Build container images on-demand for a given container file (aka Dockerfile);
• Build container images on-demand based on one or more Conda packages;
• Build container images on-demand based on one or more Spack packages;
• Build container images for a specified target platform (currently linux/amd64 and linux/arm64);
• Push and cache built containers to a user-provided container repository;
• Build Singularity native containers both using a Singularity spec file, Conda package(s) and Spack package(s);
• Push Singularity native container images to OCI-compliant registries;

Container provisioning requests are submitted via the endpoint POST /container-token specifying the target container image or container file. The endpoint return the container name provisioned by Wave that can be accessed by a regular Docker client or any other container client compliant via Docker registry v2 API.

When a disposable container is requested, Wave acts as a proxy server between the Docker client and the target container registry. It instruments the container manifest, adding on-demand the new layers specified in the user submitted request. Then the pull flow is managed by Docker client as in any other container, the base image layers are downloaded from the container registry where the image is stored, while the instrumented layers are downloaded via the Wave service.

• Java 19 or later
• Linux or macOS
• Docker engine (for development)
• Kubernetes cluster (for production)

1. Clone the Wave repository from GitHub:

   git clone https://github.com/seqeralabs/wave && cd wave

2. Define one of more of those environment variable pairs depending the target registry you need to access:

   export DOCKER_USER="<Docker registry user name>"
   export DOCKER_PAT="<Docker registry access token or password>"
   export QUAY_USER="<Quay.io registry user name or password>"
   export QUAY_PAT="<Quay.io registry access token>"
   export AWS_ACCESS_KEY_ID="<AWS ECR registry access key>"
   export AWS_SECRET_ACCESS_KEY="<AWS ECR registry secret key>"
   export AZURECR_USER="<Azure registry user name>"
   export AZURECR_PAT="<Azure registry access token or password>"

3. Setup a local tunnel to make the Wave service accessible to the Docker client (only needed if you are running on macOS):

   npx localtunnel --port 9090

   Then configure in your environment the following variable using the domain name return by Local tunnel, e.g.

   export WAVE_SERVER_URL="https://sweet-nights-report.loca.lt"

4. Run the service in your computer:

   bash run.sh
   

5. Submit a container request to the Wave service using the curl tool

   curl \
       -H "Content-Type: application/json" \
       -X POST $WAVE_SERVER_URL/container-token \
       -d '{"containerImage":"ubuntu:latest"}' \
       | jq -r .targetImage

6. Pull the container image using the name returned in the previous command, for example

   docker pull sweet-nights-report.loca.lt/wt/617e7da1b37a/library/ubuntu:latest

Note You can use the Wave command line tool instead of curl to interact with the Wave service and submit more complex requests.

• To debug http requests made proxy client add the following Jvm setting:

  '-Djdk.httpclient.HttpClient.log=requests,headers'

• Wave command line tool
• Distribution (formerly Registry) API