Registries - Use & manage your registries

Objective

This guide is intended to help you to use and manage your public and private registries through ovhai command line interface (AI CLI), API and OVHcloud Control Panel. This will help you when working with OVHcloud AI Tools, especially with AI Training and AI Deploy.

Requirements

• A Public Cloud project
• Access to the OVHcloud Control Panel

Instructions

This documentation is divided into the following parts:

• Manage existing registries through
  • OVHcloud Control Panel (UI)
  • OVHcloud AI CLI ovhai
• Create, configure and use new registries
  • Private Harbor registry
  • Private Docker registry
  • GitHub registry (packages)

Manage existing registries

The Public Cloud provides a default registry called shared registry, where users of a same Public Cloud project are able to push their custom images, and use them through our AI Tools.

This is why it can be interesting to add and manage other registries. We can either do this by using the OVHcloud Control Panel (UI) or the ovhai CLI.

Manage existing registries through UI

First, log in to the OVHcloud Control Panel and go to the Public Cloud section. Then go to the AI Training section, which is located under AI & Machine Learning.

By clicking the Private Docker Registry button, you should be able to see and manage (add, delete) your different private registries.

Note that the shared registry will not appear here. This one is displayed in the Home panel of AI Training.

Manage existing registries through CLI

If you prefer, you can also use the OVHcloud ovhai CLI:

List all your available registries with:

ovhai registry list

Add a registry by using:

ovhai registry add <registry-url>

Update one of your registries with the following command:

ovhai registry update <id>

Delete a registry by using:

ovhai registry delete <id>

If you do not have our ovhai CLI yet, learn how to install it here.

Create, configure and use new registries

Using your own and private registries is the best way to use your images without exposing them to everyone. We distinguish 3 main possibilities:

• Private Harbor registry
• Private Docker registry
• GitHub registry / packages

We will see how to implement each of these registries and use them through AI Tools.

Harbor private registry

OVHcloud Managed Private Registry service allows you to spawn your own authenticated Docker registry where you can privately store your Docker images. This registry is an instance of Harbor, an open source trusted cloud native registry project that stores, signs, and scans Docker images.

Creation

To create your new private registry, we advise you to read and follow the steps mentioned in the Creating a private registry documentation.

At the end of this step, you should have your registry identification details.

Configuration

Then, you will need to use these credentials to connect to the Harbor UI. This will allow you to manage this private registry. This simple step is explained in the Connecting to the UI documentation.

To finish setting up your private Harbor registry, you will need to create a project on Harbor, and associate a user to it (Administrator user is added by default, but you can create a new one for this project). You will find all needed information in this documentation.

Get your OVHcloud Managed Private Registry API URL

In order to add this registry to AI Tools, you will need to retrieve its URL. To do this, go to the Managed Private Registry section on the OVHcloud Public Cloud Manager, and in the "more options" button (...) at the right, click on Harbor API:

Then, copy the URL of the Harbor API, which is the URL of your private registry:

Add the Harbor registry

To add your registry, you can either use the OVHcloud Control Panel (UI) or the ovhai CLI.

During this step, you will be asked your user's credentials (user ID and password). You can use the default user (administrator) credentials, which were obtained at the first part of the Create and configure a new private registry step, or those of another user if you have created one.

ovhai registry add <registry-url>

Once your registry is added, you should see it in your registries list. You can then use any image pushed on this registry for your AI Training jobs and AI Deploy apps.

Push an image to your registry

Log in to your private registry, using a user with write rights to your registry project, with the following command:

docker login -u <user> -p <password> <private-registry-address>

Once connected, all you have to do is build, tag, and push your image. Here is an example:

Build the image from the current directory's Dockerfile:

docker build . -t <image-name>:<tag-name>

Tag the compiled image:

docker tag <image-name>:<tag-name> <private-registry-address>/<harbor_project_name>/<image-name>:<tag-name>

Push the image to your private registry's project:

docker push <private-registry-address>/<harbor_project_name>/<image-name>:<tag-name>

Use your private registry images with OVHcloud AI Tools

Now that your private registry has been created, added to your Public Cloud account, you can use its image(s) in AI Training and AI Deploy.

Here is a basic example with AI Training, when using the CLI:

ovhai job run <private-registry-address>/<harbor_project_name>/<image-name>:<tag-name>

If you prefer to use the OVHcloud Control Panel, this is also possible. In this case, you will have to indicate your custom Docker image during Step 2 of your job/app creation. This path will be <private-registry-address>/<harbor_project_name>/<image-name>:<tag-name>.

Docker private registry

If you prefer to use a private registry from your Docker account, it is also possible.

Let's say we have created a private repository on our Docker account.

We build, tag and push our Docker image in this private repository:

docker build . -t <image-name>:<tag-name>
docker tag <image-name>:<tag-name> <docker_hub_username>/<private_repository_name>:<tag-name>
docker push <docker_hub_username>/<private_repository_name>:<tag-name>

Add the private Docker registry

Once your image is pushed to your Docker account, you will need to add your private Docker registry to the registries of your Public Cloud project.

As with Harbor, this can either be done with UI or CLI. This time, your private registry's URL will be index.docker.io and the requested username and password correspond to those of your Docker account.

During this step, you will be asked your Docker credentials.

Using UI

Log in to the OVHcloud Control Panel, go to the Public Cloud section, then to the AI Dashboard section which is located under AI & Machine Learning.

By default, you will be redirected to the Dashboard panel. To add a custom registry URL, click the Docker Registries button. Once there, click the + Add button to add your private Docker registry.

Here, you will need to provide your Docker credentials along with its URL.

When you have finished filling in the form, click Add.

Using CLI

To add your private Docker registry via CLI, use:

ovhai registry add index.docker.io

Once your registry is added, you should see it in your registries list.

Use your private registry images with OVHcloud AI Tools

You can now use the images contained in this private Docker registry in AI Tools by specifying index.docker.io. Here is an example to deploy a Streamlit application on AI Deploy:

ovhai app run index.docker.io/<docker_hub_username>/<private_repository_name>:<tag-name> \
--default-http-port 8501

You can also use the Control Panel to deploy a job or an app from your private Docker registry. You will just need to specify your Docker image path, during step 2 of your job/app creation. This path will be index.docker.io/<docker_hub_username>/<private_repository_name>:<tag-name>.

GitHub registry

You can also use a GitHub registry, also known as GitHub Packages.

As indicated on their documentation, you can build and push a package image by using the following commands:

docker login ghcr.io --username <your_username> --password <your_token>
docker tag app ghcr.io/<your_username>/<repo-name>:<tag-name>
docker push ghcr.io/<your_username>/<repo-name>:<tag-name>

Add a GitHub registry

Once your image is pushed on your GitHub account, you will need to add your GitHub registry to the registries of your Public Cloud project. As with Harbor and Docker, this can either be done with UI or CLI. This time, your registry's URL will be ghcr.io.

You will be asked your GitHub credentials.

ovhai registry add ghcr.io

Use your GitHub registry images with OVHcloud AI Tools

You can now use the packages/images contained in this GitHub registry in AI Tools by specifying ghcr.io. Here is an example to deploy a Streamlit application on AI Deploy:

ovhai app run ghcr.io/<your_username>/<repo-name>:<tag-name> \
--default-http-port 8501

You can also use the Control Panel to deploy a job or an app from a GitHub registry. You will just need to specify your GitHub image path, during step 2 of your job/app creation. This path will be ghcr.io/<your_username>/<repo-name>:<tag-name>.

Go further

• You can check the official documentation about how to submit a job.
• Learn how to deploy an app.
• You can check out the documentation about the ovhai CLI.

If you need training or technical assistance to implement our solutions, contact your sales representative or click on this link to get a quote and ask our Professional Services experts for a custom analysis of your project.

Feedback

Please send us your questions, feedback and suggestions to improve the service:

• On the OVHcloud Discord server