Install Altair AI Hub
Altair AI Hub has adopted a deployment architecture consisting of several components, to ensure scalability and flexibility. We have packaged the components as a set of docker images. These images can be used to deploy the platform using the Docker-compose or Kubernetes orchestration technologies, either on-prem, or in your preferred (public or private) cloud infrastructure.
The quickest and easiest way to deploy Altair AI Hub is via our cloud images, which are available on AWS and Microsoft Azure. We recommend this deployment method mainly for development and testing purposes.
If you would like to deploy with a specific use-case in mind, and the cloud images are not suitable for your needs, here is the reading order we recommend.
- First, take the time to familiarize yourself with a high level overview of the platform deployment architecture. Here you'll learn the components used in the platform and their role. Please also read our security overview page.
- Then, check out our docker-compose templates or kubernetes templates, depending on your use-case. This allows you to select a template which is closest to your needs. Our templates are designed to work out of the box.
- If you need to fine-tune parts of the deployment, visit our docker image reference for detailed configuration possibilities.
- Learn to operate your deployment by reading Manage Altair AI Hub.
- Find changelog by reading Changlog
Docker image security
Altair AI Hub is shipped as a set of docker images which will be deployed on a Docker or Kubernetes environment. Each docker image is tested against the known vulnerabilities using the standard tools (grype, trivy):
* Majority of the components are based on the most recent Ubuntu base Docker image, that gets a package upgrade and then the software installation.
* For the 3rd party components (e.g. keyCloak, Postgres) we repackage and ship their offical docker images and if it is possible, we apply all the necessary package upgrades to decrease the risk of vulnerabilities.
With the release of 2024.0.0 besides rapidminer/postgres-10 and rapidminer/postgres-14, we published the images built using the Postgresql version 15 and 16.
Starting with the 2024.0 release Alpine-based Postgresql images are introduced as default Postgresql images, the previously used Debian-based images are available with "-debian" suffix.
Please note, we still refer the same postgresq versions in our templates but provide an option for our customers to migrate to the new version for existing installations.
For new installations please use the latest available version of Postgresql.
Altair AI Hub docker images
If you are deploying Altair AiHub in an air-gapped environment, you can download the necessery docker image exports by substituting aihub_version in following command.
wget https://s3.us-east-1.amazonaws.com/rapidminer.releases/docker-images/<aihub_version>/download-docker-images.sh
Note: This script is available for versions starting with 10.3.1.
Once downloaded, you can run the script with following command to get docker images tar.gz files
./download-docker-images.sh
Upgrade notes
- In the release 2024.0 the Jupyterhub database backend got upgaded to Postgresql 16 that is a breaking change In general Postgresql major version upgrade require migration. To do that please refer to Postgresql documentation.
Jupyterhub stores information about users, user sesssions and notebook states in this database. In case the opened sessions of the users, spwawn history, notebook state are not necessary, you can decide to stop jupyterhub database container, drop the underlying volume and start the container again with the new version. The container will start with an empty volume. User records will be recreated after they log in on Keycloak.