Kubernetes cluster requirements¶
This section describes the Kubernetes cluster requirements for running the DataRobot platform with supported Kubernetes versions. Refer to your installation plan to find out the exact Kubernetes cluster and other prerequisites for your environment.
DataRobot supported Kubernetes versions¶
DataRobot supports a rolling window of major Kubernetes releases to ensure that the latest version of the DataRobot platform can be installed on your chosen hosting provider. The table below lists the Kubernetes versions that DataRobot supports:
| DataRobot release | Kubernetes versions |
|---|---|
9.1.x |
1.25, 1.26 |
9.2.x |
>=1.26 |
10.0.x |
>=1.27 |
10.1.x |
>=1.27 |
10.2.x |
>=1.28 |
11.0.x |
>=1.29 |
11.1.x |
>=1.29 |
| --- |
Install the prereqchecker tool¶
DataRobot optionally provides a CNCF-compliant Sonobuoy Plug-in CLI tool. You can use this self-service tool to analyze your target Kubernetes environment to verify that it meets the prerequisites for installing the DataRobot platform. Run this tool before beginning the installation.
OCI registry¶
DataRobot requires a OCI registry that has basic authentication and TLS configured.
Ingress¶
DataRobot platform requires installation of the nginx Ingress Controller.
For reference, see the Ingress Settings section.
FQDN and DNS¶
DataRobot platform requires a DNS setup with a fully-qualified domain name (FQDN) so that your users can navigate to the application (e.g., https://datarobot.company.net).
A common approach is to set up a LoadBalancer service and point a DNS CNAME record to it. The LoadBalancer then sends traffic to the ingress-nginx controller.
TLS¶
The Docker registry and DataRobot platform both require valid TLS certificates signed by a publicly-known certificate authority.
DataRobot also supports using cert-manager to acquire valid certificates. For detailed guidance on configuring ingress settings, refer to the Ingress Settings section. Additionally, for instructions on configuring cert-manager, consult the LetsEncrypt settings for TLS termination section.
Persistent storage¶
DataRobot platform requires a perssitent storage deployed with default StorageClass configured.
Ephemeral volumes¶
The majority of DataRobot pods use ephemeral volumes.
Network policy engine¶
Features on the DataRobot platform, such as custom models and custom tasks, require a network policy engine to enforce Kubernetes Network Policies. While not required for base installations, a policy engine is recommended for ensuring custom model pod security.