Skip to main content
Version: v3.6.1

Deployment

Deploy H2O Admin Analytics app in HAIC environment​

This guide will assist you in deploying the H2O Admin Analytics app in the H2O AI Cloud (HAIC) environment.

Prerequisites​

Before starting the deployment, ensure the following tools are installed and configured:

  1. kubectl: Install and configure kubectl for your Kubernetes cluster.
  2. H2O CLI: Install and configure H2O CLI.

Pre-deployment configurations​

The setup.sh script will download TLS certificates from your Kubernetes cluster, and create or update all necessary H2O secrets.

  1. Execute the setup.sh script with the appropriate parameters. Use the following command format:

    ./scripts/setup.sh [MLOPS_PREFIX] [MLOPS_K8_NAMESPACE] [H2O_CLI_CONFIG] [RESTRICT_FOR_ADMINS] [ALLOWED_USER_ROLE_KEY](Optional) [ALLOWED_USER_ROLE](Optional)

    Parameters:

    • MLOPS_PREFIX - Prefix for your mlops installation
    • MLOPS_K8_NAMESPACE - Kubernetes namespace for MLOps in your installation
    • H2O_CLI_CONFIG - Path to H2O CLI configuration file
    • H2O_WAVE_PING_INTERVAL - how often should ping messages be sent via wave SDK
    • RESTRICT_FOR_ADMINS - True if you need to restrict the app only for admins else False
    • ALLOWED_USER_ROLE_KEY - Allowed user role (This is the value of Token Claim Name) value if RESTRICT_FOR_ADMINS is True else not needed
    • ALLOWED_USER_ROLE_VALUE - Allowed user role (This the name of actual user group) value if RESTRICT_FOR_ADMINS is True else not needed

Automated Docker image building and publishing​

GitHub workflows now automatically build and publish Docker images, streamlining the deployment process. This removes the need to manually create and push images.

Deploy the application​

Once the configuration is successful, deploy the application using the H2O CLI:

h2o bundle import -v ALL_USERS
note

Memory optimization: To enhance the performance of your deployments, memory optimization techniques have been implemented. This includes clearing unused variables to improve memory footprint during deployments.

Feedback