Skip to content

Latest commit

 

History

History
94 lines (66 loc) · 3.36 KB

README.md

File metadata and controls

94 lines (66 loc) · 3.36 KB

Marqo Logo

Deploying Marqo on Azure Kubernetes Service (AKS)

This guide provides instructions on how to deploy Marqo on Azure AKS using Kubernetes and Helm. The deployment process involves creating a resource group, an AKS cluster, configuring node pools, and deploying Marqo using a Helm chart.

Prerequisites

  • An Azure account.
  • Azure CLI installed and configured.
  • Kubernetes command-line tool (kubectl) installed.
  • Helm installed.

Notes

Before executing the setup script, ensure your Azure subscription has enough capacity for the following requirements:

  • vCPUs: Verify the quota for vCPUs based on the node size and count you plan to use. Default setup requires 18 vCPU of type Standard_D2s_v3.
  • Persistent Disk SSD: At least 240 GB required; adjust based on your setup.
  • If using GPU nodes, ensure your subscription supports the desired Azure VM sizes with GPU capabilities, recommended GPU instance is Standard_NC4as_T4_v3.

Deployment Steps

  1. Run the Setup Script: First, log in to Azure CLI (if not already logged in):

    az login

    Set your Azure subscription ID:

    export AZURE_SUBSCRIPTION_ID='<your-subscription-id>'
az account set --subscription $AZURE_SUBSCRIPTION_ID

    Customize the following variables in the setup.sh script to match your Azure environment:

    • APP_INSTANCE_NAME: Name your application instance.
    • CLUSTER: Define your Kubernetes cluster name.
    • LOCATION: Specify your preferred Azure region such as eastus.
    • RESOURCE_GROUP: Set the name for the new or existing resource group.
    • INSTALL_GPU: If deploying with GPU support, set INSTALL_GPU=true otherwise false.

    Run the script:

    ./AKS/setup_aks.sh

    This script performs the following actions:

    • Creates a resource group (if it doesn't already exist).
    • Creates an AKS cluster with specified configurations.
    • Configures various node pools for different purposes, including optional GPU nodes.
    • Applies necessary Kubernetes configurations.

  2. Kubernetes Cluster Validation: After the setup.sh script execution completes, it may take a few minutes for the AKS cluster and its resources to become fully operational. To verify the health of the Marqo pod in the cluster, use:

    kubectl get pods --all-namespaces -w

    Ensure that all pods reach the Running state before proceeding.

  3. Test Marqo Endpoint:

    Retrieve the IP address of your Marqo service:

    export MARQO_CLUSTER_IP=$(kubectl get svc marqo -o jsonpath='{.status.loadBalancer.ingress[0].ip}')

    Set up a Python virtual environment and test the Marqo deployment:

    python3 -m venv .venv
source .venv/bin/activate
pip install marqo
python test_marqo.py

Additional Configuration

  • GPU Nodes: If you plan to use GPU nodes, uncomment the relevant section in the setup.sh script before running it. This will create a node pool with GPU capabilities.

  • Custom Configurations: Modify the Helm chart values or the setup.sh script as needed for custom configurations.

Cleanup

To delete the AKS cluster and clean up the resources, run:

az aks delete --name $CLUSTER --resource-group $RESOURCE_GROUP --yes --no-wait

Replace $CLUSTER and $RESOURCE_GROUP with the names you have used during setup.