Skip to content

Latest commit

 

History

History
35 lines (32 loc) · 1.38 KB

README.md

File metadata and controls

35 lines (32 loc) · 1.38 KB

ML AKS Clients

This repository provides tools for setting up and managing AKS clients for a Weka cluster.

The scripts and Terraform configurations streamline the creation of AKS clusters, ML workspaces, and the installation of Weka clients on AKS nodes

How to run

  • run cd azure/ml-aks
  • Optional:
    • Create AKS cluster with dedicated client node pool
      • run cd aks
      • provide the required variables:
        • rg_name
        • vnet_name
        • subnet_name
        • ssh_public_key
        • subscription_id
        • prefix
        • client_frontend_cores
      • run terraform init
      • run terraform apply
    • Create ML workspace
      • run cd ml
      • provide the required variables:
        • subscription_id
        • rg_name
        • key_vault_name
        • prefix
      • run terraform init
      • run terraform apply
  • Setup UDP WEKA clients on AKS nodepool:
    • Run ./set_env_vars.sh && source env_vars
    • Run ./run.sh <weka_backend_ip>

This script will update the aks clients with the weka cluster configuration, and will install the weka client on the aks clients On each AKS client node, a DaemonSet ensures the creation of a pod named installer-init-xxx. This pod runs a specialized UDP mount script that seamlessly connects the node to the Weka cluster.