azurehpc/blocks

Building Blocks

This directory contains reusable configuration file building blocks which can be used with the init-and-merge.sh script to produce a unique config file. The variables have been named to avoid collision with other blocks. Most blocks assume that a VNET exists (hence the dependency on the vnet.json block) and being installed from a jumpbox.

Block list

Name	Description	Dependency on
ad.json	Create an AD Directory Services VM named ADDS with domain hpc.local	vnet, jumpbox
anf.json	Create a Premium 4TB volume with 2x2TB pools /data and /apps	vnet
anf-smb.json	Create a Premium 4TB SMB volume with 4TB pool /data	ad, vnet
beegfs-cluster.json	Create BeeGFS cluster	jumpbox, vnet
cycle-cli-jumpbox.json	Install the CycleCloud CLI on the jumpbox	cycle-prereqs-managed-identity, vnet
cycle-cli-local.json	Install the CycleCloud CLI locally	cycle-prereqs-managed-identity, vnet
cycle-install-server-managed-identity.json	Create a CycleCloud server in a managed identity context	cycle-prereqs-managed-identity, jumpbox, vnet
cycle-prereqs-managed-identity.json	Create all pre-requisites for deploying CycleCloud with managed identity
jumpbox.json	Create a jumpbox in the admin subnet	Existence of a VNET
jumpbox-anf.json	Create a jumpbox in the admin subnet with ANF mounted	anf
jumpbox-nfs.json	Create a jumpbox in the admin subnet acting as a 2TB NFS server	Existence of a VNET
vnet.json	Create a vnet named hpcvnet 10.2.0.0/20 with subnets admin, compute, netapp, viz and storage
workstation.json	Build a Windows Desktop workstation, domain joined	ad

How to use building blocks ?

It's easy as the blocks have been designed to be merged together into a single configuration file. The init-and-merge.sh will do the merge and initialization and you just need to provide an ordered list of the blocks you want to use and json file containing the variables you want to set. The init.sh script below will create a config file to deploy a VNET, a JUMPBOX with NFS and a BEEGFS cluster.

#/bin/bash
block_dir=$azhpc_dir/blocks
AZHPC_CONFIG=config.json
AZHPC_VARIABLES=variables.json

blocks="$block_dir/vnet.json $block_dir/jumpbox-nfs.json $block_dir/beegfs-cluster.json"

# Initialize config file
echo "{}" >$AZHPC_CONFIG
$azhpc_dir/init-and-merge.sh "$blocks" $AZHPC_CONFIG $AZHPC_VARIABLES

Before running that script I need to create a variables.json file which contains all the <NOT-SET> values of my blocks. In our case these are only resource_group and location.

{
  "variables": {
    "resource_group": "my resource group",
    "location": "my location"
  }
}

Once done I can run the init script wich produce a config.json that I can use to build my whole environment.

$ ./init.sh
$ azhpc-build

NOTE : Have a look at the examples/cc_beegfs directory for a detailed example