Graphistry on Azure: Manual Environment Setup Instructions#

DEPRECATION WARNING:

Get started more quickly and securely with Graphistry in Azure Marketplace.

We no longer recommend manually installing drivers via the original Graphistry-maintained bootstrap scripts. Instead, we now recommend using Graphistry in Azure Marketplace which has been preconfigured, and for advanced manual enterprise users, to use the Nvidia GPU Container (Ubuntu) “Nvidia NGC” base image.

Deprecated instructions#

Graphistry runs on Azure. This document describes initial Azure virtual machine environment setup. From here, proceed to the general Graphistry installation instructions linked below.

The document assumes light familiarity with how to provision a standard CPU virtual machine in Azure.

Contents:

Prerequisites: Azure GPU Quota
- Testing if you already have GPU Quota
- Requesting Azure for GPU Quota

Start a new GPU virtual machine
Proceed to general Graphistry installation

Subsequent reading: General installation

Prerequisites: Azure GPU Quota#

You may need to make quota requests to add GPUs to each of your intended locations:

Minimal GPU type: Standard_NC4as_T4_v3 in your region
Maximal GPU type: N-Series (NC T4_v3, NC v2, NC v3, and A100), see general documentation for sizing considerations

Testing if you already have GPU quota#

Go through the Start a new GPU virtual machine, then tear it down if successful

Requesting Azure for GPU Quota#

For each location in which you want to run Graphistry:

Start help ticket: ? (Help) -> Help + support -> New support request
Fill out ticket
Basics: Quota -> <Your Subscription> -> Compute (cores/vCPUs) -> Next
Problem: Specify location/SKU, e.g., West US 2 or East US for NCv2+ Series and ND+ Series
Contact Information: Fill out and submit

Expect 1-3 days based on your requested Severity rating and who Azure assigns to your ticket

1. Start a new GPU virtual machine#

See general installation instructions for currently supported Linux versions (subject to above Azure restrictions and general support restrictions.)

Virtual machines -> Create virtual machine
Ubuntu 22.04 LTS (or 24.04 LTS)
Basics: As desired; make sure can login, such as by SSH public key; needs to be a region with GPU quota
Size: GPU of all disk types, e.g., NC6v2 (hdd) is cheapest for development
Settings: Open ports for administration (SSH) and usage (HTTP, HTTPS)
Summary: Should say “Validation passed” at the top -> visually audit settings + hit Create

2. Confirm proper instance#

Test login; see SSH command at Overview -> Connect -> Login using VM Account
Check to make sure GPU is attached:

$ lspci -vnn | grep VGA -A 12
0000:00:08.0 VGA compatible controller [0300]: Microsoft Corporation Hyper-V virtual VGA [1414:5353] (prog-if 00 [VGA controller])
	Flags: bus master, fast devsel, latency 0, IRQ 11
	Memory at f8000000 (32-bit, non-prefetchable) [size=64M]
	[virtual] Expansion ROM at 000c0000 [disabled] [size=128K]
	Kernel driver in use: hyperv_fb
	Kernel modules: hyperv_fb

# Example output - GPU type varies by instance (T4, V100, A100, etc.)
xxxx:00:00.0 3D controller [0302]: NVIDIA Corporation [GPU Model]
	Subsystem: NVIDIA Corporation [GPU Model]
	Flags: bus master, fast devsel, latency 0, IRQ 24, NUMA node 0
	Memory at xxxxxxxx (32-bit, non-prefetchable) [size=16M]
	Memory at xxxxxxxxxx (64-bit, prefetchable) [size varies by GPU]

3. Proceed to general Graphistry installation#

For steps involving an IP address, see needed IP value at Azure console in Overview -> Public IP address

Ensure Docker includes the docker compose plugin (Docker 20.10+ includes it by default). NGC base images set the default runtime to nvidia (/etc/docker/daemon.json).

From here, you can perform a general installation using ./graphistry up -d (which is a wrapper for docker compose with GPU, telemetry, and cluster configuration).