Graphistry in Azure Marketplace#

You can now securely run Graphistry in your private Azure account with one click by launching Graphistry in Azure Marketplace!

Get it now!#

Click create (vs. Start with a pre-set configuration)
Fill most VM settings as usual: Resource group, VM Name, and User + SSH Key / Password
- Region: Pick a GPU-capable region in which you have GPU quota (see below), such as East US and West US 2
- Size: Select NC4as_T4_v3, NC6s_v2, NC6s_v3 (and not v1 series of NC6s) or bigger within the same families
  - If the first time using GPUs, you may need to make a quota request, which generally takes 1-2 days
- If asked, publicly expose (for incoming connections) http (80), https (443), ssh (22), and potentially unconfigured streamlit (8501)
Click Review and create and then Create
Go to the public IP, refresh the page until it has fully loaded, create the first user (= admin) account, and start graphing!

Please reach out to our team if we can help along the way.

Basic administration#

Security: Enable auto-TLS, restrict network access, and manage your server internals
Create users
Generate API keys for individuals without accounts
Turn server on-and-off in the Azure Console via stop and start
Advanced configuration
Update, backup, and migrate
- To simplify administration and limit downtime, we recommend creating a new Marketplace instance, copying data snapshots to it and loading it in, and switching DNS to the new instance only when tested

Common marketpace administration#

The Graphistry marketplace instance is designed for secure web-based use and administration. However, command-line administration can be helpful. This document shares common marketplace tasks. See the main docs for general CLI use.

Contents:

Azure special notes
Recommended configuration
Solve GPU availability errors
Command-line Login
Docker
Install Python packages
Install native packages
Marketplace FAQ

1. Azure special notes#

Pick an NC4as_T4_v3, NCv2-series, or NCv3-series GPU, such as an NC6s v2, from an Azure GPU-capable region near you
Install path: /var/graphistry
If many uploads are expected, you may benefit from attaching a managed disk; contact Graphistry staff for automation scripts

2. Recommended configuration#

Public ports:
- HTTP (80), HTTPS (443), SSH (22), and optionally StreamLit (8501)
  - You can lock these down further at any time, e.g., only open 22 when using
- In restricted environments, constrain networking to a a safelist, e.g., VPN, and optional, change logging drivers to stop Graphistry from recieving maintenace logs
Assign a static IP or DNS entry to your Graphistry instance
Setup TLS and restrict ports

3. Solve GPU availability errors#

Upon trying to pick a VM size or launching, Azure may fail for several reasons:

Insufficient account quota. Solve by requesting increased GPU / vCPU core capacity in your target launch region and extra regions for dev and contention periods. From your Azure Portal, go to ? (Help) -> Help + support -> New support request -> Service and subscription (Quotas) (link). Request quota for multiple NC4as_T4_v3, NC6s_v2, NC6_v3s or bigger in the same family for your intended regions that have them. We recommend at least 3x-4x quota: production, staging, test1, test2. Later, you may want more for scaling experiments.
Lack of GPU availability in the current region. In this case, try another valid GPU type in the current region, or launch in another region. Keeping the GPU close to your users is a good idea to minimize latency.

4. Command-line Login#

Logging in is configured at Azure instance start and uses your instance’s public IP/domain, custom username, and choice of password or SSH key:

ssh -i my_key.pem my_username@MY_PUBLIC_IP_HERE

Many ssh clients may require you to first run chmod 400 my_key.pem or chmod 644 my_key.pem before running the above.

5. Docker#

Graphistry leverages docker compose and the Azure image preconfigures the nvidia runtime for docker. Note that Graphistry in Azure requires sudo.

cd /var/graphistry
docker compose ps

               Name                             Command                  State                        Ports                  
-----------------------------------------------------------------------------------------------------------------------------
graphistry_celerybeat_1              /entrypoint bash /start-ce ...   Up             8080/tcp                                
graphistry_celeryworker_1            /entrypoint bash /start-ce ...   Up             8080/tcp                                
graphistry_forge-etl_1               /tini -- /entrypoints/fast ...   Up (healthy)   8080/tcp                                
graphistry_nexus_1                   /entrypoint /bin/sh -c bas ...   Up             8080/tcp                                
graphistry_nginx_1                   nginx -g daemon off;             Up             0.0.0.0:443->443/tcp, 0.0.0.0:80->80/tcp
graphistry_notebook_1                /bin/sh -c graphistry_api_ ...   Up             8080/tcp                                
graphistry_postgres_1                docker-entrypoint.sh postgres    Up             5432/tcp, 8080/tcp                      
graphistry_redis_1                   docker-entrypoint.sh redis ...   Up             6379/tcp, 8080/tcp                      
graphistry_streamgl-datasets_1       /tini -- /entrypoints/fast ...   Up (healthy)   8080/tcp                                
graphistry_streamgl-gpu_1            /tini -- /entrypoints/fast ...   Up (healthy)   8080/tcp                                
graphistry_streamgl-sessions_1       /tini -- /entrypoints/fast ...   Up (healthy)   8080/tcp                                
graphistry_streamgl-svg-snapshot_1   /tini -- /entrypoints/fast ...   Up (healthy)   8080/tcp                                
graphistry_streamgl-vgraph-etl_1     /tini -- /entrypoints/fast ...   Up (healthy)   8080/tcp                                
graphistry_streamgl-viz_1            /tini -- /entrypoints/stre ...   Up             8080/tcp   

Note: Precise set of containers changes across versions

6. Install Python packages#

If you see wheel errors, you may need to run pip install wheel and restart your Jupyter kernel.

7. Install native packages#

By default, Jupyter users do not have sudo, restricting them to user-level installation like pip. For system-level actions, such as for installing golang and other tools, you can create interactive root user sessions by logging into the Jupyter Docker container:

Admin:

Note that sudo is unnecessary:

ubuntu@ip-172-31-0-38:~/graphistry$ docker exec -it -u root graphistry_notebook_1 bash
root@d4afa8b7ced5:/home/graphistry# apt update 
root@d4afa8b7ced5:/home/graphistry# apt install golang

User:

ubuntu@ip-172-31-0-38:~/graphistry$ docker exec -it  graphistry_notebook_1 bash
graphistry@d4afa8b7ced5:~$ go version

go version go1.10.4 linux/amd64

8. GPUDirect Storage (GDS)#

Issue Overview A specific issue has been identified with NVIDIA GPUDirect Storage (GDS) on Azure Ubuntu 22.04 images using the official NVIDIA CUDA Drivers (version 550). While the same operating system and driver versions work correctly on other cloud providers such as AWS AMI and Google Kubernetes Engine (GKE), the Azure platform presents a unique challenge.

Temporary Workaround To ensure stability and performance of Graphistry on Azure, GDS support has been disabled by default. This adjustment has been applied to all Azure Marketplace images by setting the environment variable LIBCUDF_CUFILE_POLICY to OFF. See the official Magnum IO GPUDirect Storage (GDS) documentation here for more information: https://docs.rapids.ai/api/cudf/nightly/user_guide/io/io/#magnum-io-gpudirect-storage-integration

Future Considerations Monitoring of this issue is ongoing to identify a permanent solution. Once Azure resolves this issue, GDS support will be re-enabled to take full advantage of its performance benefits. Customers will be informed of any updates or changes regarding this matter.

Recommendations Users may override the default setting and enable GDS support manually by setting the environment variable in the data/config/custom.env file (a Docker Compose environment file). For example: LIBCUDF_CUFILE_POLICY=ALWAYS.

Understanding and cooperation are appreciated as work towards a resolution continues. For further assistance, please reach out to the support team.

9. Marketplace FAQ#

No site loads or there is an Nginx 404 error#

Wait a few minutes for the system to finish starting. If the problem persists for more than 5-10min, log in, run docker ps, and for each failing service, restart it. If problems persist further, please report the results of docker logs <service> to the Graphistry support team and we will help out.

I lost my admin account#

See the reset command in the main README. Requires logging in, and will delete all users, but no data.

I want to log into the server#

See section login

I want to use a private VM#

Private marketplace: See section Azure Marketplace Privacy Offer: Launching Graphistry Core BYOL

Private docker: Check Azure instructins for VM sizing and Docker instructions for manual installation.

See general installation for further information.