Ray Cluster Provisioning Module for WHMCS
Sell ready-to-use Ray clusters from WHMCS: head and worker servers are created on Cudo Compute or Hetzner Cloud and Ray is installed for you.
What does the Ray cluster module do?
Ray is an open-source framework for spreading Python and machine learning work across many machines. Building a cluster by hand means creating servers, installing Ray on each one and pointing every worker at the head node. This module does all of that from a WHMCS order.
The customer chooses the number of servers, a data center (or any data center) and an image. On Cudo Compute they also pick a GPU model, GPU count, vCPUs, memory and boot disk; on Hetzner Cloud they pick a server type. The first server becomes the Ray head node and the others become workers. Once the servers are up, a cron job connects to each one over SSH, checks the operating system, installs Python and Ray with its data, train, tune and serve libraries in a virtual environment, and registers Ray as a system service so it comes back after a reboot. The head node starts with the Ray dashboard, and each worker joins the head on port 6379.
Which clouds does it run on?
Two backends ship with the module and you choose one per WHMCS product: Cudo Compute for GPU and CPU virtual machines, and Hetzner Cloud for CPU servers. Each product stores its own API key, and one click on the product page creates the matching configurable options.
What is included?
| Area | What the module handles |
|---|---|
| Order | Number of servers, data center, server type or GPU model, image |
| Setup | Servers created, Ray installed over SSH, workers joined to the head node |
| Customer controls | Restart servers, restart Ray, VNC console, SSH key management, cancel |
| Admin controls | Start, stop, restart and rebuild servers, restart Ray |
| Admin view | Status of each node, Ray cluster stats and a link to the Ray dashboard |
| Billing | Hourly deduction from prepaid credit with markup and a grace period |
How is it billed?
Usage is prepaid. Customers buy credit through a credit product addon that ships with the module, and an order is accepted only with that addon in the cart or with a minimum credit balance you set. When the servers are created, the module records each one's hourly cost (the price Cudo reports, or the Hetzner server type price shown at checkout), adds your fixed or percentage markup and deducts an hour at a time from the customer's credit through a cron job that runs every five minutes. With no credit left the service is suspended, and after your grace hours it is terminated.
Who is it for?
GPU and cloud resellers who want to sell a ready AI compute product rather than bare servers, and teams that run Ray workloads for customers and want each cluster billed by the hour without manual setup.
What do you need, and what are the limits?
Requirements: WHMCS 8.x (confirmed at purchase), PHP 7.4 or newer, a Cudo Compute or Hetzner Cloud account with an API key, SSH access from your WHMCS server to the new servers, and cron jobs for the install and billing scripts.
Limits: Ray is installed automatically only on Ubuntu and Debian images. The Ray dashboard listens on the head node's public IP on port 8265, so add firewall rules if you do not want it open. The module sets up the cluster; it does not write, schedule or monitor the customer's Ray jobs. You pay the cloud provider for every server.
It costs $199 once for a lifetime license with twelve months of updates and support, an instant download and license key, and a seven-day refund window.
Key features
- Multi-server Ray clusters ordered from one WHMCS product
- Cudo Compute GPU and CPU backend and Hetzner Cloud backend
- Automatic Ray install over SSH on Ubuntu and Debian
- Workers joined to the head node, Ray run as a system service
- Ray dashboard link and cluster stats in the admin area
- Restart servers or restart Ray from the client area
- VNC console and SSH key management for customers
- Hourly billing from prepaid credit with fixed or percentage markup
- Credit top-up addon and minimum balance check at checkout
- Suspension when credit runs out, termination after a grace period
Why customers buy this module
- →Sell AI compute as a ready cluster instead of a pile of servers
- →No manual Ray installs or worker joins after each order
- →Hourly prepaid billing keeps short GPU jobs profitable
- →Customers restart nodes or Ray themselves without a ticket
Frequently asked questions
Does the module install Ray on the servers itself?
<p>Yes. After the servers are created, a cron job connects to each one over SSH, installs Python and Ray in a virtual environment and sets Ray up as a system service. The first server runs as the head node with the Ray dashboard, and the others join it as workers. This works on Ubuntu and Debian images.</p>
Can customers get GPUs in their Ray cluster?
<p>Yes, on the Cudo Compute backend. The order form lists Cudo's GPU models per data center along with the GPU count, vCPUs, memory and disk. The Hetzner Cloud backend offers CPU server types only. You choose the backend per WHMCS product, so you can sell both.</p>
How are customers charged for a running cluster?
<p>From prepaid credit, by the hour. Each server's hourly cost plus your markup is deducted every hour through a cron job. Customers add credit with the credit product addon that comes with the module. If the credit runs out the service is suspended, and it is terminated after the grace hours you set.</p>
Technical specifications
- WHMCS
- 8.x (confirmed at purchase)
- PHP
- 7.4+
- Cloud backends
- Cudo Compute, Hetzner Cloud
- Cluster OS
- Ubuntu or Debian images
- Billing model
- Hourly, from prepaid account credit
- License
- Lifetime, 1 year of updates and support included
- Delivery
- Instant download and license key
- Refund window
- 7 days