Skip to main content

Getting Started with Rackrr

This guide walks through how to launch and access GPU-backed compute on the Rackrr platform.

Rackrr enables users to provision GPU-enabled virtual machines (VMs) for AI workloads without managing underlying infrastructure. You can choose compute based on workload requirements, configure resources, and launch instances in minutes.

Accessing the Platform​

After signing in, navigate to Available VMs from the main dashboard or left-hand navigation. This section displays available compute capacity across regions and operators. Dashboard

Selecting Compute Capacity​

You will see a list of available servers. Each server represents a physical machine with defined limits for:

  • CPUs
  • GPUs
  • Memory (RAM)
  • Local storage
  • Network bandwidth
  • Geographic location

Pricing and maximum allocatable resources are shown per server.

Your choice should be driven by:

  • workload type (training, inference, batch jobs)
  • performance requirements
  • budget and runtime expectations

Enterprise-grade GPUs are typically used for large-scale training and inference, while lighter workloads may be better suited to workstation-class GPUs.

Creating a Virtual Machine​

Once you select a server, you can begin configuring a virtual machine.

Configure VM Resources​

Under Specifications, define the following:

VM specification configuration

  • vCPUs Minimum of 2 (even numbers only). Maximum depends on server availability.
  • RAM Minimum of 4 GB (8 GB for Windows-based instances).
  • GPUs Minimum of 1. GPUs are the primary cost driver.
  • Storage Local disk space. Minimums vary by operating system.
  • Operating System Available options may include:
    • Ubuntu 22.04
    • Ubuntu 20.04
    • Ubuntu with preinstalled AI frameworks (e.g., PyTorch, TensorFlow, Jupyter)
    • Windows (where supported)

The operating system selection does not change pricing directly, but different OS images may have different minimum resource requirements.

Networking & Access​

For security and isolation reasons, direct server ports are not publicly exposed. Instead, access is managed via port forwarding.

Default Ports​

  • Linux (Ubuntu): SSH via port 22
  • Windows: Remote Desktop via port 3389

These are automatically configured.

Additional Ports​

You may optionally forward additional ports for services such as:

  • databases
  • APIs
  • web applications
  • notebooks

Each forwarded port maps an external port to an internal service port.

Pricing Overview​

The total cost of a VM is calculated based on:

  • GPU allocation
  • CPU allocation
  • RAM
  • Storage

GPU pricing varies by model and availability. CPU, RAM, and storage pricing are consistent across servers.

Pricing is visible before launch so you can review the estimated cost.

Launching the VM​

Click Start Virtual Machine to provision the instance.

After submission:

  • Credentials will be displayed (temporary password)
  • Initial provisioning may take 10–15 minutes

During this time, the VM may not be immediately reachable.

Accessing Your VM​

Linux (Ubuntu)​

Connect using SSH:

ssh -p <ssh_port> user@<vm_ip_address>

If prompted to confirm the host key, type yes.

You will then be asked for the temporary password provided at launch.

Windows​

Access Windows instances using Microsoft Remote Desktop.

When adding a new PC:

<vm_ip_address>:<external_port>

Use the password shown at launch.

Common Connectivity Issues​

Connection Refused / Network Unreachable​

These errors usually indicate the VM is still provisioning. Wait a few minutes and retry.

Host Key Changed​

If you receive a host key warning, it typically means a new VM has been created on an IP address previously used.

You can:

  • Remove the old entry from your known_hosts file
  • Or bypass host key checking for that connection

Running Jupyter Notebooks​

If your VM includes Jupyter:

  1. SSH into the VM
  2. Start Jupyter:
jupyter notebook --no-browser --ip=0.0.0.0 --port=8888
  1. Copy the generated URL
  2. Replace 127.0.0.1 with the VM's IP
  3. Replace the port with the forwarded external port
  4. Open the URL in your browser

Stopping a VM​

You can stop a VM to retain data and pause compute billing.

When stopped:

  • GPUs, CPUs, and RAM are released
  • Storage remains allocated
  • You are billed for storage only

If capacity is unavailable when restarting, you may need to wait until resources free up.

Terminating a VM​

To permanently delete a VM:

  • Navigate to the VM
  • Select Terminate
  • Confirm the action

This permanently deletes all data and stops all billing. This action is irreversible.