Launching an instance with a GPU via Horizon (GUI)

Overview

This guide details how to instantiate a single compute instance with a GPU accelerator card using the firstcloud Horizon dashboard.

Prerequisites

• Valid firstcloud credentials (username, passphrase, and domain).
• Familiarity with terminal/SSH and SSH key management.

Launching an Instance with a GPU

Steps

1. Import SSH Keys:

   • Upload your SSH key(s) to the Horizon dashboard. Refer to the importing SSH-keys guide.

2. Create Security Group for SSH Access:

   • Create an SSH access security group. See the create security groups

3. Log In to firstcloud Dashboard:

   • Access the firstcloud Dashboard using your provided credentials.

4. Navigate to Instances:

   • Go to "Project" > "Instances."

5. Launch Instance:

   • Click "Launch Instance."
   • Enter an "Instance Name."
   • Click "Next."

6. Configure Boot Source:

   • Select a boot source and enable "Create New Volume."
   • Set the volume size.
   • Choose a source image.
   • Click "Next."

7. Select Flavor:

   • Choose a GPU flavor. Current GPU Flavors are: gpu1.xlarge, gpu1.xxlarge and gpu1.3xlarge.
   • Click "Next."

8. Configure Security Groups:

   • Select the "default" and "ssh" security groups.
   • Click "Next."

9. Select Key Pair:

   • Choose your SSH key pair.
   • Click "Launch Instance."

10. Monitor Instance Provisioning:

    • Observe the instance status in the instance list.

11. Access Console (Optional):

    • Click on the instance and go to the "Console" tab to view console output.

12. Connect via SSH:

    • After associating a floating IP, open a terminal and connect using:

      ssh debian@<public IP> -i ~/.ssh/<private SSH key>
      

    • (Replace <public IP> and <private SSH key> with your values.)

13. Install Nvidia Drivers:

    • Installing the Nvidia Drivers:

      apt update
      apt install ubuntu-drivers-common
      apt install nvidia-driver-<version> 
      reboot
      
      nvidia-smi
      

    • Replace <version> with your desired Nvidia Driver version. To get a list of available versions use apt search nvidia-driver
    • After the reboot, the GPU should be visible by running nvidia-smi
    • You can also install the open drivers instead of the proprietary drivers with apt install nvidia-driver-<version>-open

Conclusion

We have successfully created an Ubuntu Instance with a GPU accelerator Card.