Burn Testing
Machines should be thoroughly tested before being listed on the RunPod platform.
Here is a simple guide to running a burn test for a few days.
Stop the RunPod agent by running
sudo systemctl stop runpodThen you can kick off a gpu-burn run by typing
docker run --rm --gpus all gshaibi/gpu-burn -d 36000
You should also verify that your memory, cpu, and disk are up to the task. You can use the stress-ng library to accomplish this: https://wiki.ubuntu.com/Kernel/Reference/stress-ng
When everything is verified okay, you can start the RunPod agent again by running
sudo systemctl start runpodLast modified 1mo ago