Fix a problem (providers)

First, look at the agent

Run these on the GPU computer:

systemctl status gpuflow-agent      # should say "active (running)"
journalctl -u gpuflow-agent -n 50   # the last 50 lines of its log

Most answers are in those 50 lines.

My machine shows Offline

1. Is the computer on and connected to the internet? The agent needs to reach gpuflow.app and ws.gpuflow.app on port 443 (normal HTTPS).

2. Is the agent running? If systemctl status gpuflow-agent doesn’t say active (running), start it:

   sudo systemctl start gpuflow-agent

3. Still offline? Open My Machines, copy the command on that machine’s card and run it again. It reinstalls the agent and keeps the same machine, so its listing comes back by itself.

A machine that was offline shows as online within about a minute of the agent connecting.

Install errors

You see	What to do
enrollment failed and the log says the token expired	The command is valid for 15 minutes. Reload the dashboard, copy a fresh command and run it.
this installer targets systemd Linux; systemctl not found	The installer needs Linux with systemd. Windows, macOS and systems without systemd aren’t supported yet.
unsupported CPU architecture	Only 64-bit PCs (x86_64) and 64-bit ARM (arm64) are supported.
no enrollment token	Copy the whole command from the dashboard, including --token and everything after it.
failed to pull model '…'	The model name is wrong, or it doesn’t fit your GPU. Try a smaller one, for example --model qwen2.5:0.5b.

The model is slow or won’t load

• A model that’s too big for your GPU’s memory runs slowly or not at all. Pick a smaller one (see choosing an AI model).
• The first request after a restart is slower while the model loads into memory.
• Check the GPU temperature on My Machines. A hot GPU slows itself down.

A rental ended early

If your machine is offline for 10 minutes during a rental, the rental ends. You’re paid for the time up to the moment it went offline, and the renter gets the rest back. Keep the machine awake: turn off sleep and hibernate in your system settings.

Uninstall

sudo systemctl disable --now gpuflow-agent
sudo rm -f /usr/local/bin/gpuflow-agent /etc/systemd/system/gpuflow-agent.service
sudo rm -rf /etc/gpuflow /var/lib/gpuflow /var/lib/private/gpuflow
sudo systemctl daemon-reload

Ollama stays installed. Remove it separately if you don’t want it. Then remove the machine on My Machines with the trash can button.

Note

Still stuck? Open a ticket at gpuflow.app/support and paste the last 50 lines of the log (journalctl -u gpuflow-agent -n 50).