Troubleshooting

Start with friendlai-node status and friendlai-node logs: most problems show up in one of them. See Monitor for what they mean.

My node shows as offline

A node is offline when it hasn't checked in for 2 minutes.

  • Is it running? If status shows the service isn't active, start it with friendlai-node start. Check that you haven't stopped it from the tray menu.
  • Linux, after logging out or rebooting: user services only run while you're logged in unless lingering is on: 
    sudo loginctl enable-linger $USER
  • Is the computer asleep? A sleeping computer can't answer questions. Turn off sleep while you want to earn.
  • Can it reach the internet? The logs show Error: … (retrying in …) lines when it can't reach friendlai.xyz.
  • "This node has been disabled by the network operator." The network has stopped sending work to this node, for example because its answers kept failing. Email us.

My node is online but isn't getting questions

  • Is it paused?status shows Paused: yes. Run friendlai-node resume.
  • Is anyone asking for your models? Questions go to nodes with the model that was asked for, oldest question first. Popular models on the status page get the most work. If many nodes serve the same model, each one gets less.
  • Are your models being served? Look for the Serving: line in the logs. If it says None of your models are in the network catalog yet, pull a model from the catalog.

A model I have isn't being served

  • It's not in the network catalog: the logs list it under Not in the network catalog (ignored).
  • It's too big: models larger than 85% of your RAM plus GPU memory are skipped, since they'd fail or be very slow.
  • It's an embedding model: those can't chat, so they're skipped.
  • You installed with --models and it isn't in the list. status shows Serving only.

The node notices new models within a minute of ollama pull; there's no need to restart it.

"Can't reach Ollama"

The node retries every 30 seconds, so it recovers once Ollama is running. Start the Ollama app, or on Linux: 

systemctl start ollama

If Ollama listens somewhere other than http://127.0.0.1:11434, tell the node with friendlai-node install --ollama http://host:port.

Questions fail on my node

Job …: failed: … lines carry Ollama's error. The most common cause is running out of memory, often because another program is using the GPU. Try a smaller model, close other GPU-heavy programs, or pause the node while you use them. Check that the model runs on its own: 

ollama run qwen3:8b "Hello"

The tray icon is missing

  • It starts with your desktop session, so log out and back in after installing on Linux, or start it yourself with friendlai-node tray &.
  • Stock GNOME has no system tray: install the AppIndicator extension.
  • The node runs without the icon; use your account page or status instead.

macOS says the binary "can't be opened"

The node isn't notarized by Apple yet. Run these in the folder you downloaded it to, then install again: 

xattr -c friendlai-node && codesign --force --sign - friendlai-node

My node isn't on my account page

It was installed without a node key, so it isn't linked to you. See linking a node to your account.

Still stuck?

Email us with the output of friendlai-node status and the last few dozen lines of friendlai-node logs.