Troubleshooting

Most problems are fixed by stopping your server and starting it again. If that doesn't help, find your symptom below.

To restart your server: go to espm.nrp-nautilus.io/hub/home, click Stop My Server, wait for it to stop, then click Start My Server. Files in your home directory are kept.

The hub won't load at all

If espm.nrp-nautilus.io shows an error page (for example "503") or doesn't respond, the problem is on our end, not yours. Tell your instructor, including the time and what you saw.

My server is stuck starting

The first start after an update can take several minutes. Give it up to 10.
GPU environments and the largest memory options wait for a machine with room. Try a smaller size, or a CPU environment, if it keeps waiting.
If it fails or never finishes, go to /hub/home, stop the server, and start it again.

My R session or Python kernel keeps restarting

This usually means you ran out of memory. Save your work, restart the server, and choose a larger memory option. Or work with less data at a time: read only the columns and rows you need, or query files with DuckDB instead of loading them whole.

A package I installed is gone

Installed packages don't survive a restart; only your home directory does. Keep a list of what you need in your project (a requirements.txt or an install.R script) and re-run it after starting a server.

My server shut down while I was away

Servers idle for about an hour are stopped automatically. Your files are safe; anything that was only in memory is gone. For long terminal jobs, run them inside tmux so they survive your browser disconnecting. That won't stop an idle shutdown, though.

`git push` fails

Run gh-scoped-creds again. Its credentials last 8 hours. Also check that the GitHub app has access to that repository. See Using GitHub.

An AI assistant says my key is missing or invalid

For opencode with NRP, check that you saved the key under the provider id nrp: opencode auth list. See AI gotchas.
Make sure you copied the whole key, with no extra spaces.
NRP "overloaded" or rate-limit errors mean the model is busy: wait, or pick another model.

Still stuck?

Ask your instructor or TA. Include what you were doing, what you expected, the exact error message (a screenshot helps), and which environment and interface you chose.