For how notebooks work in general, see Jupyter notebooks. For
the disk limits, see Notebook files and disk space.
- Your
work folder may be full. Each time the server starts it copies the course
folders into work, and the start fails when there's no room. Open a
support ticket asking for more space; once the server starts, delete
what you don't need. See When work is full.
- A GPU node takes a few minutes to start. A CPU node starts right away.
- Maintenance: upcoming outages are listed at the top of the Server Options form.
- "Academic E-mail Required" or "Verify Your E-mail Address": see
Accounts.
- You got an email saying "Your Jupyter Notebook was stopped because it was over disk
quota." Your server wrote more than its limit outside work (12.5 GB on a CPU node,
10 GB on a GPU node). Psi4 scratch files in /tmp and installed packages are the usual
causes. See
What fills the disk without you noticing.
- You were on a GPU node: notebook servers on a GPU node are stopped, with no email,
after 2 hours without GPU activity. Code running only on the CPU doesn't count, so
pick a CPU node for work that doesn't use the GPU.
- No email: servers shut down after 12 hours idle. A notebook that is still
running code counts as active.
- It ran for 4 days: every server is stopped after 96 hours, however busy it is,
and no email is sent. Save checkpoints in work for anything that long, or run it as a
ChemCompute job.
- You're asked to log in again: a hub sign-in lasts 12 hours.
- They were saved outside
work. Only ~/work (/home/jovyan/work) is kept;
everything else is erased when the server stops, and it can't be recovered.
- You changed a file in
experiments or datasets. Those folders go back to the
originals every time the server starts. Keep your own copies in another folder in
work.
Check where the file was going:
- Into
work: work is full. Check its size in a Terminal with du -sh ~/work,
delete what you don't need, or ask for more space. See
Getting more space.
- Anywhere else: move the file, or the program's scratch folder, into
work.
The notebooks in experiments are read-only, so saving one fails or the notebook says
it's read-only. Copy it into a folder of your own in work and open the copy.
Most often the notebook used more than its 6 GB of memory. Reduce the size of the
calculation (a smaller basis set, fewer atoms, fewer trajectory frames loaded at once), or
run it as a ChemCompute job, which can use more memory.
Open a support ticket. Say whether you're on a CPU or GPU node, which
image version you picked, and the exact message you see.