I’m out of SU! What should I do?
To get more allocation, you can email to Mitchcomp Help and ask to be given hours from Prof. Toback’s allocation.

I’m out of disk space and/or file count! What should I do?
If you need a higher memory or file-count limit, you can send a request to HPRC help. You should also cc Dr. Toback (toback@tamu.edu) and/or Mike Kelsey (kelsey@tamu.edu) in your request because their approval as the MitchComp/HPRC liaisons is required.

I used sbatch a few times and all SU is consumed! scancel did not refund my SUs!
Your allocation really does only get charged for the SUs that were actually used up in running. When you submit, a deduction is taken for the total usage you estimated (nodes*wall-hours). Once the job completes, the actual usage for that job is recorded, and the original deduction goes away.
However, the accounting system is an entirely separate server from SLURM, so there are lags between changes or results at either end being communicated to the other.
SLURM remembers the original deduction amount, so if you submit a bunch of jobs at once, you can appear to burn through your allocation quickly. The idea, of course, is so that users don’t accidentally game the system, and get to run more jobs than they’re allocated for.
If you have a new job configuration that you testing, submit them with a very short time limit (5 minutes or so). If it starts running and doesn’t crash or fail submission, then kill it, and resubmit with the full time estimate.
HPRC also encourages us to try to set the wall-clock time limit “accurately.” Don’t pad out your –time argument unnecessarily, but do give yourself some overhead (10% or so).

I cannot git clone or do HTTPS requests in Jupyter Notebook and/or my batch jobs!
Working nodes does NOT have internet access.
However, all the scdms-jupyter modules after V3.3 will allow you to do HTTPS requests like GET and POST requests.
You also can do HTTPS requests with some proxy settings. If you need that, send a request to Mitchcomp Help and we will help you!

Jupyter Notebook jobs sits in queue for a long time / Jobs time out
If it is not an HPRC side problem, it might be your Home area being full (either disk space or file count). Your Jupyter Notebook session needs to write on the .local directory in your home area, and if your home area be full, it might fail.
The best way to avoid this problem is to create a directory in your $SCRATCH area and make your ~/.local a symlink.

>$ mkdir -p ~/.local
>$ mv ~/.local $SCRATCH/jupyter-local
$ ln -s $SCRATCH/jupyter-local ~/.local

This way you don’t have to worry about your home directory being full!

How to make Widgets available in my Jupyter Notebook
This solution assumes you are loading a sdcms-jupyter module (either by direct module loading or by choosing it on “Additional Module to Load” section when you want to launch a session). Guide written by “Joshua D Winchell” written at Jupyter Notebook Modifications page at confluence.
Our scdms environment on HPRC was ipywidgets installed, but they have to be activated for your jupyter sessions. Do the following:

1. Launch a jupyter notebook session from the Terra OnDemand portal
2. New -> Terminal (that is, in the notebook browser, not the OnDemand portal)
3. jupyter nbextension install –py widgetsnbextension –user
4. jupyter nbextension enable –py widgetsnbextension –user
5. Open a new notebook/reload kernel

Widgets should now be available in your notebook and both in your current session and in future sessions launched from OnDemand (the ‘enable’ line does this).

Where can I seek help?
If you had questions and you are a member of Mitchcomp group, try emailing mitchcomp_help@physics.tamu.edu first, and if your issue was not resolved, then try to email help@hprc.tamu.edu