lft is looking for trouble

For a given NetID, lft will crawl the /home directory of the user and run multiple commands before generating a report that highlights potential problems with the user's shell configuration files, software installations and jobs.

How to use it?

Add this line to your ~/.bashrc file on each cluster (Adroit, Della, Stellar, Tiger, Tigressdata and Traverse):

alias lft='/home/jdh4/bin/lft'

Then look at the help menu:

$ lft -h

lft is looking for trouble. Example usage:
     $ lft aturing

Options:
     -v  Verbose output (show packages and more jobs)
     -s  Show sponsor
     -h  Show this help menu

Example

What exactly does it do?

lft does the following:

about the user

• runs ldapsearch to get the full name, department and position of the user (this takes 1-2 seconds which explains why lft seems to hang at the start).
• converts an email alias to the NetID of the user (e.g., halverson is converted to jdh4).
• displays the groups of the user.
• shows when the user was last active as determined by the most recent between the modification time of /home and the last command. Note that last treats tigercpu and tigergpu independently.
• optionally runs sponsor_report to show the user's sponsor (e.g., lft aturing -s).
• displays the user's login shell if not /bin/bash (e.g., /bin/zsh, /bin/tcsh).

home directories

• shows all the clusters where the user has a /home directory (this information is updated every 2 hours M-F from 7:30 am to 5:30 pm).
• shows which top-level directories are world-readable (/home, /tigress, /scratch/gpfs) and the number of items in each.

.bashrc and .bash_profile

• scans .bashrc and .bash_profile looking for the loading of environment modules and the setting of important environment variables (PYTHONPATH, R_LIBS, R_LIBS_USER) while reporting the state of each startup script as either skeleton or custom.
• looks for ~/.modulerc

Python

• lists Conda environments in ~/.conda/envs as well as the packages (if -v option) with color coding.
• looks for a .condarc file and follows the conda_envs path to Conda environments.
• lists Python packages found in ~/.local/lib/pythonX.Y/site-packages.
• reports on problematic libraries such as libmpich.so.
• reports on custom installations of Anaconda or Miniconda.
• shows when OnDemand Jupyter was last used.

R

• lists existence of R libraries and packages found in ~/R/x86_64-redhat-linux-gnu-library/X.Y as well as Conda environments in ~/.conda/envs (packages are printed with color coding to highlight parallel packages).
• reports the existence of ~/.R/Makevars and LOCK files as well as ~/.Renviron and ~/.Rprofile.
• shows when OnDemand RStudio was last used.

MATLAB

• shows which MATLAB environment modules the user loaded and when OnDemand MATLAB was last used.

Stata

• shows which Stata versions the user has used and when OnDemand Stata was last used.

Julia

• lists Julia packages found in ~/.julia/packages with color coding.

Jobs

• shows the user's fairshare value, 30-day group usage, group share of the cluster and number of running and queued jobs.
• shows most recent jobs according to sacct including an MT column which reports the memory efficiency (M) and the time efficiency (T) where both of the these independent quantities are scaled to be between 0 and 9 (time efficiency is the elapsed time of the job divided by the runtime limit).
• shows F, TO, OOM and NF job states in red.
• identifies OnDemand jobs and replaces the job name with either O-JUPYTER, O-RSTUDIO, O-MATLAB or O-STATA instead of, e.g., sys/dashboard/sys/jupyter.

Reminders

• shows the previous and upcoming downtime, RC workshops, reservations on Adroit and all university holidays.

About sponsors

The cluster-specific sponsor is returned. The code for this is here (however, the pandas lines are commented out). To install lft you will need to do:

$ cd panes
$ rm sponsor.py
$ wget https://raw.githubusercontent.com/PrincetonUniversity/monthly_sponsor_reports/main/sponsor.py

And then comment-out the two lines involving pandas:

#import pandas as pd
#rk = pd.read_csv(fname)