Handling inappropriate processes on the login node

How to identify

Function to identify processes (Mike may have updates)

function cheaha-login-proclist() {
  ps \
    -eo user,pid,ppid,pcpu,pmem,stat,start_time,etime,cmd \
    --sort=-pcpu,-pmem \
    | egrep -v "  0.[0-9]  0.[0-9] " \
    | egrep -v "auditbeat|metricbeat|vscode|gnome-shell|desktop|firefox|slideshow|prometheus|systemd|glances" \
    | egrep -v " [0-9]  0.[0-9]" \
    | egrep -v "rsync|polkitd|sftp-server|notty|screensaver|mmfsd|atom"
}

Send email(s)

• Set reply-to email to be support@listserv.uab.edu:

  • Outlook:
    • Pop the email out into its own window.
    • Under "Options" ribbon, click "Direct Replies To".
    • Under "Delivery Options" add support@listserv.uab.edu to the "Have Replies Sent To" text box.
    • Click "Close". The "Direct Replies To" button in the ribbon should now stay highlighted.

• Suggested subject:

  Cheaha: Please run processes on compute nodes, not login node

• Suggested body:

  Hi there!

  As part of investigating a recent outage of service, we noticed that you have what appear to be research software running on the login node.

  Please run your research software, workflows, and other long-running or resource-intensive processes on compute nodes in Slurm job contexts.

  If you are new to Cheaha or to Research Computing, I highly recommend visting us at our Office Hours (Zoom info here: https://docs.rc.uab.edu/#contact-us) for a brief introduction. In either case, please review the documentation listed below for starting Slurm jobs.

  • How to use Open OnDemand: https://docs.rc.uab.edu/cheaha/open_ondemand/ood_main/
  • Using interactive applications on our web portal: https://docs.rc.uab.edu/cheaha/open_ondemand/ood_interactive/
  • Slurm batch jobs via the terminal: https://docs.rc.uab.edu/cheaha/slurm/submitting_jobs/

  For additional assistance and to reach us directly, please reply to this email or send an email to support@listserv.uab.edu.

  Thank you,

• Put list of recipient emails into BCC line. Allows sending multiple emails quickly without worrying about researchers seeing each other's mistakes (lets them save face).

No repository file or test is named. Start by locating the documentation page or operational guide for login-node process handling, then confirm where this procedure belongs. Done should cover identifying inappropriate processes, contacting affected users with the supplied message and links, and using BCC for multiple recipients.