Looping Shell Script Recipe

The following procedure creates and executes a looping script that executes a command with output to a file and a sleep in between each execution:

1. If you're using SELinux security, then you'll have to place the script into an allowed location. First, check SELinux status:

   sestatus | grep "Current mode"

   Example output:

   $ sestatus | grep "Current mode"
   Current mode:                   enforcing

   1. If the sestatus command does not exist or Current mode is set to anything other than enforcing, then you can place the new script anywhere; for example:

      cd /tmp/

   2. If SELinux is set to enforcing, then place the new script into an allowed directory (may require root permission); for example, a common such directory:

      cd /usr/local/bin/

2. Create a file named diagscript.sh replacing the command after the CHANGEME comment, and, if necessary, the sleep time and the output location instead of /tmp/:

   #!/bin/sh
   set -e
   # Change output directory if needed
   outputfile="/tmp/diag_$(hostname)_$(date +"%Y%m%d_%H%M%S").txt"
   while true; do
     echo "[$(date)] Executing iteration" >> "${outputfile}" 2>&1
   
     # CHANGEME: Change the `ps -elfyww` command to your desired command (but make sure to still write to ${outputfile}):
     ps -elfyww >> "${outputfile}" 2>&1
   
     # Change the sleep time (in seconds) as needed:
     sleep 30
   done

3. Manually confirm the looped command exists and works (as some distributions support different flags). Using the example command above:

   ps -elfyww

4. Make the script executable:

   chmod +x diagscript.sh

5. If SELinux is set to enforcing, change the security context of the script based on a well-known executable; for example:

   chcon --reference=/usr/bin/cat diagscript.sh

6. Check if you're running in a systemd environment:

   systemctl status

   1. If systemctl status succeeds (running in a systemd environment), start the script with systemd-run so that it's not interrupted if the user that started the script logs out:

      systemd-run ./diagscript.sh

      Example output:

      Running as unit: run-r973b3f9d949b4d60a6b9a5fc842b84d1.service; invocation ID: 800d52ff3cb541659df667804ae7b149

      Confirm the script is running by running systemctl status on the dynamically generated unit name and confirm it's active (running); for example:

      # systemctl status run-r973b3f9d949b4d60a6b9a5fc842b84d1.service
         [...]  
         Active: active (running) since Tue 2025-01-14 08:41:28 CST; 54s ago
         Main PID: 2686 (diagscript.sh)

   2. If systemctl does not exist (not running in a systemd environment), start the script with nohup so that it's not interrupted if the user that started the script logs out and send it to the background with &:

      nohup ./diagscript.sh &

7. If you'd like to watch the script output, you may tail it (replacing /tmp/ if you used a different directory):

   tail -f /tmp/diag*txt

8. Monitor that the script output does not exhaust disk
9. When you are ready to stop the script:
   1. If the script was started with systemd-run, stop it using the dynamically generated service name printed when starting it (or search in systemctl status output); for example:

      systemctl stop run-r973b3f9d949b4d60a6b9a5fc842b84d1.service

   2. If the script was started with nohup, kill it by finding the process ID:

      kill ${PID}

      Alternatively, if available, use pkill with the script name:

      pkill -f diagscript.sh

10. Upload diag*txt and any other relevant logs

Previous Section (Process Crash Recipe) | Next Section (Looping Batch Script Recipe) | Back to Table of Contents