This simple script chewed up a lot of CPU, why?

Question

I have a remote machine I need to keep an eye on. It's running Ubuntu Studio 22.04 (KDE Plasma). A few weeks ago it crashed and journalctl showed a "Bug" that occurred a couple minutes before the crash. So I wrote a simple script that follows journalctl and if the word "Bug" appears it sends a warning email. I set that script running about 10 days ago. Yesterday, I connected to the machine remotely and checked htop and found that the script was using over 90% of CPU. I killed it and CPU usage dropped back to normal. Here's the script:

#!/bin/bash

#####################
#   THIS SCRIPT LAUNCHED AT STARTUP, CHECKS journalctl for string "Bug"
######################

while true; do

nohup journalctl --follow | grep -i -q "bug" && mutt -s "ALERT - AirchainPC may be in TROUBLE" -- [email protected] < bug_issued_by_journalctl.txt &>/dev/null &

done

Anything there that might explain the high CPU usage? BTW, I don't think I need that "nohup".

Answer 1

Please, change your script to something like:

#!/bin/bash
journalctl --follow | grep --line-buffered -i bug | while read -r ; do
  mutt -s "ALERT - AirchainPC may be in TROUBLE" -- '[email protected]' < bug_issued_by_journalctl.txt
done

Here, you are following the journal trying to catch lines containing the "bug" string. If the string is matched, it will be returned by grep immediately (thanks to the --line-buffered option) into the while loop. The read command will consume each such line and return, so that the body of the while statement will be executed at that time.

I don't know what the contents of the bug_issued_by_journalctl.txt file are, but I would prefer something dynamically created and containing (part of) the string matched from journalctl's output. In such a case, replace read -r with read -r msg and use the contents of the $msg variable in the mail text. Also, if the script is to be run by another user and/or from another directory, the full path for the above-mentioned file should be used.

Because of the --follow option in journalctl command, this script will never return. If you want, you can send the script to background like this:

#!/bin/bash

check_bug () {
  journalctl --follow | grep --line-buffered -i bug | while read -r ; do
    mutt -s "ALERT - AirchainPC may be in TROUBLE" -- '[email protected]'
  done
}

check_bug &>/dev/null &

A simple advice: Always test interactively first before putting something into background! While testing, you will also want to replace the &>/dev/null in the last line, with something like &>/tmp/check_bug.out.

---

Update: As @G.Sliepen suggested, you may replace the command sequence of

journalctl --follow | grep --line-buffered -i bug

with

journalctl --follow --grep=bug

You should read man journalctl page though, for case-sensitivity and other details.