bash script to restart Apache automatically

We used to have Apache segfaulting sometimes on a machine; here's the script we used trying to debug the problem while keeping Apache up. It ran from cron (as root) once every minute or so. It should be self-explanatory.

#!/bin/sh# Script that checks whether apache is still up, and if not:# - e-mail the last bit of log files# - kick some life back into it# -- Thomas, 20050606PATH=/bin:/usr/binTHEDIR=/tmp/apache-watchdogEMAIL=yourself@example.commkdir -p $THEDIRif ( wget --timeout=30 -q -P $THEDIR http://localhost/robots.txt )then    # we are up    touch ~/.apache-was-upelse    # down! but if it was down already, don't keep spamming    if [[ -f ~/.apache-was-up ]]    then        # write a nice e-mail        echo -n "apache crashed at " > $THEDIR/mail        date >> $THEDIR/mail        echo >> $THEDIR/mail        echo "Access log:" >> $THEDIR/mail        tail -n 30 /var/log/apache2_access/current >> $THEDIR/mail        echo >> $THEDIR/mail        echo "Error log:" >> $THEDIR/mail        tail -n 30 /var/log/apache2_error/current >> $THEDIR/mail        echo >> $THEDIR/mail        # kick apache        echo "Now kicking apache..." >> $THEDIR/mail        /etc/init.d/apache2 stop >> $THEDIR/mail 2>&1        killall -9 apache2 >> $THEDIR/mail 2>&1        /etc/init.d/apache2 start >> $THEDIR/mail 2>&1        # send the mail        echo >> $THEDIR/mail        echo "Good luck troubleshooting!" >> $THEDIR/mail        mail -s "apache-watchdog: apache crashed" $EMAIL < $THEDIR/mail        rm ~/.apache-was-up    fifirm -rf $THEDIR

We never did figure out the problem...

Can the count of a process really be less than zero?

This should be sufficient:

if ! pgrep apache2 -c >/dev/null; then

You could try to send an http request to apache (e.g. using wget --timeout=10) and if that request times out or fails (exit status != 0), you kill and restart apache.