Marseglia.org

I had a problem with mysql and syslog-ng. The mysql pipe was not being re-read by syslog-ng if it was interrupted at all. If the file was deleted or the pipe command broke, whatever...syslog-ng promptly stopped reading the file.

I had created a script run by cron to combat this but it wasn't working very well. I had also told the script to run @reboot in cron, but that wasn't working either. Most of the scripts I found on the net just weren't working for me.

So I rolled my own. Here's an init.d script to make sure the pipe was created and the logging started every time the machine booted. I even added a little 'status' indicator and stuck that in cron so it emails if there's a problem.

syslog-ng.mysql init.d script
-----

#!/bin/sh
# script to start syslogng dump into mysql
# options
MYSQLPIPE=/tmp/mysql.pipe
USER=root
DB=syslog

MYSQL=/usr/bin/mysql
PIDFILE=/var/run/syslogng.mysql

case "$1" in

        start)
                if [ ! -e $MYSQLPIPE ]; then
                        mkfifo $MYSQLPIPE || (
                                echo "Can't create $MYSQLPIPE.">&2
                                exit 1
                                )
                fi
                echo -n "Starting syslog-ng mysql dump..."
                exec $MYSQL -u $USER $DB < $MYSQLPIPE &
                echo "done."
                echo $! > $PIDFILE
                /etc/init.d/syslog-ng restart
                logger -is -- Syslogng-mysql started. Process id:$!
                ;;
        stop)
                if [ -f $PIDFILE ]; then
                        logger -is -- Stopping syslogng-mysql.
                        sleep 5
                        echo -n "Removing mysql pipe...">&2
                        rm -f $MYSQLPIPE
                        echo -n "stopping input to database...">&2
                        kill `cat $PIDFILE` 2>&1 && (
                                rm -f $PIDFILE
                                echo "done.">&2
                                ) || (
                                echo "$PIDFILE exists but process not running. Removing $PIDFILE."
                                rm -f $PIDFILE
                                )
                else
                        $0 status
                fi
                exit 0
                ;;
        status)
                ps -f -p `cat $PIDFILE 2>&1` >/dev/null 2>&1  && (
                        echo "syslog-ng mysql up."
                        ) || (
                        echo "syslog-ng mysql down."
                        )

                if [ -f $PIDFILE ]; then
                        ps -f -p `cat $PIDFILE`|xargs -l1 logger -is -- Syslogng-mysql check:
                else
                        echo "Not running.">&2
                        exit 1
                fi
                exit 0
                ;;
        restart)
                $0 stop && (
                sleep 5
                $0 start
                ) || exit 1
                ;;
        *)
                echo "Usage: (start|stop|status)">&2
                exit 3
                ;;
esac
exit 0

crontab
-----

10 * * * *      root    /etc/init.d/syslog-ng.mysql status || cat EOF|mail -n you@yourhostname.org -s "ALERT: syslogng on $HOSTNAME is down."