Chrooting MySQL on Debian

I’ve posted a recipe for chrooting MySQL on Debian sarge a while ago. These instructions no longer work out of the box for newer MySQL packages from Debian and Ubuntu. The main problem is that the startup script added a few extra checks and script invocations that don’t understand the chroot environment. So here’s an updated plan:

    * Prepare the chroot directory. It’s recommended to use an extra partition/filesystem for it. I will use /srv/mysql (which is an LVM2 partition with an ext3 filesystem on my system) for the rest of the text.
    * Stop MySQL:

      /etc/init.d/mysql stop

    * Copy the databases to new location:

      mkdir -p /srv/mysql/var/lib
      cp -a /var/lib/mysql /srv/mysql/var/lib

    * Copy this script to /etc/default/mysql-chroot
    * Edit /etc/init.d/mysql:
          o Source the mysql-chroot script somewhere at the top:

            …
            test -x /usr/sbin/mysqld || exit 0

            . /etc/default/mysql-chroot

            SELF=$(cd $(dirname $0); pwd -P)/$(basename $0)
            …

          o Fix the disk space check:

            …
            # check for diskspace shortage
            datadir=`mysqld_get_param datadir`
            if LC_ALL=C BLOCKSIZE= df --portability $CHROOT_DIR$datadir/. | tail -n 1 | awk '{ exit ($4>4096) }'; then
              log_failure_msg "$0: ERROR: The partition with $datadir is too full!"
            …

          o Run setup_chroot right in the start section:

            …
            if mysqld_status check_alive nowarn; then
              echo "...already running."
            else
              setup_chroot
              /usr/bin/mysqld_safe > /dev/null 2>&1 &
            …

          o Somehow /var/run/mysqld/mysqld.pid disappears after each start. We have to create it each time, otherwise the stop command won’t work properly:

            …
            if mysqld_status check_alive warn; then
              log_end_msg 0
              ln -sf $CHROOT_DIR/var/run/mysqld/mysqld.pid
                             /var/run/mysqld
              # Now start mysqlcheck or whatever the admin wants.
              output=$(/etc/mysql/debian-start)
            …

    * In /etc/mysql/debian.cnf, change the two socket lines to:

      socket = /srv/mysql/var/run/mysqld/mysqld.sock

    * In /etc/mysql/my.cnf:
          o Change the socket line in the [client] section to:

            socket = /srv/mysql/var/run/mysqld/mysqld.sock

            Don’t change the socket lines in the other sections!
          o Add

            chroot = /srv/mysql

            to the [mysqld] section.
    * Prepend /srv/mysql to the log files listed in /etc/logrotate.d/mysql-server
    * Make /usr/bin/mysql_upgrade_shell use the chrooted socket. Note: Currently these changes must be made each time mysql gets upgraded because upgrades override this file!

      …
      −−password=*) password=`echo "$arg" | sed -e 's/^[^=]*=//'` ;;
      −−socket=*) socket=`echo "$arg" | sed -e 's/^[^=]*=//'` ;;
      −−ldata=*|−−data=*|−−datadir=*) DATADIR=`echo "$arg" | sed -e 's/^[^=]*=//'` ;;
      …
      fi
      $bindir/mysql_fix_privilege_tables −−silent −−user=$user −−password=$password −−socket=$socket $args
      exit 0
      …
      check_args="−−check-upgrade −−all-databases −−auto-repair −vuser=$user −−password=$password −−socket=$socket"
      …
      $bindir/mysql_fix_privilege_tables −−silent −−user=$user −−password=$password −−socket=$socket $args
      …

    * Start MySQL:

      /etc/init.d/mysql start

    * Check /var/log/syslog for errors ;-)
http://blog.blackdown.de/2006/12/30/chrooting-recent-mysql-versions-on-debian-and-ubuntu/