Creating a template cache : Slackware or HostGIS Linux

This process uses VMware to install the OS into a VM, then to trim down the VM's contents to only those items suitable for a VE environment, then to save a snapshot of the system as a host template cache for use in OpenVZ.

This document focuses on HostGIS Linux (a Slackware derivative) but aside from the specifics about installation settings, it should be 99% applicable to Slackware as well.

Create the VM in VMware

Technically, you could probably do this on a hardware PC without VMware, but VMware does make it more convenient.

Start by creating a new VM in VMware.

• The disk and RAM stats can be minimal, as the system will never see live use.
• There is no need to create the entire disk at once during the setup.
• Create the disk as SCSI.

Then install HGL.

• Create a small partition at the end of the disk for swap. Some swap is technically necessary, but since you'll never in fact be using it, a few MB should be fine.
• Set the passwords to 'password'
• Do set the timezone properly. The internal clock does not use UTC/GMT.
• Select the default mouse, but do NOT enable GPM at startup.
• Hostname: template
• Domain: internal.lan
• IP config: as appropriate for your LAN
• Nameserver: no

Reboot into your new HGL install, and log in.

Delete unnecessary stuff

A lot of packages aren't relevant to a VЕ setting, e.g. floppy disk utilities and kernel modules, even getty listening on the console.

 # kernel, kernel modules, documentation, mount points
 rm -rf /lib/modules /boot /dev/.udev /usr/doc /usr/info /media

 # packages not applicable to a VE
 cd /var/log/packages
 for pkg in \
    hotplug-* hdparm-* devmapper-* udev-* usbutils-* pciutils-* module-init-tools-* \
    mdadm-* floppy-* lvm2-* raidtools-* reiserfsprogs-* \
    smartmontools-* sysfsutils-* syslinux-* wireless_tools.* quota-* iptables-*
 do removepkg $pkg ; done

 # prune init's getty
 vi  /etc/inittab # delete everything after entry l6 (runlevel 6)
 init q

 # clean out the fstab and mtab files
 ( cd /etc ; rm -f fstab mtab ; ln -s ../proc/mounts mtab )
 echo "proc     /proc      proc     defaults    0  0" >> /etc/fstab
 echo "devpts   /dev/pts   devpts   mode=0620   0  0" >> /etc/fstab

 # the startup sequence and services
 cd /etc/rc.d
 rm -f rc.gpm-sample rc.gpm rc.hotplug rc.ip_forward rc.modules \
       rc.scanluns  rc.serial rc.udev rc.sysvinit
 vi rc.syslog    # delete all mentions of klogd
 vi rc.M         # delete the setterm entry
 vi rc.S         # delete the MOTD clobbering

Fix permissions and ownerships

 # clear out old/dummy SSL certificates
 mv /etc/ssl/openssl.cnf /tmp ; rm -r /etc/ssl/* ; mv /tmp/openssl.cnf /etc/ssl

 # set an ownership on any unowned files
 find / -mount -nouser -exec chown root {} \; &
 find / -mount -nogroup -exec chgrp root {} \; &

 # remove the setuid bit from programs which nobody else should use
 # you may want to review this list first, as some folks want their users
 # able to edit cronjobs and to change their own passwords, etc.
 for i in \
   /bin/ping /bin/mount /bin/ping6 /bin/umount /usr/bin/chfn \
    /usr/bin/chsh /usr/bin/crontab /usr/bin/chage /usr/bin/traceroute6 /usr/bin/traceroute \
    /usr/bin/expiry /usr/bin/newgrp /usr/bin/passwd /usr/bin/gpasswd \
    /usr/libexec/ssh-keysign /usr/libexec/pt_chown /usr/bin/wall /usr/bin/write
    do chmod u-s $i ; done

Changes to rc scripts

OpenVZ emulates rebooting with an external cronjob called vpsreboot and a dummy file called /reboot within the VE, and emulates the /etc/mtab file by pointing it to /proc/mounts So, some small changes are necessary to the rc scripts.

 # somewhere in rc.6 add this command: touch /reboot
 vi /etc/rc.d/rc.6

 # somewhere in rc.M, add this command: rm -f /etc/mtab ;  ln -s /proc/mounts /etc/mtab
 vi /etc/rc.d/rc.M

Blanking settings

Lastly, you'll want to delete or blank out a bunch of files so they start fresh when the VE is booted for its first time.

 # stop all services
 apachectl stop
 killall syslogd klogd udevd crond
 /etc/rc.d/rc.sendmail stop
 /etc/rc.d/rc.inetd stop
 /etc/webmin/stop
 /etc/rc.d/rc.pgsql stop
 /etc/rc.d/rc.mysqld stop
 killall named proftpd

 # blow away the network configuration with dummy strings for later replacement
 #    replace the IP address with __IPADDRESS_
 #    replace the netmask with __NETMASK__
 #    replace the GATEWAY with __GATEWAY__
 vi /etc/rc.d/rc.inet1.conf

 # disable the root and user accounts
 # by changing the password for root and user to a ! character.
 vi /etc/shadow

 # refresh the 'locate' cache
 /etc/cron.daily/slocate

 # blank out the system logfiles
 for logfile in \
     /var/log/messages /var/log/syslog /var/log/debug /var/log/secure \
     /var/log/maillog /var/log/spooler /var/log/proftpd.log /var/log/xinetd.log \
     /var/log/dmesg /var/log/faillog /var/log/lastlog /var/log/wtmp \
     /var/log/apache/access_log /var/log/apache/error_log \
     /var/log/webmin/miniserv.error /var/log/webmin/miniserv.pid
 do cp /dev/null $logfile ; done
 rmdir /var/log/sa

 # clear the SSH host key
 rm -f /etc/ssh/ssh_host_*

 # database server logfiles
 rm -f /var/lib/mysql/*.err /var/lib/pgsql/logfile

 # delete vi backup files, bash_history files, and other small application crud
 unset HISTFILE
 find / -name '*~' \
     -o -name .bash_history \
     -o -name .gnupg \
     -o -name .lesshst \
     -o -name .viminfo \
     -o -name .rnd \
     -delete

 # anything under /tmp
 rm -rf /tmp/*

Zipping it up into a cache image

A CT cache is just a tar.gz file of the entire filesystem, excluding some very dynamic stuff which gets populated by the OS at runtime anyway:

 tar --numeric-owner -zcvf /tmp/HostGIS_Linux_4.2_64bit.tar.gz --exclude='/sys/*' --exclude='/proc/*' --exclude='/tmp/*' /