SUMMARY: Installation identical Tru64 systems ...

From: Thomas Leitner <tom_at_radar.tu-graz.ac.at>
Date: Thu, 30 Nov 2000 09:58:04 +0100 (MET)

Hello,

Sorry for the late summary. I got lots of replies. The bottom line is,
that this is possible this way and many have done similar things.
The installation was either done via network, an externally attached
hard-disk (gold-disk) or via tape. Compaq folks also pointed me to
the installation cloning feature in Tru64 which seems easy to use.
However, I wanted to clone a machine with local customizations as well
as the appropriate patch kit already installed. This seemed not possible
with the standard installation cloning feature.

Many people also sent me their cloning scripts which gave me valuable
hints for my own script.

My original idea, swapping the Tru64 installation CD with my own recovery
CD-ROM doesn't work because the Tru64 installation CD is mounted and
many utilities are needed from it. It is possible to create a special
bootable CD-ROM but I decided against it because it seemed too much work
as a special dedicated harddisk is needed for that.

So this is how I do it now:

On the master machine:

- vdump /, /usr and /home to files, save disklabel to file disk.label,
  copy /etc config files to a config subdirectory. Together with a
  "recover.sh" script (see below) burn everything to a CD.

On an arbitrary machine in the network called "nfsserver":

- Mount the recovery (or cloning) CD-ROM created above and make it
  NFS mountable by the clone machine.

On the clone machine:

- boot Tru64 4.0f installation CD-ROM
- Press "UNIX Shell."
- Enter these commands to start cloning
  NOTE: nfsserver is 10.10.10.1, clone is 10.10.10.2

      echo 10.10.10.1 nfsserver >/etc/hosts
      ifconfig tu0 10.10.10.2 up
      mkdir /var/recover
      mount nfsserver:/cdrom /var/recover
      cd /var/recover
      sh recover.sh

  After that, just set the date with the "date" command and
  set the appropriate console variables for automatic booting etc.

>>> set boot_osflags A
>>> set bootdef_dev dka0
>>> set auto_action boot

And that's it.

Appended is my recover.sh script as well as the replies I got.

Thanks a lot to the following people (sorry if I forgot somebody) for
their replies:

    "Hummers, Rick (Frederick)" <hummers_at_enterasys.com>
    Tobias Burnus <burnus_at_gmx.de>
    Sean O'Connell <sean_at_stat.Duke.EDU>
    Tom Reu <Thomas.Reu_at_compaq.com>
    Gerhard Nolte <gnolte_at_gwdg.de>
    alan_at_nabeth.cxo.dec.com
    Richard Loken <tech_at_athabascau.ca>
    Debra LaGattuta <debra.lagattuta_at_compaq.com>

---------------------------- MY RECOVER.SH -------------------------
#!/bin/sh
#
# Script to reload a machine configuration from CD-ROM.
#
# Copyright (c) 2000, Dept. of Communications, Graz University of Technology
# e-mail: support_at_wiis.tu-graz.ac.at
#
#
PATH=$PATH:/bin:/sbin:/sbin/advfs;
export PATH
printf "Please enter the name of the system disk (e.g. rz0) : "
read sysd
if [ "$sysd" = "" ]; then
    sysd=rz0
fi
printf "Press RETURN to start the recovery process.\n"
printf "WARNING: All data on the target disk will be lost: "
read a
pwd=`pwd`
cd /dev
./MAKEDEV $sysd
cd $pwd
rdev=/dev/${sysd}a
rrdev=/dev/r${sysd}a
sdev=/dev/${sysd}b
udev=/dev/${sysd}g
scu -f $rrdev write media starting 0 limit 5m pattern 0x0
disklabel -z $sysd
disklabel -rw $sysd $sysd
#disklabel -R -r -t advfs $sysd disk.label /mdec/rzboot.advfs /mdec/bootrz.advfs
disklabel -t advfs -r -R $rrdev disk.label UNKNOWN
rm -rf /etc/fdmns/root_domain
rm -rf /etc/fdmns/usr_domain
mkfdmn $rdev root_domain
mkfdmn $udev usr_domain
mkfset root_domain root
mkfset usr_domain usr
mkfset usr_domain home
if [ ! -d /var/mnt ]; then
    mkdir /var/mnt
fi
if [ ! -d /var/mnt/root ]; then
    mkdir /var/mnt/root
fi
if [ ! -d /var/mnt/usr ]; then
    mkdir /var/mnt/usr
fi
if [ ! -d /var/mnt/home ]; then
    mkdir /var/mnt/home
fi
mount root_domain#root /var/mnt/root
mount usr_domain#usr /var/mnt/usr
mount usr_domain#home /var/mnt/home
echo restoring /
vrestore -x -f dumps/root.dump -D /var/mnt/root
echo restoring /usr
vrestore -x -f dumps/usr.dump -D /var/mnt/usr
echo restoring /home
vrestore -x -f dumps/home.dump -D /var/mnt/home
rm -rf /var/mnt/root/etc/fdmns/root_domain
rm -rf /var/mnt/root/etc/fdmns/usr_domain
mkdir /var/mnt/root/etc/fdmns/root_domain
mkdir /var/mnt/root/etc/fdmns/usr_domain
ln -s $rdev /var/mnt/root/etc/fdmns/root_domain
ln -s $udev /var/mnt/root/etc/fdmns/usr_domain
rm -f /var/mnt/root/sbin/swapdefault /var/mnt/root/sbin/swapdefault.orig
ln -s $sdev /var/mnt/root/sbin/swapdefault.orig
#
# install recovery stuff on /home/recover
#
if [ ! -d /var/mnt/home/recover ]; then
    mkdir /var/mnt/home/recover
fi
tar cpf - . | (cd /var/mnt/home/recover; tar xpvf -)
#
printf "Which machine is this wxrdecw, wxrdeck, wxrdeci or wxrdecs: "
read m
ETC=/var/mnt/root/etc
FILES="exports hosts ntp.conf passwd rc.config routes motd"
for i in $FILES ; do
    if [ -r $ETC/$i ]; then
        mv $ETC/$i $ETC/$i.old
    fi
    echo cp configs/$i.$m $ETC/$i
    cp configs/$i.$m $ETC/$i
    chmod 644 $ETC/$i
done
#
# licenses
#
echo Installing Tru64 licenses .....
lmf -d /var/mnt/usr/var/adm/lmf delete OSF-BASE
lmf -d /var/mnt/usr/var/adm/lmf register - < licenses/${m}-base.pak
umount /var/mnt/root
umount /var/mnt/usr
umount /var/mnt/home
echo FINISHED ........
#
# end of file
-------------------------- ORIGINAL POSTING ------------------------
From: Thomas Leitner <tom_at_radar.tu-graz.ac.at>
Subject: Installation identical Tru64 systems ...

Hi,

I have five identical machines (identical hardware), I want to install
Tru64 4.0f on. In order to save some work (and to save the setup for
recovery purposes later), I thought that I setup a machine as required and
try to duplicate the installation to the four other machines.

I thought that the following steps may be sufficient:

- On the already installed machine:
    + vdump /, /usr and /home to files.
    + save the "disklabel -r" output to a file
    + write a small script to restore the disklabel
       and install the boot block
    + burn all files to a CD-ROM in ISO9660 format. The CD-ROM
      is called recovery CD-ROM

- Now on the client machine which is to be installed:

   + boot from the Tru64 4.0F installation CD-ROM
   + open a shell window
   + take out the installation CD-ROM and insert the
     recovery CD-ROM and mount it.
   + run the script to label the fresh disk and create the filesystems
   + vrestore everything to /, /usr and /home

Question: Is this feasible? Any more ideas anyone?

Thanks // Tom
------------------------------- REPLIES ----------------------------------
From: "Hummers, Rick (Frederick)" <hummers_at_enterasys.com>

This is almost exactly what I did to meet my y2k upgrades, except
I was going to V4.0D and I just used a spare 2G read/write disk
rather than a CD.  Over the period leading up to and following y2k I
installed over 3 dozen client and server OS upgrades.  I found that
the R/W disk worked well since I found that I was 'tweaking' the 'final'
system gold disk quite often.   I had it customized with virtually
everything
I needed including NetWorker, LPS, automount files, ntp, hosts, routes,
shells, aliases, .rhosts, .cshrc, .login. profile, even the new motd...

I would harvest a bunch of files off the old OS over the network
for reference (I can send the script on Monday when I get back in the
office)
I have a ba353 (flat) three slot StorageWorks box for my 'gold disk',
a cd (firmware upgrades got done as needed during the process) and
a 'scratch' disk that I just carried to the target system.    I built
a small bootable 'a' partition, a small 'b' swap area (vdump/restore want
swap space even on large memory systems) on my gold disk and had the gold
files
in a 'g' partition.  Boot the gold disk '-fl s', mount the swap and gold
partitions,
disklabel, newfs, mount, cd /new_root and  vdump -f /gold/root.vdump ...

I had several different 'disklabel' model files depending on the system
drive one would work (my systems were not identical).  After slamming the
new OS, I booted genvmunix, changed rc.config, fstab, and hosts then
built a new Kernel.    Booted into the new kernel,  reconfigured sendmail
and  restored other things that were not on the gold disk and we were
done.

\s\Rick
--------------------------------------------------------------------------
From: Tobias Burnus <burnus_at_gmx.de>

Hallo,

"Best Practice for Installing Large Numbers of Systems"

http://www.tru64unix.compaq.com/faqs/publications/best_practices/BP_INSTALL_CLONING/TITLE.HTM

könnte ein guter Start sein.

Tobias
--------------------------------------------------------------------------
From: Sean O'Connell <sean_at_stat.Duke.EDU>

You could also netboot the target machines, disklabel as needed,
mount the source machine's partitions via nfs, and vdump. Not too
bad if you already have an ris environment set up (fall over to the
unix shell to do all the dirty work).

S
--------------------------------------------------------------------------
From: Tom Reu <Thomas.Reu_at_compaq.com>

If you look in the installation guide, you will find references
to "Installation cloning" which enables you to do much the same
with less effort.

Regards,
Tom Reu
Tru64 UNIX Installation Team
--------------------------------------------------------------------------
From: Gerhard Nolte <gnolte_at_gwdg.de>

Hi Tom,

das sollte so funktionieren, hab ich jedenfalls auch schon so
durchgefuehrt. Im Anhang schick ich Dir Skripts und
Handlungsanweisung (Rezept fuer mich, damit ich nichts vergesse) zu. Ich
hab allerdings nicht alles komplett automatisiert.

Eine Bemerkung noch:
Ich habe die zu installierenden Rechner erst mal als diskless Clients
gebootet (--> DMU). Damit kann man dann Konfigurationsdateien auf dem
Server editieren (zB disklabel Output) und danach auf dem Client verwenden.

Die zu clonenden Filesysteme hab' ich dann natuerlich per NFS gemountet.

Viel Spass
Gerhard
--------------------------------------------------------------------------
From: alan_at_nabeth.cxo.dec.com

        Seems reasonable. If you want to get fancy, copy the
        installation CDROM to a partition large enough that
        you can modify the contents. Then track down the scripts
        that start the installation procedure and change them
        to drop you directly into a shell. Trim out subsets and
        other things you don't need to make space for the backups.
        Verify the partition is bootable and usable, then burn
        a copy of it.

        Extra credit: modify the installation to do some of the
        recovery semi-automatically, after checking where you
        want the system disk to be.

        A problem you might run into switching the CDROM, is that
        programs are probably loaded as needed from the installation
        CDROM. Once you switch out the installation for yours you
        may not have access to all the utilities you need. The
        method works fine as long as the backup media is separate
        from the installation CDROM, but with both on CDROM, it
        may have problems. Hence my suggestion to use a common
        piece of media.
--------------------------------------------------------------------------
From: Richard Loken <tech_at_athabascau.ca>

I have done something similar using a tape and not bothering with the CDrom
and then I did a couple little twists on the clones and I was off and running.

You shoud do fine.
--------------------------------------------------------------------------
From: Debra LaGattuta <debra.lagattuta_at_compaq.com>

Are you familiar with the Installation Cloning feature?
Installation cloning sounds ideal for your situation
because you perform one Full Installation on one machine,
and then clone the rest of the machines.

Installation Cloning lets you duplicate the installation
characteristics (that is, the file systems and installed
software) from an installed system onto one or more systems with
the same or similar hardware configuration.

When you install the current version of the operating system
on a machine, the installation process automatically
generates a configuration description file (CDF) that
contains a record of the installation setup data you specified.
This file is created in the /var/adm/smlogs directory under the file
name install.cdf. The install.cdf file contains all the installation
information required to perform the same installation on one or more
target systems.

This feature is documented in the Version 4.0F
Installation Guide if you want to learn more about it.

Regards
Debra LaGattuta
-- 
--------------------------------------------------------------------------
Dr. Tom Leitner                             Dept. of Communications
                                            Graz University of Technology,
e-mail    : tom_at_radar.tu-graz.ac.at         Inffeldgasse 12
Phone     : +43-316-873-7455                A-8010 Graz / Austria / Europe
Fax       : +43-316-463-697
Home page : http://wiis.tu-graz.ac.at/people/tom.html
PGP public key on : ftp://wiis.tu-graz.ac.at/pgp-keys/tom.asc or send
mail with subject "get Thomas Leitner" to pgp-public-keys_at_keys.pgp.net
--------------------------------------------------------------------------
    Before we have the paperless office, we have the paperless toilet!
Received on Thu Nov 30 2000 - 08:59:27 NZDT

This archive was generated by hypermail 2.4.0 : Wed Nov 08 2023 - 11:53:41 NZDT