We have a locally written autoinstall system, that handles SunOS
5.[34], SunOS 4.1.4, and OSF/1 3.0. Essentially the same "finish"
script is used on each, with case statements to vary behavior as
needed; this leads to high cross-platform consistency. The SunOS
5.[34] portions lack many of the things in Casper (et al)'s system,
but I believe add some things as well. An Irix version is planned.
The SunOS 5.[34] version's pre-finish stage is a very slightly changed
Sun autoinstall. The SunOS 4.1.4 pre-finish copies itself; that is,
it boots diskless from a set of files, copies those same files to the
local disk, and tailors the local copy for local booting. The OSF/1
version boots into a ramdisk (via a slightly hacked RIS kernel), and
begins doing setld's (like pkgadd's). As alluded to above, all four
then chain into the same "finish" script.
The system makes heavy use of a local, NFS-mounted software library,
"dcslib".
One of the more interesting aspects of the system, is the usage of two
programs, "stamp" and "check-stamp", perhaps analogous to Tripwire.
Signatures of files (permissions, ownerships, filetypes, checksums of
regular files, targets of symlinks, major and minor numbers of device
files) are noted and recorded in a local, gzip'd file. This system
allows for a fast, automated inspection of changes that have been made
to a machine, since its initial install, removing a significant
portion of the guesswork associated with performing upgrades on
machines for clients with "low maintenance" contracts (if you know all
your machines intimately, it won't help you much; we don't always have
that luxury here).
The system may be inspected (at least for the time being, if there's
overload I'd need to shut this access down) via NFS mounts of:
autoinst.acs.uci.edu:/auto_install/OSF-1-3.0
autoinst.acs.uci.edu:/auto_install/Solaris-1.1.2
autoinst.acs.uci.edu:/auto_install/Solaris-2.3
autoinst.acs.uci.edu:/auto_install/Solaris-2.4
autoinst.acs.uci.edu:/auto_install/generic-after
Comments on the system are desired.
I have two questions for the list:
1) Of those of you who are successfully using an autoinstallation
procedure, how many of you are performing periodic, automated
upgrades/re-installs? How many are continuing to do
upgrades/re-installs on a case-by-base basis? What historical
considerations and politics have gone into your respective decisions?
2) Have any of you encountered problems reading/writing files of mode
2700 from/to a Solaris 2.4 NFS server? SunOS 4.1.4's /usr/kvm/crash
is the example that brought this to my attention. Is there a patch?
Received on Thu Apr 13 1995 - 02:26:02 NZST