[comp.os.vms] shutting down to install

MCGUIRE@GRIN1.BITNET ("The Sysco Kid ", McGuire,Ed) (06/03/88)

> Date: Wed, 25 May 88 00:51 MDT
> From: WHEELER@CC.UTAH.EDU
> Subject: What resources does an idle process tie up ... or why have
>
> By the way, does anyone *really* make all the user's log off and shutdown
> the network to install new products?

It depends upon the installation.  We install with the system up and during
regular working hours when possible.  This makes our clients happy and our
staff happy.

For example, we typically install maintenance FORTRAN releases that way.  We
announce the unavailability of FORTRAN two weeks beforehand.

We install ALL-IN-1 releases off-prime and with the system down.  They
typically require adjustments to system parameters as part of the installation
procedure, and our clients depend upon ALL-IN-1 availability during the
business day.

Before I decide whether to install an update with the system up, I find out
what files are modified by the update.  I consult the Customer Support Center
for their advice if I am unsure.

Just to be sure I don't miss anything, I always use the VMSINSTAL option `L',
which causes the installation procedure to log important file operations.

There are certain VMS files often modified by software updates that require
special thought.

For example, if the system help file SYS$HELP:HELPLIB.HLB is modified, a
collision may occur between someone reading help and the update.  It has never
happened at Grinnell.  I simply watch for error messages.

We use a common system disk for our cluster.  If SYS$SHARE:DCLTABLES.EXE is
modified, I go to other nodes of the cluster and manually INSTALL REPLACE it.

If the update being installed is supposed to live on several cluster members,
it may be necessary to reinstall shareable images on the other nodes, change
SYSGEN parameters, or redefine logical names.  If there isn't a lot to this
I'll do it by hand after the installation.  The more screwing around required,
however, the more likely I am to do the installation with the cluster down and
then reboot.

Comments?

Ed