Installing August 2004 Update Rollup on this Cluster?
I've got a two node Active/Passive Exchange 2k cluster at a client. I didn't
set it up, nor did the guy before me (IIRC).
With Exchange 5.5, you had to install Exchange ON the shared resource; there
was no other way to do it.
With 2000, I guess the recommendation is that you do NOT install on the
shared resource, however, there's nothing preventing you from doing so. The
cluster in question was installed with Exchange on E:, which is a
shared/clustered (fiber channel) disk.
I got involved with all of this after seeing the April security update for
Exchange 2k SP3. This cluster didn't even have SP3 installed. I started
there. I couldn't perform the SP3 update on the "offline" node, because the
install location for Exchange is on a clustered disk that it can't access...
You can probably see where I'm going with this.
I was able to get SP3 to install OK on each node - I did the node, then took
the virt. server offline so I could reboot the node (w/o having Exchange try
to fail over to the "not upgraded" node). It rebooted and then Exchange came
up all OK. Did the same thing on the other server, and I've now got SP3
across the board. This is fine.
The April security update (MS05-021, KB 894549) requires the "August 2004
Post-SP3 Update Rollup" (KB 870540) to be installed first. The installer for
this is NOT the same Exchange-type installer - it behaves more like a Windows
service pack installer.
It complains that I should "stop all clustered resources and then stop the
Cluster Service, and manually restart it after installation" when I normally
run the install file. Since Exchange is installed to a cluster disk, I can't
really do this now, can I?
I checked the command-line switches for the updater, and by passing the
"/passive" switch (no user input but display progressbars) it did actually
start the install process. However, the first thing it did was try to stop
running services... one of which was THE CLUSTER SERVICE. This obviously
would not do, and I was able to cancel before it made itself die (it would've
unmounted the E: drive by stopping cluster services, and the installer was
trying to run from there...)
So what can I do here? Do I need to reinstall Exchange on each node, putting
it on a local disk instead?
If so, is there a procedure to do this with minimal hassle? I don't want to
lose my stores by uninstalling (I don't think it will touch that though), and
I'm hoping that I can somehow uninstall Exchange from one box without
actually removing the files from the clustered drive, to allow me to keep
Exchange up on the other node in case the first node has some catastrophic
problems... I know that I can just back up the stores to tape, but I'd like
to keep the actual app on the shared storage so I can go "one leg at a time"
on the cluster. Plus, if I uninstall Exchange on one machine, and it deletes
the files from E:, what will happen when I try to uninstall on the other
node? Won't it be annoyed that its files are gone already? I'd need to
manually clean the registry, etc. to get it all out so I can reinstall...
I'm at a loss here and could really use some help. I thought (naively) that
the Security support line would give me a hand since this is all related to a
SECURITY UPDATE, but they pointed me to the newsgroups...
date: Mon, 18 Apr 2005 13:05:01 -0700
author: Braden McGrath
