Exchange Server - Add a node with borrowed hardware

Asked By CK

27-Oct-09 04:21 PM

I have a Exchange 2003 sp2 cluster and I was wondering the following. I need
to replace the backend servers since the procs are  a little under powered.
Currently we are connecting to our SAN via ISCSI using a QLogic ISCSI HBA/TOE
cards. I was wondering 2 things:

1. If I add a node using a different kind of ISCSI HBA, is there a potential
conflict that could cause corruption between the server and the storage? The
ISCSI HBA is several years old and with recent improvements to the ISCSI
initiator i'm not sure that there is reason to pony up for an HBA.

2. If it is deemed necessary to have the same model HBA, I was thinking of
doing this: Shutting down our passive Node. Pulling the ISCSI HBA and
installing it on the new box. Bring up the new box. Add it to the cluster to
see how things are working. If there are no problems then I will evict the
old node from the cluster and then move on to the next back end server. If
there are problems, then I can alwayes roll back.

Let me know what you guys think.

27-Oct-09 05:55 PM

Don't use iSCSI HBAs, they are pointless these days unless you are iSCSI
booting, and even then not so much.

What's the corruption risk  you are thinking of?

Make sure Exchang is fine on the active. Evict the node so that it
ends up as a one-node cluster. Shut down and remove the box. Bring the
new box up and get it iSCSI connected to the server. Allow that iQN to
access the shared LUNs and then install Exchange.

There will not be any rollback because having one would actually be more
dangerous as well as unecessary. Nothing says success like Cortez and
the whole boat burning thing.

Done properly there is not a risk at all. Who is your SAN vendor? Talk
to them and their PS arm will help you get it all done in the right
order.

09-Nov-09 07:20 AM

Hi,

1. Usually it is deemed 'best practice' to have the same model HBA's and the
same firmware revisions.

However in this instance I would simply consult the iSCSI SAN vendor to
confirm new model HBAs that are compatible with your storage are OK.

2. In regards to upgrading, you can add a new node to make your EVS A/P/P -
however in your instance you want to re-purpose an HBA which will leave you
on a single node until you can present a new node to the cluster. This
obviously has risk, if this is the method you choose rather than ratifying
that a new node with new HBA is supported and then presenting that to your
Cluster, I would ensure you do not remove the old cluster node - so if you
cannot get your new node working you can install the HBA back in it and
bring it online.

Ensure each element of risk is presented to management and signed off.

Me - I would speak to my SAN vendor and get that HCL off them for HBAs.

Oliver