RE: Failover/Mirror for the III sever?

[adabbah at nypl dot org]

At New York Public Library we have an Alpha ES40 production server with
RAID and a DS10 training server with a cold swap disk cabinet.  We
installed a RAID controller in the DS10, so that in the event of
catastrophic failure of the ES40 CPU, we plug the production RAID into the
DS10 controller.  Innovative then changes the IP address on the training
server (in addition to a few other adjustments to the OS) and we are back
in business.  We've only had to use this once, and it is not automatic
(human intervention is required on our part and Innovative's), but it
serves as a fairly effective failover system that can ensure relatively
short interruptions of service in the event of serious problems.  If your
production and training servers are identical, you have less of a problem -
the latest backup tape from the production restored to the training and a
change of IP address on the training server, and you would be functional
again.  An automatic failover would essentially require a different kind of
system.  You would then not be talking about a production/training set up,
it would be what Innovative calls a 2-pac system, so that the servers are
continuously updating in real time (a much more expensive proposition).
---------------------------------------------------------------------------------------------------------

Aaron Dabbah, CATNYP Systems Administrator
The Research Libraries, New York Public Library
Fifth Avenue and 42nd Street, New York, NY 10018-2788
Phone: (212) 642-0135  Email: adabbah at nypl dot org
----------------------------------------------------------------------------------------------------------