We have a pretty large ASP.NET 1.1 application at work that has about 100
programs total and about 60+ virtual directories. The application has an
editor program that lets use define the URL to each application. This
information is then somehow mapped to the user's login and the main menu
that they see on the screen. The application's editor expects all URLs to
start with http://servername/vdir/.
Anyway, to make a long story short, we use W2K3 Load Balancing for our web
farm. All works fine, but some of our customers have figured out ways to
bypass our DNS and access a particular server via that servers DNS name.
This really isn't a problem, except it defeats our load balancing or impacts
customers when we roll servers into/out of production for maintenance.
I.E., they access a specific server and we remove it from production... they
gotta try a different IP.
One of my recent challenges was troubleshooting a problem with one of our
clients. I had them access a specific server, which will bypass our virtual
load balanced IP address. This works, however, when they login and the main
menu displays, the main menu references the virtual IP or DNS name of the
server.... so when a link is clicked on a menu, they go back to load
balancer which points them to another server. Now they're basically on two
servers...
What I determined was that the editor application only allows http://
formatted URLs now not relative paths... which really doesn't make sense.
I'm currently have a long debate with our internal company writing our apps
to not expect or force users to use a FQDN for URLs. There is no reason why
I cannot use /vdir in place of http://servername/vdir when they're on the
same server/cluster.
I've basically proved this fact by manually editing the configuration file
where these URLs are saved and removing the http://server references and
replaced with blank.
The only issue I had was one Web Service that we call on any of our servers
that appears to be doing a string.parse() or something on the URL. It fails
if I do not have an http://servername defined. So the 1 program out of the
60+ programs I have to use a FQDN.
I know that either works fine when accessing web pages, but how do you guys
feel about this sort of non-standard practice. I never write design a
website and hardcode the FQDN for all my hyperlinks for pages on the same
server... so why should this application think it can get away with it?
Really stupid if you ask me.
