I am using BorderManager 3.8sp2 on a Netware 6.5 sp2 server to function as a
reverse proxy to an Exchange 2003 OWA front-end server. I have two
accelerated proxies setup, one for the front-end server on port 80, the
other on port 443. This works fine except during a certain situation. That
situation is during a reboot of the proxy server, or a reload of the proxy
service on the server. What happens is that https accelerated site
frequently fails to bind to the Exchange front-end server. When this is
happening, the server will frequently be able to bind to the accelerated
site on port 80. Other times it won't. The problem is made worse is there is
any volume of incoming attempts to go through the reverse proxy to the
front-end server. If there is little or no traffic, the binding will
succeed. I suspect it may have to do with the fact that the BorderManager
server will always resolve the ip address of the front-end server website to
it's published Internet DNS address, rather than the internal ip address
that it is actually connecting to. I have read Craig Johnson's tips on this,
and I have added a name to address mapping in the sys:etc\hosts file on the
server for the website pointing to the internal ip address. However, if I
perform an nslookup from the console of the BorderManager server, it
resolves the name of the website to it's Internet ip address. What's more,
if I look at the BorderManager Proxy Cache Server screen after proxy service
finishes loading, it shos that it is trying to fill the content for the site
from it's Internet address. This doesn't work very well, so I end up having
to go back to using the internal ip address in the configuration for the
accelerator in NWAdmin, which leaves me with the hit or miss issue on the
binding again. Is there any way to ensure that BM is going to use the ip
address in the hosts file to resolve name for this website. I haven't been
able to find anything so far that will allow me to force the server to do
so. Thanks.