I had this posted in the firewall board so I moved it to cache... Pleas Help...
I have a ISA 2004 EE array with 3 servers. I added 2 servers over the last 2 weeks. The orginial server is acting as the configuration storage server (although I dont know if that is pertinent). I am using these only for reverse proxy of one internal website and all are on the same subnet. They all 3 show up on the monitoring and all 3 say they are synced and happy. However when I am loadtesting against them the only server that handles any load is the main server. the other 2 sit there and do nothing. I am able to watch the perf mon on the other 2 and the dont do a thing.
All the cache drives appear to be set up ok and caching is enabled.
Any help would be appreciated. My first post but have ref'ed the site many a time over the implementation. Feel free to insult my intelligence.
I seem to be narrowing down some possibilities for this... I was reading some posts from Tristan on ISAuser.com and It looks like I dont need NLB for caching only, which is what we are trying to do... can anyone confirm this?
I am at a large company 60 - 80K users... so it seems that it would be difficult to configure all those browsers. (this is for a webcasting app so I potentially hit all browser) this is why I am interested in server side carp. But it doesent seem to be working very well. I.E. I dont see sessions being spawned in the monitoring screen to the other array members.
Am I missing somthing fundamental about how server side carp works?
CARP, when functioning correctly, has both a client side and server side component.
If auto-discovery is working correctly, then the clients will know that Node1 serves a particular range of websites and Node2 servers another, and so on...
The way it works is that when the clients pull down the WPAD file, WPAD contains a hash from each node of the array. This hash is used by the client to determine which node "most likely" has a particular URL cached, and will send the request to it.
If auto-discovery isn't working, or isn't setup, then the clients send the request to the proxy arrays address (which is resolved through a host record that points to all IP's of the nodes). Since you can't guarantee who will receive the request (let's say Node1 receives the request), Node1 will service the request and talk to the other nodes on the intra-array adapter to retrieve the content from whover is supposed to have it.