ad-hoc link causing loop
Ok, here's the deal, me and a friend are going to be linking up eachothers local networks with eachother via an ad-hoc link. The ad-hoc link is with 2 cisco 350 aironet pci cards (100mW), eventually we will be attaching parabolic antenna's to these but that's a way's off yet. So I brought my server to his house, set it up and attached a computer to it so we could figure out how to ping my client computers from his and vice versa. So first step is create an ad hoc network (same as using wireless bridges) with the two cisco cards with static ip's 192.168.0.252 and 253. That went perfect, the servers could see eachother, all was happy. Then came the time to set up routing and remote access with static routes (windows 2003) so that we could forward requests from my local network (on a different ethernet card of course) across the ad-hoc link to his local network. After some head-banging we finally got the routing right and could see that routing requests were being fulfilled when we pinged, but the requests would always time out. That made no sense, finally by dumb luck I happened to look up at our switch and notice that activity on the network was pegged. I found that if I disabled the rf radio (the ad-hoc network) then network traffic would stop. This leads me to believe that somehow something is causing an internal loop in the intranet we are trying to set up because once the ad-hoc link is up I can't even ping my client computers from my server. So, basically, I have come to the point of not being able to look beyond my own nose at this problem, it's driving me bonkers (we spent 12 hours straight working on it last night only to run into this). So I'm going to provide you a visual representation of what we want to do hoping that someone better at this than me (or with a clearer head) can shed some light on the issue. Thank you very much, if anyone figures this out I owe them a coke :).
here's an update.
basically, we've narrowed it down to a problem with routing and remote access on the servers. Basically, it does route between the cisco cards and the 10/100 nics, however, nothing will go past either of the 10/100 nics. At one point during this evening I was able to set static routes on a client machine on his network and ping the 10/100 on my server, but was not able to ping that client from the server. So, incoming requests are not going through those nics. We have no idea why. Rip and OSPF should both make everything automatic and it's not working correctly at all. We even plugged both 10/100's into the switch and turned off the cisco cards to see if we could do it that was just to test, still no luck. So tomorrow we're going to format and try server 2000 instead and see if it's any better (I'm guessing 2003 has some sort of bug ::shrugs::). If anyone out there has experience with routing and remote access via windows server please give me some feedback, I'm dying over here. It's driving me insane because I can set up routing on an actual router till the cows come home...but windows doesn't want to cooperate. Thanks again for any help you guys can give me.
Your setup doesn't sound particularly unusual so you would probably get help in a specialist forum such as microsoft.public.windows.server.networking or microsoft.public.win2000.ras_routing.
I have no experience with the scenario so the following is just speculation. Do the wireless cards need to be on their own subnet? If possible, could you give them IP addresses on their respective lan subnets? Example 126.96.36.199 and 10.0.0.100. This would remove the complexity of the extra subnet.
thanks for the suggestions, I actually resolved this issue sunday, putting the grand total hours spent at ~20. Basically, he has dialup on his home network, and the machine is running ICS. So for everything to work correctly he needed to set client computers default gateway to that computer and not the server in the dhcp pool. This meant that they ignored static routes set on his server and forwarded return echo requests (pings) to that computer and the server, in turn causing it to time out because the computer running ICS had no clue how to get back to my network. This explained why adding static routes on each client would allow me to ping them and vice versa. So we just added a static route of 00.00.00.00 255.255.255.255 to the computer that has the dialup so that any requests to the internet would go to it, and everything else to the server. So everything works great now (we even played counter strike against eachother across the networks with a maximum of 5ms latency).
And just so anyone else who reads this looking to do what I did, here's the skinny:
Two computers running windows server 2000 or 2003. Turn on RRAS but disable remote access (otherwise your server steals ip addresses from dhcp). Each server has at least one nic connected to the local lan (a second nic is good for load balancing) and a wireless nic in ad hoc mode. So there are three subnets of a class c network (we used 10.0.x.x). Each lan is on a different subnet, and the wireless cards are on their own subnet. Microsofts implementations of RIP and OSPF do NOT route across more than 2 subnets correctly (because they're lame). So, one must add static routes to tell lan 1 how to reach lan 2, this is done by making a route that says any traffic intended for lan 2 must go to the wireless card of lan 2's server as the default gateway. Also, it is important that no matter what, dhcp clients must have their default gateway set to the server and not the source of the internet otherwise all hope is lost. Only the servers must point their default gateway to the internet, and in our case, only my lan has internet so lan 2's server points it's default gateway to my wireless card.
My advice is, use hardware routers if it is feasible because their implementation of RIP and OSPF are more standardized and will avoid static routes and other MS-induced headaches.
Hopefully this will someday save someone a GREAT deal of suffering.