hello all
I am on a Wireless ISP. recently we are facing some disruption. After having a complete health check, we have found lotts of microsoft-ds stufs. i have stopped those loosy packets bypassing my network. but recently there are thousands of arp packets hitting the third internal LAN card of our Gateway server. but using iptables, there are no way to drop or reject those packets.

Acctually our network is now almost flooded with those arp requests. can anybody help me to solve the issue?

thanx in advance.

You rather need to find out who is that and stop that machine/person. Even if you block the flood on your server, the network will still be not very usable.

How? You know the subnet. If subnet is related to physical area, you have the area. Now... write down a moment of such traffic (using tcpdump or similar tool). Analyze it - maybe you'll be able to find out who's sending it (user ID?).

Thnx for your reply.
I know from where those requests are coming. but the prob is, it wont be possible for me to physically disconnect them from my network. coz they are some banks and they use some kind of sensative software like Royeter or something. They need to be online every single ms.
so, all i have to do is, cleare my network from my end. this is the main prob.

we use ALLOT's NetEnforcer. It's not possible also from this end to stop those arp requests. nor from my main linux boxes.

i'm in deep ocean of problems. give me some clue to find the land.

thanx again.

You use wireless. It means you use radio. It also means that anyone can interfere with your transmission. That's sad fact. So the only thing you can do is to try to do something to move the problem to small part of your network, as small as possible.

The most important thing is if you can separate the 'messy' part (small area etc). If you can, install a new switch to handle those clients that cause all that. Switch will handle ARPs (or should, at least). If you have one big subnet, divide it. Those 'messy' clients should be in one - ARPs are probably requests to get (effectively) all machines in a subnet (I'm guessing, you should know for sure). So, when you separate them to one subnet, there will be (1) smaller subnet, so probably lower number of requests and (2) they will be handles inside the small subnet.