ProCurve 2448 NATed connections: Disconnects

[john_1114]

Hi Folks
I have a terminal server setup for users to access our network remotely. At present, a user connects from the internet to a load balancer on our network (2X - formerly Gateway Pro) then from there connecting to a terminal server. (because the onnections are through our firewall the packets are NATed)
The problem I am having is that users are disconnecting randomly. I think it has something to do with the HP ProCurve 2848 switch I have.
Has anyone come accross this type of thing before?
Any Suggestions would be most appreciated.
Thanks in advance
John

[john_1114]

Oh yeah, I forgot to mention we are using Windows 2003 SP1 on the terminal server.
Thanks
John

[Matt Hobbs]

Is there any reason you feel like the issue is related to the 2848?

Are the users being disconnected from the VPN session? (is it a VPN?) or just from the terminal server?

I'd be checking the terminal server and switch event log and looking for any clues there to start with.

[john_1114]

The reasoning why I thought it may be my 2848 is by process of elimination.

I don't think it's my terminal server itself that is having a problem. There is no evidence of this in the server logs, and, we have a number of people connecting to it across the WAN (without going through the load balancer) and they never have any connection problems.

That got me to thinking to look at the load balancer itself, I tried chaning the network card, but, alas, it had the same problem.

This got me to thinking to the switch.

Since my original post I have hard coded the load balancer and switch port on the 2848 to 100M Full Duplex, and things have been running much smoother since then, with far less drops in connection.. which is confirming to me that the problems lie in the switch. I have also tried changing the network cables to all devices concerned.

In answer to your other query,
The users who connect from the internet do not connect through a VPN, they connect using encryption on their TS client.

Thanks for your help
Look forward to hearing from ya soon
John

P.S. I have now reset the stats on the switch port, to see how many errors etc I am getting

[Les Ligetfalvy]

Have you optimized the buffers on the 2848 with the QOS Passthrough?
See http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=1002810

I cannot help but think that it may be related to the firewall and load balancer. Many firewalls don't take well to traffic that hops from source to source. You also said "people connecting to it across the WAN (without going through the load balancer) and they never have any connection problems". I don't see how changing a NIC on the load balancer rules it out.

Run a packet trace and look for syn-ack mismatches.