Networking Frustration

[Shinei]

I've been banging my head against the wall for two years with this chronic problem. I bought an entirely new NIC to replace the old one, thinking that was the cause; the problem persists, regardless of NIC.
So, this is my problem:

I have a cable modem (old Motorola Surfboard), which, for the duration of what I will describe, indicates that it is powered on, connected to the network, and sending/receiving information from the outside world.
This modem is, in turn, connected to a D-Link DGL-4100 wired GbE router; this also will indicate network activity, on any connected ports (there are two, besides the WAN, for this description).
The router connects to two separate ethernet devices: A 3COM 3C2000 wired gigabit card, and an onboard Realtek RTL8169/8110 ethernet device (the device manager lists this as a GbE part, but this motherboard is an nForce2 board, and I doubt it's actually GbE).

Now, for the problem. During any internet usage, the network will suddenly drop off on the 3COM card, without any cause or indication; I'll be watching a video off YouTube or something, and the feed will stop streaming, or I'll be on AIM and get disconnected for no apparent reason. Originally, I attributed this to ISP throttling, since I only used to suffer this when I downloaded large amounts of data in a short period of time, but the problem has been occurring lately without any heavy bandwidth applications running (IRC and Firefox hardly qualify as saturating a gigabit network).
If I disable the 3COM device through the Network Connections window, I can reacquire the network with the RTL8169, but this will also fail after about an hour of usage, for no apparent reason.
Restarting resolves the problem temporarily, since it typically blows up after about an hour of use, and will continue to do so after any reboots.

Things I've ruled out:
Modem (like I said earlier, it's powered on and getting full connectivity for the duration of these incidents)
Router (also indicates connectivity during the network blackouts, and other computers can use the router after the blackout starts)
Static IP address (problem persists regardless of IP being static (as it is configured on the 3COM), or dynamic (as it is configured on the RTL8169))
Power management issues (I disabled the "allow computer to disable this device to save power" option, which reactivated the network for a short while but ultimately led to the same problem)
Cables (the network is working fine until the moment the network devices stop being able to receive any data, so it's not a cable problem)

Things I'm not sure about:
I can't /release or /renew either device once they stop being able to receieve data from the network. I can't figure out why, and it just times out during the operation.
3COM's diagnostic indicates that the NIC is working fine, but obviously it isn't since it isn't receiving any data.

Things I've done to try to fix the problem:
Replace the NIC. I've gone from an old Netgear 10/100 card to a D-Link 10/100 card, to the current 3C2000 card, and the problem persists across all NICs I stick in the computer. The problem also applies to the integrated network device, which is puzzling.
Replace the router. The DGL-4100 is an upgrade from an old, old Linksys 10/100 wired router, and the problem was just as bad on the old Linksys as it is now.
The only thing I haven't tried yet is replacing the cable modem itself, but I suspect this is a software issue more than a hardware issue, considering all the things I've ruled out so far.

Relevant hardware/software:
Windows XP Professional, SP2 with all patches (except for the latest WGA patch)
GigaByte GA-7N400 (nForce2 chipset)
Athlon XP 3200+ (Athlon XP 2800+ with a minor overclock, AGP/PCI frequencies locked at 66/33)
3COM 3C2000 GbE network card
Realtek RTL8169/8110
BFG 6800GT (not sure if this is relevant per se, but I want to rule out power fluctuations)
Antec 430W TruePower PSU

So, the conclusion: How do I fix this problem so that I don't have to keep restarting every two hours to stay on the internet?

 

[nweaver]

1. Disable one of the cards. Only use one at a time, or you are asking for trouble.
2. Start a set of pings, one to your default gateway (your router). One to the default gateway of the router (next hop, onto the ISP's network), one to a common web site, such as google, and one to a root DNS server. When you lose access, find where you are losing pings, that will start to show you where the issues exist.

 

[Shinei]

I hadn't thought about the second NIC being connected as an issue, since the problem had existed before I attached both. The only reason I had both wired in is because I had bridged them a couple of months ago, in the hopes that bridging might at least keep me from suffering lack of service by failing over to the functioning network connection; this solution didn't really do anything, and I just haven't disconnected the integrated device because I've been using the "one breaks, disable it and use the other" method.

As for your second suggestion, I haven't been able to talk to the outside world at all once the blackout hits. The devices just stop being able to receive data, and 3COM's diagnostic confirms this (the link icon is blue, the transmit icon is green, but the receive icon is empty). I'll try the gateway ping the next time I black out, but I won't be able to give you an update right away (such that I can't actually communicate with the internet in the meantime). Thanks for the quick reply.

 

[nweaver]

dual gateways (that would be the setup with both nics) = bad mojo scooby.

Plus make sure you have all that bridging crap disabled.

I would set up the pings now, and minimize them. You might ping your own IP as well.

 

[Shinei]

I've been trying to ping the gateway as it's listed in the connection information (192.168.0.1 for D-Links, I guess), but it keeps timing out on all ping requests. The connection status for the 3COM says it's still sending and receiving data though, so I'm not sure why it's not getting the ping replies.
I can ping my own IP and localhost fine, though, since I just tried both options.

Probably in a twist of irony, the problem hasn't resurfaced since I made the thread. I'm sure it will soon, though.

Edit: Why is having two NICs connected a problem? I thought Windows switched to whichever had a stronger connection?

 

[spidey07]

Windows does not support more than one default gateway. That's bad mojo and cause a whole nightmare of sporadic connectivity problems.

There is never a reason to use more than one NIC unless there are very specific reasons to do so and one fully understands the consequences of doing it.

 

[Shinei]

Both NICs are rigged to connect to the default gateway at 192.168.0.1; is this still causing a problem? The network was suffering the same problem before I connected the second NIC...

 

[spidey07]

Absolutely. Plus you bridged them. Another no-no. Essentially you've created a bridging loop where frames never die and circle around constantly. In fact this bridging loop may be the root of your problems given that you see activity but can't communicate.

If I were you I'd take one card out and completely remove it from the system. Then clean up your network connections (make sure you reboot between every change).

That might just clean up your "mess". Then possibly re-install your service pak.

Windows has a nasty habit of getting a TCP stack competely fubarred, so your stack may be having problems. All kinds of things could be going on. The one thing though is the more you mess with it, the worse it gets. Wouldn't be a bad idea to search microsoft on how to reinstall/repair your TCP/IP stack.

 

[Shinei]

I've since unbridged them, since the problem wasn't fixed by bridging; at the moment they're two, separate, standalone connections. I'll try scrubbing out the networking things, but this problem's been persistent across three installations of Windows.
Additionally, the first bridge I built can't be deleted because I removed the constituent connections from the bridge properties manually, rather than dissolving the bridge. Is there a way to remove that bridge? Every attempt I've made so far tells me that there was an unexpected error that prevents the bridge from being deleted.

Edit: The problem is also intermittent, since I'm currently posting from the afflicted machine with no problems. As a matter of fact, it hasn't collapsed at all since I posted this thread, which is kind of annoying because I can't take screencaps of what happens when it stops working.

 

[spidey07]

Unplug everything, focus on a single machine.

Post ipconfig/all and "route print" commands.

 

[Shinei]

This is the only computer I have, though. :s
Before I go entirely into diagnostics mode, could you tell me how route print works? Of all things, networking is not my strong point.

 

[spidey07]

it's a command you run in a command prompt.

start/run - cmd

It's troublesome that you can't ping your default gateway (router). That definately says the provider isn't involved and it's your home gear.

Another thing - you said "it was working before so it's not a cabling problem". That is not correct. Cabling problems or homemade cables can exhibit these kinds of problems as well and should be considered the number one cause.

 

[Shinei]

It's more puzzling than anything else. I can't ping the gateway, but I can ping the other NIC device just fine, and I can also communicate with the internet anyway (after all, I am posting here from the busted machine).
I'll try route print now and see what happens. I'll edit in the results, so watch this space.

Edit: http://img134.imageshack.us/img134/6830/routeprinteh6.jpg

 

[Shinei]

Originally posted by: spidey07
Another thing - you said "it was working before so it's not a cabling problem". That is not correct. Cabling problems or homemade cables can exhibit these kinds of problems as well and should be considered the number one cause.

The fact that it works, and then doesn't, should rule out the cables, shouldn't it? It just seems like any electrical problems would be apparent all the time, not just some of the time.

 

[spidey07]

Originally posted by: Shinei

Originally posted by: spidey07
Another thing - you said "it was working before so it's not a cabling problem". That is not correct. Cabling problems or homemade cables can exhibit these kinds of problems as well and should be considered the number one cause.

The fact that it works, and then doesn't, should rule out the cables, shouldn't it? It just seems like any electrical problems would be apparent all the time, not just some of the time.

No. First mistake when dealing with cabling problems.

Looking at your "router print" output, you still have two network cards active. As far as windows/stack is concerned.

 

[nweaver]

rule #1, it's not the network
rule #2, check the physical layer

 

[Shinei]

I did the route print before disabling one of the devices; even with one disabled (the 3COM failed in the interim of spidey07 posting and this reply), the route print doesn't seem to tell me much.
I also tried rebuilding the TCP stack with netsh, which didn't solve the problem, either.

Things I can't do in a blackout:
Re-enable the disabled devices. It reports "connection failed" when you try to re-enable the disabled device.

 

[spidey07]

Originally posted by: nweaver
rule #1, it's not the network
rule #2, check the physical layer

Rule #3, it's an application problem. Pound rule number 1# into their thick, dumb skulls.

 

[dmcowen674]

Sounds like packet fragmentation.

You're probably still set on dial up MTU instead of 1500 Broadband MTU.

DrTCP

Tweak Test

That should fix you up.

 

[Shinei]

Update: MTU setting is 1500, per the ping test results (this is set on the router as well); these settings have been like this since the devices were installed.
Also, I disabled the onboard NIC through the BIOS, and am only riding the 3COM for now; at the moment it doesn't appear to me that there are any problems. Thanks for all the help, guys.

 

[her209]

I've had this happen to me a couple of times. I would be browsing the web, then Windows would freeze up for a few seconds and then come back and my network connection would be gone even though the link lights are still on. A restart would fix the problem. Although I never actually bothered to figure out what was causing Windows to freeze, my hunch would be a dying hard drive.

 

[vailr]

I'd suggest: try DSL for 60 days, instead of cable. May be less expensive than cable.
If you don't like the DSL, then there's always "free cable modem after rebate" deals, when returning to broadband via cable.
Also, check for the latest NIC driver at:
http://www.station-drivers.com/