Network Cable Disconnected? Not!

[houseisland]

I have a few XP Pro workstations which lose their network connection a couple of times a day -- the network connection shows cable disconnected. Link lights show an active link on both the NIC and the switch in the "disconnected" state.

All of the PCs are connected to the same main switch in the server closet.

The problem persists with the installation of different NICs in the workstations -- Intel, DLink, 3Com -- all the same. Drivers all up-to-date. Power saving turned off on NICs.

If the PCs are separted from the main switch by a little 5 port switch at the workstation site the problem does not occur.

The main switch in the server closet had a meltdown yesterday, taking down the entire network. It has been replaced. However, the problem of the PCs losing their connection to the network persists.

The PCs cannot be shut down while the connection is lost. If the PCs are left in the "disconnected" state, the connection will evenutually right itself. Most of the time, though, a forced ungraceful shut down is required because of time constraints.

PC event logs show nothing other than what would be expected under the circumstances.

I am thinking cabling problems..... But the cables all test OK for continuity, and there are no mismatched pairs. When the PCs are working, network connections seem quite fast.

I wonder if there are different wiring standards in the walls and ceilings different from those of the patch cables. I seem to remember reading that if one cable is A and the other is B, continuity will test OK and pairs will test OK but that there can be a gradual impedence build up in the cables that will cause a port to drop. However, I have never seen this happen in reality. Most places have very sloppy cabling and they seem to get away with it.



Thoughts?

[houseisland]

Interesting:

http://support.microsoft.com/default...;en-us;Q239924

It may be a media sense client issue, a result of cabling problems?

[Percy]

Can you replicate the problem on demand? For instance, by moving large files back and forth on the network? Also, are you using DHCP or static IP addresses?

[houseisland]

I cannot replicate the problem on demand.

Address are assigned by DHCP. I suppose I should try giving the problem workstations static addresses.

Note that it is only a few workstations that are affected and that these are identical to the ones that are unaffected.

[slgrieb]

Well, there could be problems with the cabling, but A and B standard wiring mixed in a circuit really shouldn't be a problem. If you have an A standard patch cord from a computer to a B standard wall jack, the send and receive signals still match up.

I've seen several installations where the modular plug wasn't crimped correctly so that the insulation jacket on the cable wasn't held tightly, and the cable untwisted over time (as well as some where the cable was just untwisted too much to start with) so the impedence of the cable, in particular, the receive pair (which is split) goes up as high as 200 Ohms. Some of these cables may test good with a cable tester and a 9 Volt battery, but will fail to work with the power levels used by NIC signals. I don't believe you would see this as an intermittent problem.

Since the new switch exhibits the same behavior as the old one, I would look for external sources of electrical interference. Do any cables run parallel to electrical conduits, are they draped over fluorescent ballasts, etc. Could be that the problem is as simple as a poor cable run and someone turning lights off and on.

[houseisland]

It is interesting that a little 5 port switch, acting as a repeater between the main switch and the workstation, makes the problem go away. This would lend credance to the theory of signal degradation as the root of the problem. The cables are for the most part short run, 50 to 100 feet max. But they are in the ceiling where there are electrical bundles and flourescent lights.

[houseisland]

Well ...... up into the ceiling.....

One bundle of cables was running directly over the balast in a large eight tube flourescent light fixture.



Could explain some of the problems. Luckily whoever pulled the cable left enough slack at either end that the bundle could be moved away from the fixture.

We shall see if there is a difference........