Machines dropping off the network randomly

[MK-2]

This is an extremely weird problem.
We've had a new ICT suite built, new data points (all FLUKE tested) and new HP machines.
For two weeks all was fine. As of last week, machines would randomly drop off the network. Link light on the switch goes off, machine says cable unplugged, can't ping it. If you switch off the machine and power it back on the link light comes back and all is OK.
It happens at any given time to any given machine.
At first I thought it could be DNS as we had someone working remotely on the server who was looking through DNS and made some changes.
Unfortunately I don't have a spare switch to replace and test if it's the switch. All I know for now is it's not the data points as they have been tested again yesterday and definitely work.
Anybody have any ideas?
Thanks

[siuko]

What sort of switches?

If they are managed is it anything silly like broadcast storm being on and dropping the connections due to lots of activity?

[MK-2]

they are allied telesyn switches. the thing is that this problem didn't occur prior to last week and no changes have been made on the switches. and really the only thing that changed was we had an engineer remotely looking at something for us and said he had made some changes in dns.
would dns changes stop the link light coming on until a reboot?

[K.C.Leblanc]

would dns changes stop the link light coming on until a reboot?

No.

Sounds like the switch is packing up. Is it gigabit or 100mb? If it's 100mb try altering the driver settings on the machines to force them to 100mb.

EDIT: Just had a brain wave, check to see if there are any power saving settings in the NIC driver. If there are disable them.

[MK-2]

the switch is 100mb, i'll try setting the speed to 100mb and see how we go.
just wish i had a replacement switch to try!!

[ricki]

Hi

I have the same problem with some stone machines. They all have intel pro 100 cards and are connected the various switches. The nic cards just appear to stops working. I have tried different cables points and drivers but that does not work. I have even tried ghosting machines and installing them again from scratch. But if I plug the machine into a different switch it starts working ok. I can then put the pc back and it might never be a problem or it could be back in a few hours.

As it only happens couple of times a month its not too much of a hastle

I have tried to find out what it is but have never got an answer. I do suspect that power surges might be the problem but have never managed to prove it.

Richard

[Butuz]

Sounds like a switch issue to me. Or it could be some kind pf power saving affecting the NICs on the PC's? Turn the power saving off?

Butuz

[MK-2]

i went back into the nic settings and there was a power management tab (which i must have overlooked before!!). all enabled to "allow pc to turn this device off to save power". so ive unticked that and see how we go. i was sure i'd looked in there before for power management and didn't see anything....
grrrrrrr

[MK-2]

not NIC power management. all are off and a further two machines have just dropped off.
looks like ill be buying a replacement switch then. bloody hope it is that otherwise i'm really stuffed.

[xathros]

Daft question but have you checked the switches to see if someone has put a loop in the system by patching a switch to itself?

[MK-2]

xathros, not a daft question at all! but i have checked.
i even rejigged my cab here to free up a switch that i know works and put that in there. just looking now as i get in and it looks like about 9 are offline.
if its not the switch i dont know what else it can be!!

*seems someone had switched off a few machines last night, so at least that accounts for the 9 being off*

*but it doesn't account for one just dropped off again. no ping reply, and this is with a different switch!*

[mac_shinobi]

on a few computers just to test if you set there bios back to factory defaults and then ensure Wake On Lan ( WOL ) is disabled in the bios and re set the boot order and ensure the bioses are password protected

Also try updating the NIC drivers to the latest revision and also Once updated the NIC drivers ensure the power management on the NIC drivers are disabled again then try again.

[ozzy]

Do you have a couple of spare network cards you could try for a while in some of the machines? sounds much like a problem I came across a while ago and it turned out to be dodgy nic's in them. 8 out of 15 new machines, not HP though

[MK-2]

very weird. just went into device properties to see which driver it was, and the power management was back on to enable device to be switched off.
i disabled it and something has turned it back on!

i know it might be silly, but the bios has the XD feature which reading about it it said some device drivers try to write to kernel stack which this feature disables. could that cause it? im seriously clutching at straws now