I am a serial NIC Killer

[bkasten]

Most of the computers on my network are Linux. One of ways I use unRAID is to have common folders on these Linux clients, for all the larger areas (ie Downloads, Documents, Videos, etc.). I find this very convenient. I mount the remote folders via the /etc/fstab.

 

So far, so good.

 

Now, on my primary workstation (and mine alone) I have been killing NIC's at a pace of about 1 every 6 months.

 

At first I used the on-board NIC, RealTEC 8111E.

 

http://www.newegg.com/Product/Product.aspx?Item=N82E16813131646R

 

That lasted a little more than a year, about 14 months. So then I installed my first Intel NIC (that will fix it!). Well, it didn't. In the last 10 months I have put in 3 Intel NIC's Like this one:

 

http://www.newegg.com/Product/Product.aspx?Item=N82E16833106122

 

Which is to say, I have killed two Intel NIC's so far. About 1 every 5 months. I am on my third card now.

 

==========================

 

All my Linux clients are configured this way, but only on my workstation do I do video editing/processing, over the network. I am sure that is what's killing my cards. None of the other workstations are experiencing the NIC killing fun. Only my workstation.

Am I expecting too much from my NIC's?

 

Should I really not be editing/processing video over the network?

 

Would I perhaps do better with a PCI-e NIC?

 

http://www.newegg.com/Product/Product.aspx?Item=N82E16833106033

 

The culprit programs would be:

makemkv

DVBCut

vobcopy

 

It is so convenient to work across the network, but I may have to change my work-process, or maybe buy stock in Intel.

 

I am looking for any feedback on this issue.

 

Am I the only one editing/processing video over the network... and destroying NIC's?

 

Thanks

 

bkasten

[JonathanM]

I find it hard to believe any normal software usage, even if extreme amounts, would cause outright failure. The only causes I can think of are electrical, or software attacking the firmware. Different vendors and cards would seem to rule out the firmware issue, so I can only go back to electrical.

 

Could you please describe your wiring, both high and low voltage? What I'm asking for is the physical layout, as in, workstation in one room on a surge protector, network plugged into a patch cord plugged into a wall outlet that runs to a patch panel in the server room, with a patch cord plugged into a 16 port netgear gige switch powered by a battery backup.

[dirtysanchez]

I find it hard to believe any normal software usage, even if extreme amounts, would cause outright failure. The only causes I can think of are electrical, or software attacking the firmware. Different vendors and cards would seem to rule out the firmware issue, so I can only go back to electrical.

 

Could you please describe your wiring, both high and low voltage? What I'm asking for is the physical layout, as in, workstation in one room on a surge protector, network plugged into a patch cord plugged into a wall outlet that runs to a patch panel in the server room, with a patch cord plugged into a 16 port netgear gige switch powered by a battery backup.

 

+1.  Completely agree. 

 

After 16 years in IT (and many more as a geek) I can say that THE ONLY THING I have ever seen take out a NIC is an electrical surge of some sort, lightning mostly.

[WeeboTech]

Or static electricity...

 

 

perhaps there is a floating ground somewhere in the machine or network.

[JonathanM]

Or static electricity...

 

 

perhaps there is a floating ground somewhere in the machine or network.

My money is on the grounding issue. I worked on one network where if the machine was unplugged from the wall power, but plugged into the network, you would get a pretty good shock from the case. Turns out the other end of the network cable was connected to a switch with a different grounding point, and the difference in ground potential was enough to shock you. Needless to say, the network communicated very poorly over that link.

[dirtysanchez]

Agreed.  I've seen exactly that situation before.  It never killed the NIC, but I'm sure it certainly could have.

 

I have seen a number of lightning strikes take out NICs.  Mind you the strike didn't necessarily have to hit the site itself, it only had to be nearby, and the static buildup in the air somehow got into the structured wiring and blew out dozens of NICs.  It happened numerous times at this particular site until extensive grounding eventually solved the issue.

 

Not saying that lightning in particular is what's causing the OP's issue, just that as you guys also suspect, it's related to an electrical surge of some sort.

 

 

[bkasten]

I find it hard to believe any normal software usage, even if extreme amounts, would cause outright failure. The only causes I can think of are electrical, or software attacking the firmware. Different vendors and cards would seem to rule out the firmware issue, so I can only go back to electrical.

 

Could you please describe your wiring, both high and low voltage? What I'm asking for is the physical layout, as in, workstation in one room on a surge protector, network plugged into a patch cord plugged into a wall outlet that runs to a patch panel in the server room, with a patch cord plugged into a 16 port netgear gige switch powered by a battery backup.

 

I will try and answer all the questions so far:

 

On my workstation, it, and everything else computer-wise is plugged into a surge protector, a Triplite Isobar, and that is plugged into a APC Battery Back up., XL 1300.

 

As for the network, the computer NIC is plugged into a Netgear Pro-Safe GS105. From there, the switch is plugged into the wall-mount connection. I had a professional Network installer wire my house. Next, the wiring comes out in the "server" room, the utility room for the house, furnace, water heater etc. In this room I have all my servers and other IT style equipment. This then plugs into am Apple Airport Extreme router. It is then plugged into a Netgear 8 port GS108, which the unRAID server is them plugged into.

 

Electrically, in the Utility room, all equipment is plugged into another battery backup like before (same model). To get enough plower plugs, I have multi-port surge protectors for all equipment.

 

When I burned out the first Intel NIC, I plugged the next card into a different PCI slot on the motherboard, just to make sure that PCI slot was not "bad". No difference.

 

When the NIC is working, I get fantastic through-put, in the upper 90's for read/writes to the cache drive. 40's to 50's for protected disk writes.

 

This has stumped me for some time.

 

bkasten

[unevent]

From the gs105 to the workstation, how long is the patch cable?

How old is the gs105?  How old is the patch cable?

Is there anything else plugged into the gs105 other than the patch to the wall?

Are you using shielded or unshielded patch cables (including the in-wall cable runs)?

Do you have any Ethernet cables plugged into a surge suppressor of any kind/type, anywhere?

Do you get transmit/receive errors reported on the workstation in question?

 

[bkasten]

From the gs105 to the workstation, how long is the patch cable?

How old is the gs105?  How old is the patch cable?

Is there anything else plugged into the gs105 other than the patch to the wall?

Are you using shielded or unshielded patch cables (including the in-wall cable runs)?

Do you have any Ethernet cables plugged into a surge suppressor of any kind/type, anywhere?

Do you get transmit/receive errors reported on the workstation in question?

 

Patch cable is 4 ft.

The gs105 is about 2 years old.

The patch cable is less than 1 year old.

I have my work laptop plugged into the gs105 as well.

The patch cables are all Cat 6. If that is not automatically shielded, then they are unshielded. In the wall should all be shielded

I have no Ethernet cables plugged into any surge suppressors anywhere.

 

Output from ifconfig on workstation:

$ ifconfig eth2

eth2      Link encap:Ethernet  HWaddr 90:e2:ba:4d:d3:37 

          inet addr:10.0.1.24  Bcast:10.0.1.255  Mask:255.255.255.0

          inet6 addr: fe80::92e2:baff:fe4d:d337/64 Scope:Link

          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1

          RX packets:429787660 errors:38 dropped:2720 overruns:0 frame:19

          TX packets:224595612 errors:0 dropped:0 overruns:0 carrier:0

          collisions:0 txqueuelen:1000

          RX bytes:629460081348 (629.4 GB)  TX bytes:298357404550 (298.3 GB)

 

[lionelhutz]

To troubleshoot I'd first unplug the cable from the PC and measure the AC and DC voltage from the connector to ground. A meter with decently sharp probes should be able to fit between the dividers and connect to the terminals.

[bkasten]

I find it hard to believe any normal software usage, even if extreme amounts, would cause outright failure. The only causes I can think of are electrical, or software attacking the firmware. Different vendors and cards would seem to rule out the firmware issue, so I can only go back to electrical.

 

Could you please describe your wiring, both high and low voltage? What I'm asking for is the physical layout, as in, workstation in one room on a surge protector, network plugged into a patch cord plugged into a wall outlet that runs to a patch panel in the server room, with a patch cord plugged into a 16 port netgear gige switch powered by a battery backup.

 

OK, now I think I am on to something.

 

It would appear jonathanm was right on from the first. I did describe my setup, but I left out one thing, because it has been here for so long, it was kind of invisible.

 

I have a thirty year old florescent desk lamp, on the same isobar as my computer. Every time I turn it on or off, my USB hub makes it's connection/disconnection noise. never really thought about it, but it having some affect on the computer if it is causing enough disturbance to make the USB hub react. The USB hub is not plugged in, so the only way it can be reached is through the computer.

 

Could that be the real issue?

[JonathanM]

I have a thirty year old florescent desk lamp, on the same isobar as my computer. Every time I turn it on or off, my USB hub makes it's connection/disconnection noise. never really thought about it, but it having some affect on the computer if it is causing enough disturbance to make the USB hub react. The USB hub is not plugged in, so the only way it can be reached is through the computer.

 

Could that be the real issue?

I think you may have a winner. The ballast in that old lamp is a BIG coil transformer type affair. The surge created from turning it on and off is not well isolated from the PC, since it's on the same surge protector. Try running that lamp straight to the wall outlet and see if it still interferes, if so, get rid of it, or leave it on all the time. Most modern fluorescent lamps are much better behaved, they use electronic ballasts that don't use huge magnetic coils.

[JonathanM]

I have seen a number of lightning strikes take out NICs.  Mind you the strike didn't necessarily have to hit the site itself, it only had to be nearby, and the static buildup in the air somehow got into the structured wiring and blew out dozens of NICs.

It's not static buildup in the air, it's an EMP event. Any length of wire that passes through a magnetic field or a magnetic field passes through it will have a voltage generated along it. That is how transformers, and generators work. In the case of a lighting strike, a huge "wire" is momentarily created from the cloud to the ground, the electricity gets dumped along that wire, and as it passes, a huge magnetic field is created and dies. Any wire nearby will have a voltage generated on it. If one end of that wire isn't grounded, the voltage will seek the easiest way to get to ground, usually frying anything in its path.

 

Most telephone and cable modem lightning damage occurs because the lightning hit a power pole and travelled alongside the telephone or cable conductor until it got to ground. The telephone or cable line rarely takes a direct hit, it's almost always induced current.

[dirtysanchez]

Excellent explanation johnathanm.  Thanks!

[bkasten]

I have a thirty year old florescent desk lamp, on the same isobar as my computer. Every time I turn it on or off, my USB hub makes it's connection/disconnection noise. never really thought about it, but it having some affect on the computer if it is causing enough disturbance to make the USB hub react. The USB hub is not plugged in, so the only way it can be reached is through the computer.

 

Could that be the real issue?

I think you may have a winner. The ballast in that old lamp is a BIG coil transformer type affair. The surge created from turning it on and off is not well isolated from the PC, since it's on the same surge protector. Try running that lamp straight to the wall outlet and see if it still interferes, if so, get rid of it, or leave it on all the time. Most modern fluorescent lamps are much better behaved, they use electronic ballasts that don't use huge magnetic coils.

 

I have moved the lamp off the isobar, and it no longer is making my USB Hub signal. I am taking that as a sign that it is no longer interfering with the computer.

 

Hopefully this will fix the issue.

 

I would like to thank everyone that offered help on this... It was driving me crazy.

 

bkasten

[bkasten]

I have a thirty year old florescent desk lamp, on the same isobar as my computer. Every time I turn it on or off, my USB hub makes it's connection/disconnection noise. never really thought about it, but it having some affect on the computer if it is causing enough disturbance to make the USB hub react. The USB hub is not plugged in, so the only way it can be reached is through the computer.

 

Could that be the real issue?

I think you may have a winner. The ballast in that old lamp is a BIG coil transformer type affair. The surge created from turning it on and off is not well isolated from the PC, since it's on the same surge protector. Try running that lamp straight to the wall outlet and see if it still interferes, if so, get rid of it, or leave it on all the time. Most modern fluorescent lamps are much better behaved, they use electronic ballasts that don't use huge magnetic coils.

 

I have moved the lamp off the isobar, and it no longer is making my USB Hub signal. I am taking that as a sign that it is no longer interfering with the computer.

 

Hopefully this will fix the issue.

 

I would like to thank everyone that offered help on this... It was driving me crazy.

 

bkasten

 

And the final verdict appear to be in.

 

Not the lamp

 

Not the Isobar

 

Not the network cable

 

Not a bad port.

 

It appears to be the NetGear gs105!!!

 

Gone through a few more cards, the last one lasting less than one month. Replaced the gs105 switch, with another one, and the gigabit speed came back to my computer, and has stayed. I had tried different ports on the switch too, so it's not an individual port, it's has to be the whole switch.

 

Just thought I would let everyone know.

 

Bruce

[lionelhutz]

Did you ever measure the voltage like I suggested?

 

I would suspect the power supply has an issue. It might be passing 120VAC to the ports or some switching transient noise to the ports.

[bkasten]

Did you ever measure the voltage like I suggested?

 

I would suspect the power supply has an issue. It might be passing 120VAC to the ports or some switching transient noise to the ports.

 

I must confess, I did not.   

 

In my defense, I got a lot of good suggestions, which I followed up on, but checking the voltage of the switch power supply fell off the radar. Too bad too, it was in retrospect, brilliant. I am thinking your idea of "transient noise" is plausible. The cards themselves seem to be fine, and operational. They just cannot use that switch.

 

I have been pretty busy, new job and all, but one of these weekends I will dig out my vintage multi-tester and take some measurements.

 

I think that if the switch were passing too high a voltage the cards would be destroyed. As it is, they seem to be working fine, and at full speed too.

 

Thank you

 

Bruce

[lionelhutz]

If the cards started working again with a new switch, then that would just indicate the switch quit working and didn't actually kill them.