4

About two weeks ago I replaced a switch because it had spawned a number of bad ports over the last year. Now, the new switch also has failing ports. Of course I'm looking into the environment. I've already ruled out power issues (tried installing a newer, good UPS before replacing the first switch). That leaves me looking at network cards (or a wiring fault) somewhere in the building. Unfortunately, it's not necessarily the same ports that have failed on the new switch as on the old. Any ideas for nailing down what might be causing this?

It could, of course, also just be coincidence, but I need to start exploring sources before I RMA the new switch.

Joel Coel
  • 12,932
  • 14
  • 62
  • 100
  • 1
    Ensure every pc connected to the switch is properly grounded? Any poe involved? –  Aug 21 '12 at 14:45
  • There is a 12 port PoE injector directly below the switch in the rack for a number of IP phones, but it's using the same UPS. I don't doubt a power issue is at the source of this, but the equipment in the wiring closet is at least well-checked. – Joel Coel Aug 21 '12 at 14:48
  • If the grounding path does not follow the ethernet back to the switch, you've just created a ground loop that acts like an antenna and picks up magnetic surges from nearby (not direct hit) lightning. If the grounds are separate, it picks up "step potential" from the ground, with current going from ground to ground. – Skaperen Aug 21 '12 at 14:48
  • "but it's using the same PoE" ... what does this mean? Are these switch ethernets connected to the PoE injector? – Skaperen Aug 21 '12 at 14:57
  • @Skaperen was a typo. It reads "same UPS" now. – Joel Coel Aug 21 '12 at 14:58
  • Then that simply adds all the PoE phones to the source mix in the room. This is actually good when the phones are not connected to power (or anything) at the distant end, but not perfect (you still have an antenna, although no "through current" risk). – Skaperen Aug 21 '12 at 15:01
  • It seems like any surge damage coming through the phones would stop at the PoE injector, rather than the switch. I'm inclined to think the problem is the elevator (or other source) creates a surge large enough to damage a port, but not enough to kill the switch. It hits and passes through the PC power lines, goes out the PC network line, and hits the switch network port. – Joel Coel Aug 21 '12 at 15:09
  • Surge does not "stop". When there is a voltage difference between two point, current flows in inverse proportion to the resistance. Run a wire between the points and you have a higher current. A surge starts with a voltage change somewhere. Current flows. The only wait to stop the current flow is extremely high resistance (disconnect the cable). A surge protector cannot stop current. It can eliminate voltage difference between all wires connected to it. – Skaperen Aug 21 '12 at 15:26
  • My bet is on crappy PoE injectors (insufficient clearance/isolation between power and data lines) that are letting a [ground loop](http://en.wikipedia.org/wiki/Ground_loop_%28electricity%29) sneak over to the data lines and burn out the PHY. You didn't by chance connect the PoE injector to your own power supply, did you? Was it designed to plug into exactly the power source it is actually plugged into? – David Schwartz Aug 21 '12 at 15:39
  • @Joel Where do the phones plug into, a separate phone switch, or are they IP devices that can run from any switch on the LAN as long as they get power? – Skaperen Aug 21 '12 at 16:51
  • Phones use the same wiring as the PCs: the PC at a desk with an IP phone will connect to a pass-through port on the phone. We have a dual-personality phone system, so not everyone in the building gets an IP phone. – Joel Coel Aug 21 '12 at 16:59

2 Answers2

2

The path is most likely this: far end device ground, far end device's power supply, Ethernet cable, PoE injector, switch port PHY, PoE injector, PoE injector's power supply, near end ground. This is a form of ground loop that is burning out the PHY by subjecting it to excessive current.

The most likely cause is connecting the PoE injector to a power supply other than the one it was intended to be connected to. This is common when you replace high-quality telecom DC supplies with cheap "wall wart" transformers. Not that you can't get away with that. You can. But you have to match characteristics.

Check for grounding of the wrong side of the PoE power supply. For example, if the power supply was designed to be driven from a positive ground power supply but was being driven from a negative ground power supply. Or if it was designed to be driven by an isolated power supply and is being driven by a negative ground or positive ground power supply.

Make sure the PoE injector is connected to precisely the type of power supply it was intended to be connected to.

David Schwartz
  • 31,449
  • 2
  • 55
  • 84
  • The injector has an integrated power supply. It's a 1U box with 12 ports that plugs directly to a wall outlet or UPS via standard power cord. – Joel Coel Aug 21 '12 at 16:13
  • Are ANY cables from the switch connected to PoE in any way (even if, and especially if, looped back from the far end where PC and phone are ... such as someone connecting PC and phone together)? FYI, sharing the same UPS as the switch is no more a risk than sharing the same outlet (but that is a risk either way, from surges coming in from other than power). – Skaperen Aug 21 '12 at 16:38
  • I wonder if the PoE injector's power supply could have failed some way that's causing it not to be isolated and the PoE was designed requiring an isolated power supply. I don't think that's very likely, but it's possible. – David Schwartz Aug 22 '12 at 01:52
0

I would be looking for possible electrical surge paths. Surges are "through paths" that can go from one point to another point (including ground). This include coming in on power and going out on an ethernet port ... or the reverse ... or even just between ethernets. A "surge protector" does not prevent this unless every metal path (yup, all ethernet ports, too), go through the same protector. The function of a "surge protector" is to equalize the voltage on all paths so the surge current does not flow inside the device.

Lightning from storms is not the only source of surges. Heavy motors that start and stop can also cause a brief surge.

Distributing networking wiring around a building is always a big surge attractor. If your environment is at high risk for surges at any time, and you cannot eliminate the causes, you may need to consider switching over to fiber optic.

Skaperen
  • 1,094
  • 2
  • 11
  • 23
  • > "`Heavy motors that start and stop can also cause a brief surge`". Interesting: we've had trouble with the elevator in that building for years. Tempting to throw that in the ring as the probable cause. – Joel Coel Aug 21 '12 at 14:49
  • When the motor contactors release, the current has to go somewhere else. It can literally reflect a high voltage back down its power lines. This is one reason you see sparks when relays, contactors, and even switches, are opened (when not contained to prevent exposure to the spark). This voltage can travel over power lines through the UPS and to the switch. A surge protector would stop surges that go in AND out on the power wires, but NOT stop them from proceeding out the ethernet ports (need to grasp this "flow through" concept of surges). – Skaperen Aug 21 '12 at 14:53