Cisco 3560X Capacitor Failures

[jason.boyer]

Hi all,

I teach at a networking academy at a community college and we have a few donated 3560X switches supplementing our student lab. Six are deployed in pods. The room has adequate climate control

In the last 10 months we have had two fail, announcing their failure in a most pungent manner, the second in the last week or so. I popped the case to determine the cause, and in both cases it seems the same capacitor, a 47 uF 100V cap. Both switches were powered, and neither was carrying traffic, nor were any PoE devices connected when the failures occurred.

I searched briefly but did not see any other notices about this. One was manufactured in 2010 and seemed to have a bit more of an interesting history; the other manufactured in 2012 and was clean.

Given that 33% failed in the same way, we expect additional failures. I wonder if anyone else has experienced this, or if Cisco has any additional history to share?

R/

Jason Boyer, Information Systems Instructor

Reedley College

[Leo Laohoo]

That's why we kept a hundred of our 3750X for spares since this model is past end-of-support date.

[jason.boyer]

In a production environment, that will keep you going. These are for students to get some hands on experience with Layer 3 switches. It's a bit of an inconvenience and certainly a safety concern.

[Leo Laohoo]

I understand the concern about safety, however, the concern about learning is a little bright -- Everyone is dumping old 3560X/3750X in the used market.  Cheap.  

[Ramblin Tech]

Two units with arcing/failing capacitors is quite odd. Are you located in an area prone to thunderstorms and lighting strikes? Any copper/Cat5 cabling exiting the building? Have you checked the grounding of your power circuits? Are your units on a UPS with power surge protection?

[jason.boyer]

Great points, Ramblin.

We are in California's Central Valley. Very few thunderstorms, and the copper runs are contained in the building - the lab is contained in the same building between two adjacent rooms (one for the equipment, the other with the student computers). It has its own climate control and security.

The equipment runs on rack mount UPSs, one UPS per bay. New electrical circuits were run when the racks were installed two years ago and so far those have been the only issues. I can't completely rule out power stability as the wiring in these buildings is older, but we haven't seen any other electrical issues thus far. The IDF at the other end of the building houses similar equipment, and with computers and other sensitive devices on the line, I would expect to see more instances of similar failures.

I would also expect damage to the power supply modules if the problem were from the AC mains. The modules were pulled and tested in a working 3560 and fire right up with no problems.

I am still suspicious of the caps - perhaps a batch of bad ones that are sensitive to dielectric punchthrough?

[Ramblin Tech]

The voltage feeding the PCB and electronic components from the PSU is almost certainly in the 3 - 12VDC range, so it is quite surprising to see the extent of the damage. I am still inclined to believe that there was a high voltage surge into the gear, with the power circuit or copper ethernet cables being the most likely paths. Can a cap of that size discharge enough current to cause all that damage?

I am at a loss at this point, as you appear to have already checked all the usual suspects.

[Georg Pauwen]

Hello,

I don't want to add anything redundant to this discussion, but I guess the problem with donated equipment is that you don't know where it came from, that is, in which environment it was operating. The damage shown on the pictures looks like the switches might have been installed in a closet with bad ventilation/high humidity, or experienced power surges while not being connected to a surge protector.

[jason.boyer]

Good points. One for sure was in a harsh environment. The other was quite clean inside and didn't show any *obvious* signs of previous overheating, but that is a definite possibility.

These devices have run here for about 3 years, powered down for 1-2 months every summer. Aside from that, they run continuously with no PoE load and carrying little to no traffic in a clean, climate controlled lab.