[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: mii-diag returns strange errors



Wojciech Ziniewicz wrote:

I'll check the cable - i don't see any errors and users don't report
any problems except acces to the machine that has theese eth's
onboard... (transfer is breaking sometimes)

It still looks like a bad/noisy cable, or a bad autonegotiation.

From many many bad experiences with autonegotiation, I *never* use it on production servers, unless they're fed with fiber. Those tend to work correctly. :-)

Force both your machine and ask the provider to force the Cisco to the same settings (100/FDX, or whatever you like) and see if the errors continue. If they do, consider running new cable and make sure the cable meets the distance and other technical standards for 100 Mb/s Ethernet.

You can also test this theory without having to pull new cable, since you have one NIC that operates correctly. Swap the cables (both ends if you're just testing the cable, one end if you're testing the Cisco ports, etc... draw yourself a picture, you'll see how it works), and see if the errors continue on the NIC that originally had problems.

You just have to be cautious and have your Cisco admin available in case of trouble, so you can put things back the way they were prior to the test if something goes wrong.

If the problem follows the NIC card, you may have to start a search for updated drivers or other known issues with the card, or install something else that works.

If the provider has a good quality cable tester, and you can afford the down-time, have them test the cable. Continuity testers aren't good enough -- you need something that will measure reflectivity and other factors that can tell you if the end-to-end cable run meets the technical standards for Ethernet. I recommend the high-quality Fluke brand testers, but there are others that are just as good. And just because a cable tests out okay doesn't mean your NIC will work with it.

Other things I've done that are "quick and dirty" fixes that sometimes work include cutting the ends off the cable and taking 10 minutes to do a proper re-crimp job on the RJ45 connectors, so I know *I* did them, and they're done right with proper tools. If you have such tools available to you, it's an option.

I once saw a call center with a bundle Ethernet cables draped over the top of flourescent light fixtures and argued for hours with a CCIE about why that would (and eventually did) cause problems. Not to mention that it wasn't up to meeting Electrical Codes for that location.

This well-paid, supposed "Expert" about networking didn't have any working knowledge of Voltage, Induced Voltage (noise), or the inductive properties of cable, nor did he know that flourescent light fixtures have ballasts that create all sorts of electrical (and sometimes RF) noise. I was not impressed at his knowledge level, CCIE or none.

Good luck, your problem likely won't be that difficult to track down.

All of the above said, it could also be a driver issue with the card... hunt Google and other sources for reports about your particular Intel chipset cards.

Nate



Reply to: