Weird Cabling Issue: Cat65xx+x6148-RJ45V & x6348

[AndrewAllen]

I have a 6509. In it are some 6148 line cards and some 6348 line cards. Most devices I connect to those cards that are more than 160' away will not establish a link at 100/Full. 10Full or Half or 100/Half works fine. Now, I plugged an old 6248 in there and it handled all 4 permutations well into the 200' distance.
If the devices connected to the 6148s and 6348s are around 150' or less, they work ok at 100/Full.

I have not been able to find anything defninitive on this after a number of Google searches. Has anyone else seen or heard of anything like this? Previously I just assumed that if you were running Cat5/5E and the cable/terminators were in good shape you could consistently get 100/Full well into the high 2xx' range.

Also note that once we started down the voltage troubleshooting path we made sure to build clean cables that did not pass by or amongst power cables or any other sort of ambient electronic noise. Still failed as noted above.


Andy Ford
Austin

 

[ADB100]

The standard says 328 feet @ 350Mhz over Cat5E, this equates to just under 100-Metres, so I think 150-metres is stretching it a bit. If you attempt to use cable lengths this long then you can pretty much say goodbye to any support on these connections.

As to why the 6248 has better tolerance, I can only guess it's due to different components being used in the different linecards. Cisco only need to make them work on Cat5E cable runs of up to 100-metres; why should they use components that can tolerate longer runs?

This probably isn't much help, but this would be Cisco's response I think.

Andy

 

[jneiberger]

I think he said he's been trying lengths up to 150-200 feet, not meters.

This does seem unusual. It sounds like you've already tested the cables to make sure they're up to spec, so I'm not sure why those blades would behave that way. If all of those cables work fine on the other blade, though, that does seem to point toward the 6148 and 6348 blades.

 

[ADB100]

I think he said he's been trying lengths up to 150-200 feet, not meters.

Ahh, mis-read that ;o)

Well 100-Metres is 328-feet so yes it should work.

I assume you are running both the switchport and the attached device at Auto speed & duplex? Once this didn't negotiate 100/Full you then attempted to hard-code both the switchport & the device?

I have seen issues where PC's drivers ignore the duplex settings and always run Auto. However this is unlikely to help.... You are using Cat5E patch cables?

Have you got access to any cable testers such as a Fluke etc that will give you a detailed report about the cable? It might give you some idea?

I think there must still be a tolerance difference between the 6148 & 6248, either that or you have been really unlucky and have some faulty 6348 & 6148 modules.

I suggest you get a detailed report from a Fluke or equivilent cable tester and if the cable checks out (include the patch cables at either end), then raise a TAC case.

HTH

Andy

 

[AndrewAllen]

ADB100,

Re: Fluke/cable testers, we tried several different Cat 5E cables, both self built and mass-produced. Same behavior.

Re: Speed/Duplex, this problem initially surfaced due to a customer. They had our product, in some instances, over 150' away from the switch. Their installation of 6500s had 61/62/6348s. On the linecards that failed the distance, they initially hard strapped our device and their switch port to 100/Full. No link light. No nothing. Like no cable was plugged in, or like nothing on the other end. If they went to 10/Anything, it would work. 100/Anything. No work.

My main intent of this post was to see if anyone had seen this behavior. Since the problems follow the 61 and 6348s I assume there is something particular to them, voltage-wise. The 6248s are rock solid at 100/Full well past 200'.

Since this is a customer issue, we're trying to figure out what the root cause is so we can advise our customers accordingly. Just saying, "Well, 10/half works, use that." doesn't cut it.

Concerning the TAC case, doesn't that require an active service contract????

Andy Ford
Austin

 

[whykap]

You said you have tried several cables self built and maas produced. Have you or the customer certified the cable with a CAT-5e tester and has it passed the certification?

I would eliminate this first. That way if the wiring is good you know for a fact that it is the blade. Without a cable certification and an active cisco support contract (seems like you don't have one) you will be just spinning your wheels.