N5K - locate source causing output errors

[cableguy]

Hello, 

we're receiving output errors on two interfaces of a Nexus 5672.  

Setup: Two Nexus 5672 in one vPC domain, multiple C2232TM FEX connected, each  Nexus 5672 has an uplink to N7K core switches. The N5Ks are in cut-through switching mode, MTU on N5Ks is default 1500, MTU on N7K connected to the N5Ks is set to 9216. 

A colleague from the server departement told me that devices attached to one of the FEX report input CRC errors. I checked the interface error counters on all N5Ks, one of the uplinks of the affected FEX reports output errors. I saw that another interface of the same N5K has simultaneously increasing output errors, that is the uplink to one of the N7Ks. 

Cisco advises to look for interfaces that report input errors, which identifies the source. But there are no other interface errors on all N5Ks and N7Ks except the described two interfaces with output errors.  

I assume, that some device is sending jumbo frames, as the N7K with MTU 9216 has no errors at all on the interface where the N5K's uplink with output errors is connected to. Jumbo frames could also explain, why devices attached to the regarding FEX report input CRC errors, as the NICs of these devices have an MTU of 1500. 

I wonder how to locate the source causing these output errors?

Any hint is appreciated.

Best Regards

[marce1000 · Accepted solution]

 - You may find this thread informational :

                https://community.cisco.com/t5/routing/output-errors-on-port-nexus-5020/td-p/3339391

 M.

[cableguy]

Thanks for the link, that thread was very helpful in troubleshooting this issue.

It turned out that one of the SFP+ transceivers of the affected FEX caused the errors. Weird, because double-checked all transceivers' Tx and Rx power, they looked good. One of the transceiver had a Tx power level quite close to warning threshold, but still "good enough". I replaced it, and the errors immediatly disappeared.