We've been experiencing a really annoying issue with our two 6807s (each with two SUP2-2T and two C6800-32P10G modules, XL variant), ever since we got them some five years ago: every once in a while ports start failing the TestNonDisruptiveLoopback test and traffic stops flowing properly through those ports (this gets especially serious when that traffic is PVST for a large-traffic vlan or when neighboring devices err-disable their ports because of improper UDLD operation).
It has happened on both routers, on all four modules, so I don't think it's a hardware failure. Also, there's a pattern, ports don't just fail randomly - if we divide the 32 ports on the module in quarters (odd 1-15, odd 17-31, even 2-16, even 18-32), one such incident always affects either 1st+2nd+5th+6th or 3rd+4th+7th+8th ports of a quarter (for example, Te1/17+Te1/19+Te1/25+Te1/27 or Te2/6+Te2/8+Te2/14+Te2/16).
I found a couple of bugs that describe similar behaviour (CSCvg70513 for 6880-X-LE and CSCvn62647 for C68xx Gigatron based modules (which I think is the case here) under heavy traffic and link flap conditions), but their resolutions didn't work for us, on the contrary: we upgraded from 15.2(1)SY* (one or two incident occurences a year) to 15.5(1)SY4 (one occurence every 2-3-4 days). Also, I see that 15.5(1)SY3 (which is supposed to fix CSCvn62647) has a bad review that mentions exactly one of the symptoms we experience (UDLD-caused err-disable).
Has anyone else encountered this behaviour? Any tips for solving (or at least working around) it? Unfortunately TAC is not an option anymore...
Thank you.
