[c-nsp] Parity Problems with 7513...Again!

John Neiberger John.Neiberger at efirstbank.com
Mon Nov 14 12:45:49 EST 2005

Between February and May of this year, we had parity errors on various
line cards in a 7513 that progressively got worse. Cisco TAC ended up
replacing the RSP (twice), a VIP (twice), an ATM PA (twice), and three
DS3 PAs. They eventually decided that the chassis was the problem so it
was replaced, as well. We haven't had a problem since...until several
days ago when a VIP with two DS3 PAs reloaded. That was how the last
problem began so I was obviously worried. TAC said not to worry about it
unless it happened again.

Well, last night the RSP reloaded due to a memory parity error! This is
exactly what happened earlier in the year and I have no explanation for
it. Parity errors are often caused by cosmic rays but it certainly
appeared to be a hardware problem given the frequency of the problems.
However, it is now six months after the last event and I'm wondering if
we've run into a seasonal issue whereby our building is pelted with
cosmic rays twice a year. :-)

Seriously, this is just weird. Why would we be having the exact same
sort of problem with new equipment? It can't be a coincidence. Well, it
could be but I doubt it. Here are the commonalities between the two

* Same power circuits (protected by a Powerware UPS the size of my car.
It should be clean.)

* Same IOS (I've checked Bug Tracker and I see no reason to suspect an
IOS problem.)

* Same location (this chassis is actually in the rack next to the old
one; data center is on the bottom floor of a three story building, so I
would think that would eliminate cosmic rays as the culprit)

* Some line cards (serial, Ethernet, Channel Interface Processors) were
moved from the old chassis to the new chassis. They didn't seem to be
having a problem so TAC did not replace them

That's about it. Any thoughts? I opened a new TAC case this morning and
it has already been escalated by the engineer. Is this just bad karma?
Must I atone for some previous wrongdoing?


More information about the cisco-nsp mailing list