[c-nsp] 7200 PA-GE interface resets

[Bruce Pinsky]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Beatty Lane-Davis wrote:
| Hi all,
| I've been having a problem with the PA-GE's in our 7200's for a while
| now where they drop all sorts of packets under light load: input drops,
| output drops, resets, flushes, ignoreds, overruns etc etc etc.  I've had
| a TAC case open for months now trying to figure out what's going on, but
| that's gotten me both SFA and bitter, I must be asking the wrong
| questions.
|
| We found that by increasing the input and output hold queue's to the
| maximum 4096(thanks Jared :-) _most_ of the symptoms went away, however,
| we continued to see a constant trickle of ignoreds and resets on the two
| boxes acting as our core routers.  TAC has come back with a possible
| explanation for the ignoreds, which I'll get to trying at some point...
|
| My main point of concern at the moment has to do with the interface
| resets.  I've been advised by a TAC engineer that there is a known issue
| with the PA-GE's where it resets periodically so I shouldn't be
| concerned at all, that's just how it is.  Right.  Does anyone else have
| this problem?  He's identified two potential bugs that he's saying are
| causing the resets, CSCdt37135 which don't look like it to me and
| CSCdm84052 which I can't view in bugtracker.  None of this has sounded
| very right to me but with everything else we've got on it hasn't been
| enough of a problem to get back to until yesterday afternoon when the
| interface bounced on both the router and switch and a burst of resets
| was incremented.  Yikes!
|
| I'm starting to wonder if there is some problem on the config between
| the routers and switches?  I understand from the colleague that built
| this infrastructure that there was some tweaking that needed to be done
| with regards to getting flow control et al right.  Perhaps the
| combination of parameters they hit upon are causing this problem now?
| While these configs have been looked over more than a few times by
| Cisco, I'm wondering if we're missing something basic and simple?
| Configs and show's below.  Is it possible that because the router is
| having to ignore some of these packets that it's freaking out and
| calling for an interface reset?
|
| One key thing to keep in mind is that these routers should be really
| bored, they're NPE-400's with no more than 60-70M mostly switched in and
| out the same GE interface.  Cricket shows that the CPU never runs hotter
| than 25%.
|
| Any advice or input would be very much appreciated.
|


Hmm, both of those DDTS's you reference are about stuck interface hold
queues, not interfaces that spontaneously reset.   Got a version number?

The interface resets would occur for failure to transmit after a number of
retries.  Any number of reasons for that, including the media, bug, lack of
resources, etc.  Any messages in the log prior to the interface resetting?

The "mostly switched in and out the same GE interface" would grab my
attention WRT the ignores.  I'm wondering if that interface is starved for
buffers given that inbound and outbound packets are fighting for the same
resources and queue allocations rather than the load being shared across a
separate inbound and outbound interface.

- --
=========
bep

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (MingW32)

iD8DBQFBdbLrE1XcgMgrtyYRAvv5AKD1O6OeRQ/l5Hr+OKdbsIZTU2Jq9gCgwuPo
8W4JvmwL5djraP8QOL9zKh0=
=BIbx
-----END PGP SIGNATURE-----