[c-nsp] 3850 and output drops

Sebastian Beutel sebastian.beutel at rus.uni-stuttgart.de
Tue Sep 6 12:48:01 EDT 2016


Hi List,

    this maybe sounds like a novice question but the phenomenon i observe
here is so peculiar to me that i like to share it and ask whether someone
else has made similar observations.
    Everything started when a co-worker tried to troubleshoot a rapidly
increasing output error counter on a 3850 Switchport. Neither replacing the
switchport nor the patch cord or even the nic on the other side helped. For
no precise reason i tried to compare the numbers with our older 3750
switches and it turned out, that around 5% of our 3850 switchports had
"output errors" while not a single one our 3750 switchports had any. Even
though we have five times more switchports on 3750 and these stacks have
typically much higher uptimes (because they are operational for many years
now). But they had output drops. So my suspicion is as follows:

The output errors counters count different things on a 3750 and a 3850. 
While the 3750 only counts "frames dropped due to physical media problems"
the 3850 also adds "frames dropped due to filled buffer" to this number.

As the numbers of the "Total output drops" and "output errors" counters on a
3850 are usually identical and if i assume that our cabling is of the same
quality as it is on 3750 (where we have no "output errors" at all) this
makes perfec sense. 

Any comments on that? Is that a stupid consideration?

Best,
   Sebastian.


More information about the cisco-nsp mailing list