[c-nsp] 3550 IO memory fragmentation

[Jon Lewis]

We had a recent network event during which all of our 3550 access layer 
switches started logging things like:

%SYS-2-MALLOCFAIL: Memory allocation of 1680 bytes failed from 0x17FC48, 
alignment 0
Pool: I/O  Free: 5936  Cause: Memory fragmentation
Alternate Pool: None  Free: 0  Cause: No Alternate pool

-Process= "Pool Manager", ipl= 0, pid= 5
-Traceback= 1C919C 1CA760 17FC4C 1DB43C 1DB658 1EBACC 1EF0FC

A few minutes later, all was back to normal, though one 3550 did a 
software forced crash / reload.

Under normal circumstances, these switches have 4.5-5mb of free IO memory.

This looks very similar to what was posted several years ago at

http://www.velocityreviews.com/forums/t31947-catalysts-trouble.html

Searching bug toolkit, I didn't find anything that looked relevant.  Has 
anyone else run into this sort of thing with 12.1EA software or have an 
idea what the cause/solutions might be?

----------------------------------------------------------------------
  Jon Lewis                   |  I route
  Senior Network Engineer     |  therefore you are
  Atlantic Net                |
_________ http://www.lewis.org/~jlewis/pgp for PGP public key_________