[AG-TECH] Cisco 6500 code?

[Gregory Bell]

Martin,

Thanks for this information.  Have you by any chance disabled multicast
multilayer switching on any interface (ie, "no mls ip multicast")?  

The 6500 problems we've seen involve input buffer queues becoming
"wedged" with packets that can't be cleared.  TAC has explained this as
analogous to a memory leak.  Buffer space eventually diminishes to zero.

We've only noticed this on GigE interfaces that see a sustained, high
rate of multicast traffic.  The symptoms look something like this:

ir2000gw#sho int gig 1/1
GigabitEthernet1/1 is up, line protocol is up 
  Hardware is C6k 1000Mb 802.3, address is 0004.c028.f80a (bia
0004.c028.f80a)
  Description: Gigabit backbone
	<snip>
  Input queue: 41/75/1453/1264 (size/max/drops/flushes); Total output
drops: 0

Instead of returning to zero, the buffer size grows, slowly approaching
the max value.

- Greg


"Huerter, Martin F" wrote:
> 
> Greg,
> 
> We have been running 12.1(12c)E2 for 6 months now and have not encountered
> any problems with multicast. Actually this is the most sound version of
> native IOS we have run on our Cat. 6500's.
> 
> Martin Huerter
> Network Design Engineer
> University of Kansas
> 
> -----Original Message-----
> From: Gregory Bell [mailto:GRBell at lbl.gov]
> Sent: Friday, February 21, 2003 1:35 PM
> To: ag-tech at mcs.anl.gov
> Subject: [AG-TECH] Cisco 6500 code?
> 
> Has anyone found a multicast-bug-free version of IOS for the Cisco 6500
> platform (that is, for the 6500 running native IOS)?
> 
> I'd be interested in trading notes.
> 
> We had serious problems with 12.1(8b)EX3, then switched to 12.1(12c)E4
> after quite a bit of testing, and on Cisco's advice.  But yesterday we
> discovered what appears to be another multicast bug in that code (it was
> interfering with OSPF).
> 
> - Greg Bell, LBNL