[c-nsp] C4948 total output drops / bit error problem

[Nemeth Laszlo]

Hello

I have lot of C4948 switches in our network. All switch have 2x1 gb/s 
uplink (etherchannel) to a 6509 switch.

The problem is, if the traffic "big" on the etherchannel ( ~2x700mbit/s) 
the "total output drops" counter increase.

Yesterday the counter was:

   Input queue: 0/2000/0/0 (size/max/drops/flushes); Total output drops: 
131796

Today morning:

   Input queue: 0/2000/0/0 (size/max/drops/flushes); Total output drops: 
133483

I find some sites where use the hold-queue command but it isn't on the 
C4948 platform.

I used a packet generator device, and put 1.500.000 packet (size was 
64byte) across the network. After 5 minutes the packet generator signal 
an error (bit error, lost packet), and the total drops counter increase 
suddenly.
I tried lot of platform like C2970, C4948-10G, C6509/6724sfp, 3Com 
switch, but this loss pattern problem apply only on the C4948 devices 
with different IOS version. But all of C4948 devices show this problem.

More interesting, if i tried this 1.5M packet 64byte test only a C4948 
that no connection to other devices this can produce this problem after 
5 minutes on optical and copper interfaces too. Every time when I did 
this test, the problem apply on the 5. minutes.

Any idea?

Thanks.

Laci