[c-nsp] CRC errors on multiple ports of WS-C2950T-48-SI switch

[Martin T]

Hi,

I had two WS-C2950T-48-SI switches in the network which all of the
sudden started to show large amount of CRC errors on about dozen
ports. Those two incidents are not related to each other, i.e those
two incidents did not happen at the same time or in the same location.
I simply received those switches now from remote sites. Unfortunately
this is all the information I have about the incidents. My first
thought was that maybe it has to do with the fact that switch has 48
ports and there was increased bandwidth between the ports behind
different ASICs. I made a following physical setup:

Fa0/1 <-> Fa0/48
Fa0/2 <-> Fa0/47
Fa0/3 <-> Fa0/46
...
Fa0/24 <-> Fa0/25

Image of the setup can be seen here:
http://s15.postimg.org/fmsj753qz/WS_C2950_T_48_SI_test_setup.png


VLANs were configured like this:

WS-C2950T-48-SI#sh int status

Port      Name               Status       Vlan       Duplex  Speed Type
Fa0/1                        connected    1          a-full  a-100 10/100BaseTX
Fa0/2                        connected    2          a-full  a-100 10/100BaseTX
Fa0/3                        connected    3          a-full  a-100 10/100BaseTX
Fa0/4                        connected    4          a-full  a-100 10/100BaseTX
Fa0/5                        connected    5          a-full  a-100 10/100BaseTX
Fa0/6                        connected    6          a-full  a-100 10/100BaseTX
Fa0/7                        connected    7          a-full  a-100 10/100BaseTX
Fa0/8                        connected    8          a-full  a-100 10/100BaseTX
Fa0/9                        connected    9          a-full  a-100 10/100BaseTX
Fa0/10                       connected    10         a-full  a-100 10/100BaseTX
Fa0/11                       connected    11         a-full  a-100 10/100BaseTX
Fa0/12                       connected    12         a-full  a-100 10/100BaseTX
Fa0/13                       connected    13         a-full  a-100 10/100BaseTX
Fa0/14                       connected    14         a-full  a-100 10/100BaseTX
Fa0/15                       connected    15         a-full  a-100 10/100BaseTX
Fa0/16                       connected    16         a-full  a-100 10/100BaseTX
Fa0/17                       connected    17         a-full  a-100 10/100BaseTX
Fa0/18                       connected    18         a-full  a-100 10/100BaseTX
Fa0/19                       connected    19         a-full  a-100 10/100BaseTX
Fa0/20                       connected    20         a-full  a-100 10/100BaseTX
Fa0/21                       connected    21         a-full  a-100 10/100BaseTX
Fa0/22                       connected    22         a-full  a-100 10/100BaseTX
Fa0/23                       connected    23         a-full  a-100 10/100BaseTX
Fa0/24                       connected    24         a-full  a-100 10/100BaseTX
Fa0/25                       connected    25         a-full  a-100 10/100BaseTX
Fa0/26                       connected    24         a-full  a-100 10/100BaseTX
Fa0/27                       connected    23         a-full  a-100 10/100BaseTX
Fa0/28                       connected    22         a-full  a-100 10/100BaseTX
Fa0/29                       connected    21         a-full  a-100 10/100BaseTX
Fa0/30                       connected    20         a-full  a-100 10/100BaseTX
Fa0/31                       connected    19         a-full  a-100 10/100BaseTX
Fa0/32                       connected    18         a-full  a-100 10/100BaseTX
Fa0/33                       connected    17         a-full  a-100 10/100BaseTX
Fa0/34                       connected    16         a-full  a-100 10/100BaseTX
Fa0/35                       connected    15         a-full  a-100 10/100BaseTX
Fa0/36                       connected    14         a-full  a-100 10/100BaseTX
Fa0/37                       connected    13         a-full  a-100 10/100BaseTX
Fa0/38                       connected    12         a-full  a-100 10/100BaseTX
Fa0/39                       connected    11         a-full  a-100 10/100BaseTX
Fa0/40                       connected    10         a-full  a-100 10/100BaseTX
Fa0/41                       connected    9          a-full  a-100 10/100BaseTX
Fa0/42                       connected    8          a-full  a-100 10/100BaseTX
Fa0/43                       connected    7          a-full  a-100 10/100BaseTX
Fa0/44                       connected    6          a-full  a-100 10/100BaseTX
Fa0/45                       connected    5          a-full  a-100 10/100BaseTX
Fa0/46                       connected    4          a-full  a-100 10/100BaseTX
Fa0/47                       connected    3          a-full  a-100 10/100BaseTX
Fa0/48                       connected    2          a-full  a-100 10/100BaseTX
Gi0/1                        connected    1          a-full a-1000
10/100/1000BaseTX
Gi0/2                        connected    25         a-full a-1000
10/100/1000BaseTX
WS-C2950T-48-SI#


While I don't know the internal architecture of WS-C2950T-48-SI switch
and it has no "sh platform" command which might reveal anything, then
at least based on traces on the PCB first half of the ports seem to be
connected to first Broadcom BCM5649B0KPB ASIC and second half of the
ports seem to be connected to second Broadcom BCM5649B0KPB ASIC. Now
if I send 100Mbps of traffic through all the switch ports, then there
are frame drops because of ingress buffer overflow, but no CRC errors.
In addition, if I decrease bandwidth to 70Mbps, then there are no more
ingress buffer drops. Last but not least, those two switches served
fairly low-speed connections and most of the traffic was between
FastEthernet and GigE uplink-ports.

What might cause vast amount of transient CRC errors on multiple
switch ports? Maybe the PCB of WS-C2950T-48-SI switch is known to be
especially receptive to EMI? Any other ideas?



thanks,
Martin