[c-nsp] cat6k / error analysis

[Gert Doering]

Hi,

I'm a bit puzzled.

Catalyst 6509, Sup1A-2GE, WS-X6516A ("brand new"), SX GBIC, connected 
to 7301, SX SFP, about 1m cable.


Upgraded the 6509 today (from cat OS 6.5 to 7.6(14), to get support for
the 6516A), and since then, I regularily see

2005 Sep 28 07:57:59 UTC +00:00 %SYS-3-PORT_IN_ERRORS:Port 5/1 detected 10796 input packet(s) contained errors in last 30 minutes
2005 Sep 28 07:57:59 UTC +00:00 %SYS-3-PORT_RX_BADCODE:Port 5/1 detected 146163 bad code errors in last 30 minutes
2005 Sep 28 07:57:59 UTC +00:00 %SYS-3-PORT_RUNTS:Port 5/1 detected 2905 runt(s) in last 30 minutes

"show port 5/1" displays this:

Port  Align-Err  FCS-Err    Xmit-Err   Rcv-Err    UnderSize
----- ---------- ---------- ---------- ---------- ---------
 5/1           0       2013          0          0         0

Port  Single-Col Multi-Coll Late-Coll  Excess-Col Carri-Sen Runts     Giants
----- ---------- ---------- ---------- ---------- --------- --------- ---------
 5/1           0          0          0          0         0     12910         -

Port  Last-Time-Cleared
----- --------------------------
 5/1  Wed Sep 28 2005, 03:31:01

- far less errors than what the syslog message ("10796 input packet(s)
containing errors") claims, but a fair number of Runts.

The port is running as dot1q trunk, with quite some traffic on the native
VLAN (about 50 Mbit) and more traffic on individual VLANs (about 200 Mbit
in total).  So compared to the total number of packets, the error rate
is low - but I still don't like it, "0" would be an acceptable target.

Any suggestions where to start looking for these?  GBIC/SFPs + Cabling?
(I'd rather not "just try it", as testing these means "customers offline
while PVSTP reconverges").


The 7301 is running 12.2(18)S9, and its counters are pretty clean:

  Last clearing of "show interface" counters 3w4d
...
     2467997774 packets input, 1276539188 bytes, 0 no buffer
     Received 0 broadcasts (13597195 IP multicast)
     2 runts, 0 giants, 7 throttles
     0 input errors, 3 CRC, 0 frame, 954833 overrun, 0 ignored
     0 watchdog, 35003768 multicast, 0 pause input
     2835420544 packets output, 650859354 bytes, 0 underruns
     0 output errors, 0 collisions, 0 interface resets
     0 babbles, 0 late collision, 0 deferred
     3 lost carrier, 0 no carrier, 0 PAUSE output
     0 output buffer failures, 0 output buffers swapped out


I understand that a runt is an ethernet packet that's too short, but
otherwise well-formed.  How can this happen?  Bugs?  Different 
interpretation of standards?

gert

-- 
USENET is *not* the non-clickable part of WWW!
                                                           //www.muc.de/~gert/
Gert Doering - Munich, Germany                             gert at greenie.muc.de
fax: +49-89-35655025                        gert at net.informatik.tu-muenchen.de