rwebb at ropeguru.com rwebb at ropeguru.com
Thu Mar 27 17:04:28 UTC 2014


It is actually a 4001i for an IBM Blade Chassis. Sorry for that.

So in this setup, port a would be a trunk with multiple vlans 
connection back to a 6509. port b would be a switch port in access 
mode that connects to an IBM blade in the chassis.

Not sure that this situation fits either of those scenarios.

Overall problem is that we are seeing performance issues between 
servers. These servers are all AIX based. We believe/know that we have 
some misconfigurations in the environment with jumbo frames and flow 
control. My curiosity about the discards is due to those 
misconfigurations. The port I mentioned in my original email has 
around 480 million output packes to the 1.1 million discards.

We do have IBM and Cisco support engaged, I am just trying to make 
sure I understand enough to be dangerous when I am working with them.

Robert

On Thu, 27 Mar 2014 12:55:46 -0400
  Lee <ler762 at gmail.com> wrote:
> On 3/27/14, rwebb at ropeguru.com <rwebb at ropeguru.com> wrote:
>> So I certainly admit I am a basic networking guy and in the past 
>>have not
>> had to get into the nitty gritty of port statistics.
>>
>> I am trying to understand some statistics off a switch port in a 
>>Nexus
>> 4001i.
> 
> Good luck.  I couldn't find anything for a nexus 4000, but did find
> this for IOS:
> In-Discard - The result of inbound valid frames that were discarded
> because the frame did not need to be switched. This can be normal if 
>a
> hub is connected to a port and two devices on that hub exchange 
>data.
> The switch port still sees the data but does not have to switch it
> (since the CAM table shows the MAC address of both devices 
>associated
> with the same port), and so it is discarded. This counter can also
> increment on a port configured as a trunk if that trunk blocks for
> some VLANs, or on a port that is the only member of a VLAN.
> 
> so if you've got something like
> switch a: switchport trunk allowed vlan 1-5
> switch b: switchport trunk allowed vlan 1-4
> 
> when switch a sends a frame on vlan 5, switch b counts it as an 
>input discard.
> 
> Lee
> 
>>
>> All TX and RX counters look normal except on the TX side, I am
>> showing 1107597 input discards. Last clearing of show counters is 
>>1d8h ago.
>>
>> I have it in my mind that this particular counter is dropping 
>>packets coming
>> in from another port inside the switch that are to be transmitted 
>>out to the
>> end server.
>>
>> So lets say the interface I am looking at is port 2 on the switch. 
>>So server
>> 1 sends a packet to port 1 on the switch. That packet then traverses 
>>to
>> backplane, or inside the same ASIC, to port 2 on the switch. It is 
>>then
>> dropped and not transmitted out to server 2.
>>
>> Is the scenario I just presented correct? Not looking for the reason 
>>in this
>> email, just that my logical understanding is correct.
>>
>> Robert
>>
>>
>> Sent from my Verizon Wireless 4G LTE smartphone




More information about the NANOG mailing list