[c-nsp] Constant output drops on etherchannel

[Tóth András]

What load-balance method do you use? If you see an uneven distribution
of output drops across the 2 L2 interface, consider changing the
etherchannel load-balance method to have a more even traffic
distribution (src-dst-ip for example).

I understand that you have 5 switches in the stack. You might consider
spreading the etherchannel member interfaces across the stack members,
in a way that the 2 ports should connect to different member switches
(you can have the members on different members as cross-stack
etherchannel is supported). This will ensure that you don't
oversubscribe buffers on one port-asic (buffers are allocated per
port-asic on 3750 switches).

You might try to create a SPAN monitor session with the etherchannel
member interfaces being the source ports to see if you find a burst of
packets in a short interval which could lead to output drops due to
buffer starvation. This will help you identify what's the source of
the bursts. You might also look for unexpected or flooded traffic and
consider pruning unnecessary vlans (if it's a trunk) to prevent
flooding of broadcast and unknown unicast traffic on the etherchannel.

Ultimately, you can increase the bandwidth by adding more links to the
port-channel or using 10G links instead of GigabitEthernet.

Best regards,
Andras


On Fri, Jan 14, 2011 at 10:08 PM, Robert VanOrmer <vanormer at gmail.com> wrote:
> I am experiencing a very similar issue on the C3750E platform w/ 12.2(52)SE.
> We have 5 switches in a cluster, port-channeled with (2) Gig interfaces L2
> to distro (6500's) with excessive output drops even at low utilization.  No
> QoS enabled.  I am using the Cisco TwinGig Converter Modules for uplinks..
> We have a TAC case open, but no progress yet.  I'd be interested in knowing
> if you find a cause. I am going to try swapping GBICs over the weekend.
>
>
>
> Thanks,
>
> Rob
>
>
>
>>Date: Fri, 14 Jan 2011 12:28:03 -0600
>
>>From: Dan Letkeman <danletkeman at gmail.com>
>
>>To: Phil Mayers <p.mayers at imperial.ac.uk>
>
>>Cc: cisco-nsp at puck.nether.net
>
>>Subject: Re: [c-nsp] Constant output drops on etherchannel
>
>>Message-ID:
>
>>             <AANLkTiktxHw2Za3yWjU4rmPaa3q_-EU7N_dKtxUgMaz=@mail.gmail.com>
>
>>Content-Type: text/plain; charset=ISO-8859-1
>
>>
>
>>3560 or 3560G.
>
>>
>
>>(C3560-IPSERVICESK9-M), Version 12.2(53)SE2
>
>>
>
>>Interface config:
>
>>
>
>>interface Port-channel2
>
>>switchport trunk encapsulation dot1q
>
>> switchport trunk native vlan 3009
>
>> switchport trunk allowed vlan 8,10,1008,1101,3009  switchport mode trunk
> end
>
>>
>
>>I see more output drops during higher traffic, but I still see drops during
> low traffic rates.  Always more on one interface.
>
>>
>
>>I do have auto qos enabled for some of the phones I have connected to the
> switches, but I don't have any qos on the etherchannel trunks.
>
>>
>
>>I'm just using the default etherchannel load balancing algorithm.
>
>>
>
>>Thanks,
>
>>Dan.
>
>
>
>
>
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/
>