I think Cassidy's solution will fix your issue. By default, queuing on a 3600 only allocates a percentage of port speed to class 0 traffic (untagged). His commands will grant 100% of port speed to class 0 traffic.

Jesse DuPont

Network Architect
email: [email protected]
Celerity Networks LLC

Celerity Broadband LLC
Like us! facebook.com/celeritynetworksllc

Like us! facebook.com/celeritybroadband

On 4/14/16 7:50 PM, Craig Baird wrote:
I'll give it a try.  Thanks Cassidy.

Craig


Quoting "Cassidy B. Larson" <[email protected]>:

You might want to increase the queue limit on your 1G interfaces.

We do this on our ASR920’s:

class-map match-any cos_all
 match cos  0  1  2  3  4  5  6  7
!
policy-map MC_1G_1M
 class cos_all
  bandwidth percent 100
  queue-limit 1024000 bytes
!
int g0/0/0
 service-policy output MC_1G_1M

Haven’t used M3600X’s before, but it depends on what the buffer size of the total ports are, and probably how many ports are in use as well..

-c


On Apr 14, 2016, at 2:28 PM, Craig Baird <[email protected]> wrote:

I've got a Cisco ME3600X switch connected up to a pair of Cambium PTP820S radios via gigabit fiber SFPs.  I'm seeing significant outdiscards, resulting in lost frames on the switch on both ports going to the 820S radios.  The links are not congested--generally pushing less than 100 Mbps (80 MHz channels, capable of 650 Mbps).  I've got other 820S radios connected to the same model switch in other places, and am not seeing this.  The only difference is that this switch is fed with 10 gig fiber.  I'm wondering if it's a queueing problem of some sort.  Anybody have any ideas?

Craig








Reply via email to