[c-nsp] ASR A9K-8T-L certain ports limited to 8 Gbps.

[Adam Vitkovsky]

Hi 

I'd try the "sh controllers NP" to figure out what's wrong, on A9K-8T-L each
port has its own NP. 

So "sh controllers NP ports all" will show you which NP belongs to which
port. 

Anyways TAC engineer will provide you with some fancy show commands to
determine where the bottleneck is. 

 

adam

From: bas [mailto:kilobit at gmail.com] 
Sent: Wednesday, November 20, 2013 11:13 AM
To: Adam Vitkovsky
Cc: mark.tinka at seacom.mu; Cisco
Subject: Re: [c-nsp] ASR A9K-8T-L certain ports limited to 8 Gbps.

 

Hi,

 

On Wed, Nov 20, 2013 at 10:31 AM, Adam Vitkovsky <adam.vitkovsky at swan.sk>
wrote:

On Wednesday, November 20, 2013 01:15:36 AM McDonald Richards wrote:

> What framing mode are you running and what is the underlying
> transmission?
>
> I have seen this before on 10G circuits running in wanphy mode and the
> only fix was to get better transmission (ie. not an STM-64c) and run
> lanphy :)

 

 

Hi,

 

Thanks, but it isn't WAN-PHY

 

---------

RP/0/RSP1/CPU0:ams-5345#show controllers wanphy 0/0/0/2 all

Wed Nov 20 10:58:43.952 CET

Interface: wanphy0/0/0/2

Configuration Mode: LAN Mode

Operational Mode: LAN Mode

---------

 

I'm starting to think one of the A9K-8T-L cards is messing about on the
fabric signalling, or some funky QoS bug somewhere. 

 

Interesting that only couple of the 3 upstream ports have this issue

Are these ports in a bundle?

 

No, none are in a bundle.

There are cases when an ECMP egress path is fine on one, and limited to 8 on
another interface.

 

Or maybe some pushback is going on that's not related to the box

 

I don't think so, paths are ebgp upstreams, but also ibgp peering boxes.

One interface is afftected and another is not to the same egress path.

 

Thanks if anyone has other clues where to look.

 

Bas