[c-nsp] Input Errors

[Rodney Dunn]

Very unlikely the ignores are causing latency that would be
noticed for transit packets going through a 72xx due to it's architecture.

It will introduce a very small amount because it signals the rx ring
filled up with packets coming in. But given the transit packet forwarding
speed to drain that rx ring it would be very very minimal.

These ignores ususally come from microburst on the wire where the upstream
sent a gige linerate type burst that the 72xx couldn't pull off the wire
fast enough. It's the most common issue seen on 72xx's (or any sw forwarding
gige box) that connects to a device that can do linerate gige on transmit.

There have been some other improvements to prevent things like 'sh ver'
from blocking interrupts for a very small amount of time so make sure
you are on something pretty recents such ast 12.4(20) or later mainline
or 12.4(20)T or later.

Rodney

 
On Fri, Mar 06, 2009 at 10:12:09AM -0800, Shaun R. wrote:
> I seam to be having a few customer complain about latency and I'm not sure 
> what's going on.  There traceroutes always show the latency on my layer2 
> side of the connect from my upstream.  It always seams to be the same 
> upstream but my other two don't really move too much traffic.  One thing I 
> did notice is that I'm seeing input errors on some interested.  I'm seeing 
> input errors on the interface that connects to the upstream, as well as 
> input errors on interfaces that are connecting my borders to my core/access 
> switch.  I try to trace from my gear out to the customers ip and the trace 
> goes out the same upstream but latency looks normal and is half what they 
> see when coming in.  I'm been trying to figure out what might be going on 
> but it's proving difficult.
> 
> So what's causing these input errors?  I have them on every interface on 
> both my borders.  Whats weird is that on my core/access switches that 
> connect to the borders i dont see any errors on those interfaces.
> 
> My network design is simple.
> 
> * two border routers, 7206VXR-NPE-G2
> * 3750 stack as core/access
> * gigE connection between borders and core/access
> * gigE connection between borders
> * gigE connection to all 3 upstreams.
> * full BGP between borders
> * OSPF between borders and core/access
> * full BGP to upstreams
> * Primary upstream pushes about 200mbit
> * secondary upstream pushes around 40mbit
> * third upstream pushes around 10mbit.
> 
> Any help would be appreciated, thanks in advance!
> 
> border1#sh int | inc input errors
>     16 input errors, 0 CRC, 0 frame, 0 overrun, 16 ignored  [ GigE 0/1 ]
>     0 input errors, 0 CRC, 0 frame, 0 overrun, 0 ignored [ FE/02 Shutdown]
>     6 input errors, 0 CRC, 0 frame, 0 overrun, 6 ignored [GigE 0/2]
>     6464 input errors, 0 CRC, 0 frame, 0 overrun, 6464 ignored [ GigE 0/3 ]
>     0 input errors, 0 CRC, 0 frame, 0 overrun, 0 ignored [FE1/0 Shutdown]
>     0 input errors, 0 CRC, 0 frame, 0 overrun, 0 ignored, 0 abort 
> [Loopback0]
> 
> border2#sh int | inc input errors
>     6712 input errors, 0 CRC, 0 frame, 0 overrun, 6712 ignored [ GigE 0/1 ]
>     0 input errors, 0 CRC, 0 frame, 0 overrun, 0 ignored [ FE 0/2 
> Shutdown ]
>     23 input errors, 0 CRC, 0 frame, 0 overrun, 23 ignored [ GigE 0/2 ]
>     14805 input errors, 0 CRC, 0 frame, 0 overrun, 14805 ignored [ GigE 
> 0/3 ]
>     0 input errors, 0 CRC, 0 frame, 0 overrun, 0 ignored [ FE 1/0 
> Shutdown ]
>     257 input errors, 0 CRC, 0 frame, 66 overrun, 191 ignored [ GigE 2/0 ]
>     0 input errors, 0 CRC, 0 frame, 0 overrun, 0 ignored, 0 abort 
> [Loopback0]
> 
> 
> ~Shaun
> 
> 
> _______________________________________________
> 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/