[c-nsp] Best practice - Core vs Access Router

[David Freedman]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Andy B. wrote:
> On Wed, Feb 10, 2010 at 7:48 PM, David Freedman
> <david.freedman at uk.clara.net> wrote:
>> So, are you checking your interfaces for incrementing drop/error counters?
>>
>> Are you seeing any of this when there is the problem occuring?
>> (clear counters , sh int summ etc..)
>>
> 
> I am having input drops all the time, no matter how high or low I set
> the incoming hold-queue.
> 
> The OSPF and IBGP interfaces approx. 30 minutes after I cleared the counters:
> 
> TenGigabitEthernet8/1 is up, line protocol is up (connected)
>   Input queue: 0/2000/622/622 (size/max/drops/flushes); Total output drops: 0
> 
> TenGigabitEthernet9/1 is up, line protocol is up (connected)
>   Input queue: 0/4096/1664/1664 (size/max/drops/flushes); Total output drops: 0
> 
> TenGigabitEthernet9/2 is up, line protocol is up (connected)
>   Input queue: 0/4096/1916/1916 (size/max/drops/flushes); Total output drops: 0
> 
> 
> These links are not congested! Te9/1 is the busiest with maybe 6.5 out
> of 10 Gig. The other two are below 5 Gig.

Are these supervisor ports or on a card (i.e 6704/6708?)

Things I would check:

- - That I understand 6704 has pathetically small per port buffers
- - Hold queue input appropriate (for punt to MSFC), usually set to 4096
for these
- - No IGP hello padding (if you have large MTU and pad then you must punt
these big things
- - Check SPD headroom (show ip spd)
- - The drops are not being reported on input due to lack of transmit
buffer on output (i.e to lower speed card), check traffic flows/pps to
low speed interfaces and adjust buffers appropriately

Dave.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAktzBVEACgkQtFWeqpgEZrIB+QCeKT5sqezBtRp5DWXD71VwH6Ke
tJUAnRyC67nIKx1NpYBB+g+854TtBUq3
=g6FU
-----END PGP SIGNATURE-----