[c-nsp] Problems with bandwidth on interface tunnel

[Church, Charles]

I'm sure there could be many reasons, such as a real brief burst of traffic, maybe a frame that needed fragmentation, but DF was set (just guessing on that one).  I'm sure a cisco.com person would know better than me though... 

--- Original Message ---
From:"Adrian Chadd" <adrian at creative.net.au>
Sent:Sat 3/3/07  8:51 pm
To:"Church, Charles" <cchurch at multimax.com>
Cc:"cisco-nsp at puck.nether.net" <cisco-nsp at puck.nether.net>
Subj:Re: [c-nsp] Problems with  bandwidth on interface tunnel

On Sat, Mar 03, 2007, Church, Charles wrote:

> That 'transmit bandwidth' command can set a cap on how much data can be sent into a tunnel.  It defaults to 8 mbit, but that probably is close to what the router can do in software.  When you were getting the 8 mbit transfer rate, were the router CPUs running close to 100%?  You've got drops on the tunnel interfaces, but I'm guessing not on the physical interfaces.  The output drops on the 
tunnel interface are probably a result of running out of CPU on the router.  At least it probably is when the transmit bandwidth was set to 40 mbit.  I'd try just plain IPSec, see if that improves things.

Is that the only thing that causes output drops on tunnel interfaces?
I've noticed the occasional tunnel drop between a 2821 and 2801 w/ a
GRE tunnel. They're directly connected on the test bench and pushing
~500kbit between the 2821 and each 2801. The CPUs on the ISRs aren't
anywhere near max.

I tried looking for documentation on tunnel drops but I couldn't
find anything conclusive.

Thanks,


Adrian