I've seen cases where a VIP2-20 FE -> CX-HSSI w/Frac-T3 at ~80% usage
doing flow switching would report aborts and ignores on the FE when it
tried to switch to the HSSI and it was overloaded at the moment.
Try increasing the outbound buffer in the hssi ... ie
int hssiX/X/X
hold-queue 2048 out
> We have been working with TAC on a 1-3% packet loss problem we are having
> with a C7507, and I wanted to see if anyone had any ideas here, or had run
> into this type of problem. This is an RSP/4 with 160M. We are running
> 11.1(20)CA1. It has two VIP2-40's, each with two PA-H port adapters, (4
> HSSIs total), an FEIP, an EIP-6 just for some handy ethernet ports, and an
> AIP. Its doing bgp and ospf. Thypical processor load is 28%. We were
> seeing lots of ignores on the HSSI and FE interfaces, lots being 1-3%, and
> at times we would lose three or four keepalives in a row, and drop the line
> protocol. We are forwarding about 35,000 pps (adding up all the ins and
> outs and dividing by 2). We have swapped out all the hardware, thinking
> that if the box is having pain at 35,000 pps there must be something
> broken. That did not help. Yesterday we set the mtu to 1500 on the local
> HSSI point-to-points and that helped the hssi ports grandly. However, we
> still see a lot of ignored packets on the FE port.
>
> We are doing flow/distributed switching.
>
> None of the ports is maxing out bandwidth, but they are mostly pretty busy,
> around 50%.
>
> Are we just hitting the performace limit of this box? does anyone have any
> ideas or experience they could share?
>
> Regards,
>
> Dave.
---------------------------------------------------------------------------
** Andrew W. Smith ** awsmith@neosoft.com ** Chief Network Engineer **
** http://www.neosoft.com/neosoft/staff/andrew ** 1-888-NEOSOFT **
** "Opportunities multiply as they are seized" - Sun Tzu **
---------------------------------------------------------------------------
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:13:18 EDT