On Thu, 2008-10-16 at 10:24 -0700, Sashi Kant wrote:
> We have bonding setup in all the servers and it works fine.
Well... if it's working fine, what's the problem here then? (Yes,
tongue-in-cheek I know).
> I agree with your observation about packet per second problem , rather
> than MB/s issue. I was able to push > 800 Mb/s of traffic to similar
> setup (bonding-rr + Link aggregation )using a single machine as traffic
> We start seeing dropped packet when traffic goes above 40k pps (tx+rx).
Using bnx2... the only things I can see for that driver are some
complaints that rx_filtered_packets goes up for high packet rates, so it
could be the driver itself at fault.
Please remove bonding and try again using a single NIC. You'll know then
if the driver, the hardware, the switch or your network cable is at
fault. Or maybe it'll be a phase of the moon instead!
If you can simplify the system, then add complexity on steps you might
find the trigger. I'll surmise now that it's simply the system cannot
process 40kpps with two NICs on the same bus.