Peter Mueller wrote:
> Hello Nicolas,
>
Hello Peter,
>
> Maybe I missed something, but how do you know you are at a
> bottleneck?
Well I'm not quite sure there is a bottleneck, although it
looked like on my MRTG graphs monitoring director bandwidth
usage. I saw a flat line when having peak traffic of the day,
around 3000 inbound / 3000 outbound packets according
to ipvsadm -ln --rate output.
> Are you using testlvs (available on Julian's page) or
> apache bench or something similar?
I played with testlvs, but I'm not sure how to interpret
these figures :
client# time ./testlvs VIP:80 -tcp -srcnum 100000 -packets 5000000
Sending 40-byte TCP packets from 10.0.0.1[100000] to VIP:80
real 0m25.612s
user 0m02.460s
sys 0m22.380s
On the other side, I had 2 RS behind director and
show_traffic.sh script reported only(?) around 300 packets/s.
Seems to me that many many packets were dropped...
> BTW, this may be a stupid question but you aren't running
> into strange ethernet errors (carrier, error, etc) as reported
> by ifconfig?
No at all, all error counters are equal to zero. I used to have some
problems because NICs were configured in half-duplex instead of
full-duplex. Since then, everything's fine.
> Also what model eepro100 are you running?
Two Compaq NC3163 Fast Ethernet NICs.
Regards,
Nicolas.
|