W dniu 20.02.2015 11:57, Alfredo Cardigliano napisał(a):
This is strange, could you provide us more info? In the second case
you see correct stats for both directions or just one direction?
For testing purposes I am pushing one direction packet flood like this:
flood -> recv on zc:eth2 (zbounce) zc:eth3 xmit -> flood traffic copy
There is also a minor problem that statistics are one-direction only.
How to reproduce:
// start pushing packets from network to eth2
$ sudo ./zbounce -i zc:eth2 -o zc:eth3 -c 2 -b
// reports 0 traffic
=========================
Absolute Stats: 0 pkts - 0 bytes
Actual Stats: 0.00 pps - 0.00 Gbps
=========================
Now, if I send incoming ping-packets to eth3, it will report only them:
=========================
Absolute Stats: 2 pkts - 244 bytes
Actual Stats: 1.00 pps - 0.00 Gbps <---- in reality it pushes 11.5Mpps
received on eth2 on 1 pps received on eth3
=========================
$ sudo ./zbounce -i zc:eth3 -o zc:eth2 -c 2 -b
// reports correct level of traffic
=========================
Absolute Stats: 46'031'487 pkts - 3'866'644'908 bytes
Actual Stats: 11'575'432.89 pps - 7.78 Gbps
=========================
So it seems traffic gets accounted only if incoming traffic comes to
interface specified as outgoing in zbounce.
(Actually -i and -o interfaces become both incoming and outgoing, when
-b is used).
From the host perspective counters are zero (procfs, netlink) but I
assume this is a "feature" of external ZC packet processing.
From ethtool point of view traffic can be counted and it looks OK
(~11,5Mpps as rx_packets on eth2 and tx_packets on eth3).
HTH
--
Paweł
_______________________________________________
Ntop-misc mailing list
[email protected]
http://listgateway.unipi.it/mailman/listinfo/ntop-misc
