Thx for the quick reply.

Switch among the NICs and see if you have any values for that host other
than on the netFlow device?

well, yes ( but only for that one device I chose to illustrate the problem
because the switch is managed by a server local to the LAN that the ntop
host is on.)  I understood that -M forced packet aggregation off. I also
only have managment traffic on my NZ based subnet, not any customer
client/server traffic paths, which is the bulk of the data in the displays.
Please note that this doubling holds true for all displays ( i.e.) for R-R
flows, the summary is correct, the host detail is doubled, and the time
stamped ( hourly) buckets are correct

example here. This host flow ( verified by last contact) is for a flow that
is local to my clients network only, not in any way routed down the
management VPN tunnel.

Remote to Remote IP traffic
192.168.114.58 336 0.0 % 367 0.0 %  - this is correct

Host detail
Host Location Remote (outside specified/local subnet)
Total Data Sent 672/16 Pkts/0 Retran. Pkts [0%]  - this is double
Broadcast Pkts Sent 0 Pkts
Data Sent Stats Local 50.0 %
 Rem 50.0 %

IP vs. Non-IP Sent IP 50.0 %
 Non-IP 50.0 %

Total Data Rcvd 734/18 Pkts/0 Retran. Pkts [0%]  - this is double
Data Rcvd Stats Local 50.0 %
 Rem 50.0 %

IP vs. Non-IP Rcvd IP 50.0 %
 Non-IP 50.0 %

Sent vs. Rcvd Pkts Sent 47.1 %
 Rcvd 52.9 %

Sent vs. Rcvd Data Sent 47.8 %
 Rcvd 52.2 %

Host Traffic stats
Midnight - 1AM 0 0.0 % 0 0.0 %
1AM - 2AM 336 100.0 % 367 100.0 %  - this hourly bucket total is correct.
2AM - 3AM 0 0.0 % 0 0.0 %



What do you see if you use tcpdump to capture the packets in a combined
format?
tcpdump -c mmm  "(udp and dst port 2055) or (host 192.168.42.xx)"
That is any netFlow record (you could add host dst yourip to limit it) or
anything to/from the USA host...

I only ever see netflow records, as the customer flows are not local to my
LAN
02:06:17.275615 router.clientdomain.com.50228 >
myntophost.mydomain.com.2055:  udp 1464 (frag 52065:[EMAIL PROTECTED])
etc.
I agree I would see ethernet packets for the illustrative example, but I
only chose that for debugging purposes (ie) I needed  a flow that was
regular and known, so I could verify each record at source and be certain of
the data coming in. But as before, this doubling is for all customer flows,
not just those that may also happen to be present on my LAN.


Question - since you have a VPN connection, what makes you think the ONLY
traffic over that tunnel is the netFlow records???

Because it is routed that way. I have management data and netflow records on
that connection. But -M should be in force, so we don't count the local
traffic, and R-R flows, or any flow that is not a management flow, will
never come my way to be on my LAN. However, the point of all this, is that
the remote router counts it all, wraps it up in a netflow packet, and ntop
should display what it gets. Ntop is very very good at what is does, if I
can just nail the anomaly I am seeing on the host detail stats.

Thanks

Eric

_______________________________________________
Ntop mailing list
[EMAIL PROTECTED]
http://listgateway.unipi.it/mailman/listinfo/ntop

Reply via email to