Hi Paolo,
thanks for your reply.
> * are you getting the funky numbers in the web interface? If that's the
> case, can you please check also the underlying database to see whether
> the problem is generated by the application topped-over pmacct?
mysql> SELECT ip_src,ip_dst, sum(bytes) FROM acct GROUP BY ip_dst;
+-------------+-------------+-------------+
| ip_src | ip_dst | sum(bytes) |
+-------------+-------------+-------------+
| X.X.X.X.162 | 0.0.0.0 | 496993114 |
| 0.0.0.0 | X.X.X.X.160 | 248784 |
| 0.0.0.0 | X.X.X.X.161 | 18178595 |
| 0.0.0.0 | X.X.X.X.162 | 18501584877 |
| 0.0.0.0 | X.X.X.X.163 | 215319081 |
| 0.0.0.0 | X.X.X.X.164 | 7807499 |
| 0.0.0.0 | X.X.X.X.165 | 82180 |
| 0.0.0.0 | X.X.X.X.166 | 82180 |
| 0.0.0.0 | X.X.X.X.167 | 82564 |
| 0.0.0.0 | X.X.X.X.168 | 82584 |
| 0.0.0.0 | X.X.X.X.169 | 82132 |
| 0.0.0.0 | X.X.X.X.170 | 82256 |
| 0.0.0.0 | X.X.X.X.171 | 82980 |
| 0.0.0.0 | X.X.X.X.172 | 82288 |
| 0.0.0.0 | X.X.X.X.173 | 82192 |
| 0.0.0.0 | X.X.X.X.174 | 109908 |
| 0.0.0.0 | X.X.X.X.175 | 115696 |
+-------------+-------------+-------------+
17 rows in set (0.00 sec)
The data is similar to the data shown in the web interface.
> * is eth1 the interface where the PPP session to the ISP is established?
> If that's the case, just for testing purposes, would you give a try
> binding pmacct to eth2 instead?
PPP was only involved on a test maschine where everything went smooth.
The above data is from the gateway maschine I want to account traffic
on.
Yes, eth1 is the interface to my ISP. It is connected to some Cisco or
Juniper port. The interface eth1 is in a seperate /30 subnet, which is
used to route the /28 subnet, which I need to do accounting for.
I configured pmacct to listen on eth2 now and the data shown is much
more realistic. I will compare the data from eth2 to the data my ISP
provides to make sure it is correct. However, having pmacct listen on
eth2 instead, I will loose some traffic (e.g. traffic the gateway
blocks) and some traffic will add, even though the traffic inside the
/28 subnet is free of charge.
> * if nothing of the above helps. Would you make each plugin writing to
> a separate table (ie. acct_in, acct_out) and check whether results
> get ok?
Sure, I can try that once I gathered enough data on eth2 to compare it
to the data my ISP provides.
> * I'm assuming bridging is not involved on the "Gateway" box at all -
> it should make no sense unless there is some kind of a router (not
> in your graph) connecting to the ISP and holding the PPP session. Am
> I correct?
This is a more complete graph:
+------- Host A
+---------+ |
ISP (Cisco/Juniper) --- STM-1 --- eth1 | Gateway | eth2 ---+------- Host B
+---------+ |
+------- Host C
No, bridging is not involved on the gateway.
Thanks.
--
Daniel Bramkamp
_______________________________________________
pmacct-discussion mailing list
http://www.pmacct.net/#mailinglists
