Get a faster NIC/Processor.

You're going to have to do some research to figure out what that counter
means on your OS.  All ntop is doing is echoing the value from libpcap:

myGlobals.device[myGlobals.actualReportDeviceId].droppedPkts.value =
pcapStats.ps_drop;

How libpcap sets that value depends on the OS.  On some it's the ifconfig
value, on others it's other things...

So it might be ntop not accepting packets fast enough, or it might be a true
drop in the kernel before ntop ever sees them - typically, the kernel drops
packets when it can't keep up with the interrupt rate (either the NIC buffer
being overflowed or the kernel buffer).



-----Burton

-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] Behalf Of David
Touitou
Sent: Monday, October 20, 2003 11:06 AM
To: [EMAIL PROTECTED]
Subject: [Ntop] Dropped by the kernel


Hello all,

I've just setup a test probe on my network.

The network is about 20 hosted servers, connected to the internet with a
3com switch. I'm using port mirroring on the switch to get the full
traffic on the "probe".

The global traffic is a bit less 10Mbps servers to internet and 2Mbps
internet to servers.

The probe is a P3-733 with 256 MB of RAM (for the moment).
The OS is FreeBSD 4.9-RC.
There are four network cards in the probe, one with an IP (used for ssh
and ntop web), one in promiscuous mode which gets the traffic, the two
other are not connected.
The cards are fxp (intel), with microcode loaded.
The disk subsystem is hardware RAID mirror with two 60GB UDMA66 drives.

ntop is cvs version from today.

I'm using the rrd plugin to get statistics on the hosts (only my subnets).

I'm having more and more packets dropped by kernel...
I saw the metrics going up and up and they are more than 70% packets
dropped right now (50 minutes of ntop sampling).


Here is current top output :

last pid: 50265;  load averages:  0.41,  0.30,  0.21         up
0+05:10:53  17:59:42

22 processes:  3 running, 19 sleeping

CPU states:  0.0% user,  0.0% nice,  0.0% system,  0.0% interrupt,  0.0%
idle

Mem: 91M Active, 92M Inact, 43M Wired, 11M Cache, 35M Buf, 10M Free

Swap: 2048M Total, 2048M Free

   PID USERNAME PRI NICE  SIZE    RES STATE    TIME   WCPU    CPU COMMAND
49994 nobody    34   0 93436K 84532K RUN      9:53 27.10% 27.10% ntop


Do you people have any idea why these metrics are going so high ?

A rapid browse in the archive showed that "bigger cpu" or "better nic"
was the answer... It does not seem to be my problem here.

What does this metric exactly show ?
Packets lost by ntop for its stats ?
Packets refused by kernel ?
Anyting else ?

Thanks,
David.

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

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

Reply via email to