Debian Administration

There's a nice piece of software called ipaudit which runs on a gatway machine and keeps track of the internal machines external connections.

Once that is installed you can run another piece of software called ipaudit-web which makes pretty web pages from the data.

The output shows the amount of bandwidth used by each internal machine on a LAN and statistics about the most active external hosts - it might work for you if each of your services (like the voip box you mention) are on seperate hosts.

Here's a link to some screenshots.

I'd be interested in other solutions too - but I will say that there are a lot of cron jobs involved and it doesn't seem to load the gateway box unduly.

Steve
-- Steve.org.uk

I think you should check ntop. It has a nice web interface with graphics.

Horacio

You also may want to check out cacti as an alternative to mrtg I found it a little simpler to setup.

http://www.cacti.net/
http://packages.debian.org/stable/web/cacti

john

check out iftop: simple, ncurses-based, does the job for me.

Michel

I just got MRTG working really good today, after about a month of messing with it.

We are running it on a sarge VM. I currently am monitoring a W2K3 server and a ubuntu box. I dont really know what to tell you that cant be answered at this site:
http://www.linuxhomenetworking.com/linux-hn/mrtg.htm
http://www.linuxhomenetworking.com/linux-hn/mrtg-advanced.htm

I spent a bunch of time looking all over the web for info and I always came back to that site.
My one big issue was not using the correct MIB's, you have to load them up in the beginning of the script.
I am getting most of the info with SNMPgets and such.

Here's a mrtg example that simply uses ifconfig to obtain statistical info: ** /etc/mrtg.cfg **

WorkDir: /var/www/mrtg
WriteExpires: Yes
Refresh: 300

WithPeak[^]: wym
Suppress[^]: y
MaxBytes[^]: 96

Title[^]: Traffic Analysis for HOST.NAME

Target[eth0]: `/usr/local/bin/mrtg-data eth0`
MaxBytes1[eth0]: 13107200
MaxBytes2[eth0]: 13107200
Title[eth0]: Traffic Analysis for network device eth0 [External]
PageTop[eth0]: Internet Link

Target[lo]: `/usr/local/bin/mrtg-data lo`
MaxBytes1[lo]: 13107200
MaxBytes2[lo]: 13107200
Title[lo]: Traffic Analysis for network device lo [localhost]
PageTop[lo]: Local lo / loopback device

And /usr/local/bin/mrtg-data:

#!/bin/bash

DEV="$1"
if test -z "$DEV"; then echo No device specified; exit 1; fi

INFO=`grep $DEV /proc/net/dev | tr -s ' ' ' ' | cut -d: -f2`
RECEIVE=`echo $INFO | cut -d" " -f1`
TRANSMIT=`echo $INFO | cut -d" " -f9`
UPTIME=`uptime | tr -s ' ' ' ' | cut -d" " -f4-`

echo $RECEIVE
echo $TRANSMIT
echo $UPTIME
echo "HOST.NAME"

Enjoy, -doc

> not a complete article but here are a couple of things to try. Maybe you have already tried them?
* install snmp pkgs
To find them:
apt-cache search snmp | less

try using snmpwalk. Info on how to use it is in
"man snmpwalk"
snmpwalk should print out the mib tree for you. You *should* see an OID string for your ethernet card.

I don't remember the specifics for MRTG, but if you don't see a mib string for you network card, you're not going to get any snmp - data

* also check your /etc/snmp/snmpd.conf
The directions claim that you should not have to muck with this conf file. But just make sure that your snmp trap phrase is set to "public". There could be security implications for this. You *CAN* and *should* change the passphrase to something not so common, but make sure MRTG knows what it is.

* with CACTI, it *should* work out of the box for your if you use a template. I personally haven't gotten it to work since I can't find my MIB string for my card which runs the tulip driver.

* OH there is one more thing to look into. If snmpwalk can't find the mib string for your card, you can try to find a mib file for your card. I hear they are provided by the manufacture. If you are using 3-Com cards, I suppose you'd have better luck and things should just work easily in general. Perhaps others have a diff opinion on this.

- my 2 Cents.


The things that I definately want to be able to do is monitor the bandwidth that the VoIP box is using, monitor the bandwidth that the linux desktop box is using, and monitor the bandwidth that the connection from the server to the internet is using.

Fancier things that I would like to be able to do is to break this usage up into the seperate ports (maybe using a pie chart). And I am not sure if there would be able to get any stats on the responsiveness of the VoIP box, not even sure if this comment makes sense. Basically the purpose would be to see if there is any need to set up QoS for the VoIP box.

And a final question. Does monitoring network usage significantly affect network responsiveness or throughput?

If you're looking for a quick and easy view of your network traffic in "real time" try the ncurses based iptraf http://packages.debian.org/stable/net/iptraf

Hi there,

We have a 'Best of Breed' product called CODIMA which is a low cost app that does everything and more regarding VoIP. The Traffic Blaster in Codima helps to scale the network for pre deployment, offers a Visio plan for network management(with drill down facility) and gives a dynamic view of all usage.
It is a great tool for monitoring suppliers to nail down SLA performance and bandwidth provision. The reports are great too.
If you want to see it let me know
Kind Regards

Roger Curson

+44 (0)208 422 9699