[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [hobbit] Graph Network Usage

Have you tried any recent snapshots of Hobbit? It already creates an inbound,outbound Bits/second graph for all of your interfaces, as well as another graph which shows Packets/second (In,Out,& Retrans). I also like the Clock offset graph, which comes in handy for pointing out how poorly vmware instances keep their clocks synced :)

-Charles

Jeff Stuart wrote:

As we all know, you can use MRTG + SNMP to graph network usage. BUT instead of having to install 2 new programs + an external hobbit script that runs either on each box or on a centralized location, you can use this external script that runs on the client.

Basically it just grabs the data from the /proc/net/dev file. Note: this probably only works on Linux.

Then you add the following lines to your hobbitserver.cfg:

TEST2RRD="cpu=la,disk,inode,...,hobbitd,bandwidth=ncv"
NCV_bandwidth="bytesin:COUNTER,bytesout:COUNTER"
GRAPHS="la,disk,inode,...,hobbitd,ncv,bandwidth"

Then in your hobbitgraph.cfg, add this entry:

[bandwidth]
       TITLE Network Traffic
       YAXIS Bits/Second
       DEF:inbytes=bandwidth.rrd:bytesin:AVERAGE
       CDEF:in=inbytes,8,*
       DEF:outbytes=bandwidth.rrd:bytesout:AVERAGE
       CDEF:out=outbytes,8,*
       CDEF:total=in,out,+
       VDEF:pct95=total,95,PERCENT
       LINE2:in#00FF00:Inbound
       GPRINT:in:LAST: \: %9.1lf (cur)
       GPRINT:in:MAX: \: %9.1lf (max)
       GPRINT:in:MIN: \: %9.1lf (min)
       GPRINT:in:AVERAGE: \: %9.1lf (avg)\n
       LINE2:out#FF0000:Outbound
       GPRINT:out:LAST: \: %9.1lf (cur)
       GPRINT:out:MAX: \: %9.1lf (max)
       GPRINT:out:MIN: \: %9.1lf (min)
       GPRINT:out:AVERAGE: \: %9.1lf (avg)\n
       LINE2:total#00CCCC:Total Usage
       GPRINT:total:LAST: \: %8.1lf (cur)
       GPRINT:total:MAX: \: %8.1lf (max)
       GPRINT:total:MIN: \: %8.1lf (min)
       GPRINT:total:AVERAGE: \: %9.1lf (avg)\n
       LINE1:pct95#0000FF:Outbound 95th percentile
       GPRINT:pct95:\: %6.2lf %Sbps \n

And then finally setup this script as an external script for each client:

#!/bin/sh

/bin/grep eth0 /proc/net/dev | /bin/awk '{print "bytes_in : " $2 "\nbytes_out : " $10}' > $BBTMP/bandwidth.txt

$BB $BBDISP "status $MACHINE.bandwidth green `date`

`cat $BBTMP/bandwidth.txt`
"

/bin/rm $BBTMP/bandwidth.txt
exit 0

Hope this helps someone!

P.S.
Thx Henrik for your EXCELLENT tutorial on the NCV graph stuff. Made my job a 1000x easier! :) Also, in the next week, I'm gonna be blowing my boss's mind with some of the stuff I'll be able to track and graph. Hehe.. there's just SOO MUCH you can do when you can do it on the command line vs a windows program that can only access stuff via SNMP/TCP/UDP probes.