Blob Blame History Raw
=encoding utf8
=head1 NAME

mrtg-squid - using mrtg to monitor Squid

=head1 DESCRIPTION

Squid 2.3 knows SNMP and you can therefore use mrtg to
monitor it quite easily.

I have made some modifications to mrtg which simplify this.
My work is based on earlier modification made by:
matija.grabnar@arnes.si and kostas@nlanr.net.

=head1 MODIFICATIONS

I added new code for displaying correct units to the
previous patches "perminute" and "perhour" ("option" tokens),
which allows other measurement in addition to "persecond".

Then I created a new option token "dorelpercent" which
allows the calculation of the percentage of IN-stream / OUT-stream on
the fly and then displays it on a fixed scale from 0% to 100%. For my
requirements, this does good work. Maybe someone wants a floating
scale. It should not be a problem to implement it, too (but 
give me an option to keep my fixed scale). If IN-stream is always
less than OUT-stream both lines (OUT-stream and relative percent)
are always displayed on top of IN-stream bulk. Otherwise this
option makes no sense. With this option you can display hitrates,
errorrates (for router monitoring: rel. droprates) easily now.

If you use this options please consider that you need a 5th
colourname/value pair in your Colours statements!

Due to some discussion on this list, I have implemented two
tokens too:

"kilo" and "kMG"

"kilo" should contain the value of k (1000 or 1024), where 1000 is the
default.

"kMG" is a comma separated list of multiplier prefixes, used
instead of "", "k", "M", "G", "T" on the MRTG display. Leave the
place free, if you want no prefix.

Also an incomplete list of OIDs for the new SQUID release is
added.

You may need to turn on snmp_port in squid.conf to as it is disabled by default.

I hope you enjoy it.

=head1 CONFIG EXAMPLE

First load the squid mib

 LoadMIBs: /usr/share/squid/mib.txt

You can measure responsetimes in ms and display it with MRTG
correctly with:

 kMG[measure-ms]: m,,k,M,G,T
 short[measure-ms]: s

You can display now MB/s as 1024*1024 B/s with:

 kilo[volume]: 1024

Assuming you're not running squid's SNMP on the default snmp port, you need
to include a port number in your target line:

 Target[proxy-hit]: cacheHttpHits&cacheProtoClientHttpRequests:public@localhost:3401


A sample config for squid:

 Target[proxy-hit]: cacheHttpHits&cacheProtoClientHttpRequests:public@proxy
 Title[proxy-hit]: HTTP Hits
 PageTop[proxy-hit]: <H2>proxy Cache Statistics: HTTP Hits / Requests</H2>
 Suppress[proxy-hit]: y
 LegendI[proxy-hit]:  HTTP hits
 LegendO[proxy-hit]:  HTTP requests
 Legend1[proxy-hit]:  HTTP hits
 Legend2[proxy-hit]:  HTTP requests
 YLegend[proxy-hit]: perminute
 ShortLegend[proxy-hit]: req/min
 Options[proxy-hit]: nopercent, perminute, dorelpercent

 Target[proxy-srvkbinout]: cacheServerInKb&cacheServerOutKb:public@proxy
 Title[proxy-srvkbinout]: Cache Server Traffic In / Out
 PageTop[proxy-srvkbinout]: <H2>Cache Statistics: Server traffic volume (In/Out) </H2>
 Suppress[proxy-srvkbinout]: y
 LegendI[proxy-srvkbinout]:  Traffic In
 LegendO[proxy-srvkbinout]:  Traffic Out
 Legend1[proxy-srvkbinout]:  Traffic In
 Legend2[proxy-srvkbinout]:  Traffic Out
 YLegend[proxy-srvkbinout]: per minute
 ShortLegend[proxy-srvkbinout]: b/min
 kMG[proxy-srvkbinout]: k,M,G,T
 kilo[proxy-srvkbinout]: 1024
 Options[proxy-srvkbinout]: nopercent, perminute

=head1 AUTHOR

Andreas Papst E<lt>andreas.papst@univie.ac.atE<gt>
Dirk-Lüder Kreie E<lt>deelkar@gmx.deE<gt>
Chris Chiappa E<lt>chris+debian@chiappa.netE<gt>