Status

chevy is running Cacti, and monitoring all of the deployed metrixes and ciscos. The graphs are available at

http://chevy.personaltelco.net/cacti/

Use username: guest, password: freewifirocks.

Management interface at https://chevy.personaltelco.net/cacti/

Set up

Metrix

To set up a metrix box for monitoring:

metrix# remountrw
metrix# apt-get install snmpd ash
metrix# vi /etc/default/snmpd # set TRAPDRUN=no
metrix# cd /var/lib
metrix# cp -a snmp/ /ro/var/lib
metrix# mv snmp/ /rw/var/lib
metrix# ln -s /rw/var/lib/snmp/ .

// Get snmpd and assoc_count from another metrix
naya$ scp snmpd.conf root@metrix-ed.mississippi:/etc/snmp
naya$ scp assoc_count root@metrix-ed.mississippi:/usr/local/bin

metrix# /etc/init.d/snmpd restart
metrix# remountro

Add the device to Cacti with the Metrix Box template.
Create the Associated Stations, ath0, ath1...athN graphs.
Add the device to the main graph tree.
Add the Assoc. STAs graph to the Assoc STAs page.

WGTs

The WGTs are using the ipkg repository

http://www.personaltelco.net/~russell/kamikaze/r3291/packages/

and have the snmpd package (with its dependencies) installed. They use the same snmpd.conf and assoc_count script as the metrixes, so they look pretty much the same to cacti.

Diagnostics

/usr/local/bin/assoc_count is a homebrew script. It reports the number of associations on ath0. It only works with madwifi-ng. snmpd.conf has the line

exec assoc_count /usr/local/bin/assoc_count

to enable it. This exports the association count on OID 1.3.6.1.4.1.2021.8.1.101.1. Note that this OID is the standard for external exec'ed scripts and a second exec would use the same OID, except ending with a .2. Other than that, everything is standard net-snmp exports. The assoc_count script, is quite simple, it looks like:

echo $((`wc -l < /proc/net/madwifi/ath0/associated_sta`/3))

TODO

Figure out why we are getting SNMP errors
Enter remaining devices
A (remote) backup strategy for psql tables
Get a PTP signed cert for chevy, so we don't warnings on https.