There ought to be some sort of monitoring for the current/past state of the network. Currently this page is a log of BenjaminJencks attempts to set this up.
Vision
Something, probably Cacti, monitoring all the APs for
- Number of associated stations (done with assoc_count script)
- Link throughput (built into snmpd, unless bridges cause weirdness (I don't see why they would))
- Link status? This might be derived from throughput
So Far
- mississippi has snmp tools installed
- metrix-naya-{n,s}w has snmpd installed
- metrix-naya-sw has /usr/local/bin/assoc_count, which prints the number of associated stations.
snmpd is configured to use assoc_count for OID 1.3.6.1.4.1.2021.8.1.101.1, otherwise known as UCD-SNMP-MIB::extOutput.1. This is a string, I don't know if that will cause problems later. If it does, it's simple enough to make assoc_count return the number as an exit status, which snmpd reports as an integer.
- This means you can count stations with snmpget -c public -v 1 10.11.104.2 1.3.6.1.4.1.2021.8.1.101.1
Next Steps
- I want to set up the monitoring and see that it works before I go installing the agent on all the metrixes.
- It seems like a good idea to make the names of the APs resolve, so the monitoring software can use symbolic names. DNS or /etc/hosts? Should the mississippi network get a subdomain?
- Is mississippi the right box to run the monitoring on? It's old, slow, and has very little HDD space. But, it's on the network so we don't have to cross the internet to get stats. Is there a Cacti, MRTG or similar installation somewhere we should reuse? If so, we need a VPN to mississippi.