Monitoring if our wireless network is actually working in locations

November 16, 2022

We provide a multi-building wireless network to the department (in addition to the university wide wireless network provided by the central IT people). For reasons beyond the scope of this entry, we don't have much innate visibility into how this network is doing in the various places it exists. This means that our only current ways of finding out whether or not it's currently working right somewhere are to wait for people to notice and then report problems (which doesn't always happen) or going there ourselves. We've recently decided that we'd like to do better than this.

The obvious thing to do is to put some boxes on the wireless network at various (physical) locations, and then have our monitoring system check to see that they're still visible (for example, through pinging them). This is fairly brute force but it's also a good end to end test; if we can reach these boxes over the wireless network, they can see the network, associate with it, and contact our DHCP server to lease addresses.

We need these boxes to be inexpensive, because to be useful they'll have to be in relatively exposed locations where they might walk off. We have strong opinions that they need to be wall powered, not battery powered, because we don't want to be going around every year or so to re-find them all and replace their batteries with a new set. Ideally they'd do something useful, like report the ambient temperature around them. Unfortunately we haven't found anything that combines these three attributes together; inexpensive wifi temperature sensors seem to all be battery powered, for example.

What we're currently experimenting with is wifi controlled smart power plugs. These tick two out of the three ideal features; they're generally inexpensive, and they're powered by the wall instead of batteries. They're also available from reasonably reputable brands, which helps give us more assurance that a unit isn't going to burst into flames some day, although in actual deployment I suspect that we'll tape over their power outlet with a 'do not use' label.

(If you get the right model, you can even get Unix tools to talk to it.)

We've only just put a couple of these wifi smart power plugs on our network, but so far they do seem to ping consistently. We'll have to see if that keeps up, but I'm cautiously optimistic. Of course I'd love to discover a better option, but I honestly suspect that there isn't any (for example, I can see why hardware designers would want to avoid wall power if they can).


Comments on this page:

By Keith at 2022-11-17 04:10:58:

We once tried this with raspberry pi nodes powered by PoE to USB adapters, and used the Ethernet link to pull telemetry data.

We had complaints about coverage in a few places, and half a dozen rpi 2B nodes left over from a prior project, so we put a few in static places, and kept a couple available for testing on-demand after new complaints (e.g. in conference rooms).

It worked okay.

I no longer work there, so parsing wireless interface scan status and generating meaningful reports / alerts is left as an exercise for a future implementer.

By Keith at 2022-11-17 04:34:15:

Also, the WLAN at work used 802.11X authentication, so the Ethernet link was useful for updating the X.509 client certificates for wpa_supplicant on the rpi 2B nodes.

By Keith Farrar at 2022-11-17 05:14:58:

Typo: "802.11X" should be "802.1X".

Also, on the raspi linux monitoring nodes, WLAN service info was pulled from NetworkManager utility nmcli scan dumps output.

One could stuff values from some subset of relevant fields in "nmcli -f all dev wifi list" into a time series database (or sqlite, or

Also, the raspi nodes had static DHCP lease reservations and ran iperf3 servers, so I could poke them from my laptop or other wireless clients.

Hope someone finds this mini braindump useful

-Keith Farrar
By Miksa at 2022-11-17 11:26:33:

Would it be possible to monitor the amount of traffic going through them? I would think the access points are connected to managed switches and it could be possible to gather traffic statistics from the ports. Possibly even what IPs have shown up in the port, indicating which VLANs are working at the access point.

By cks at 2022-11-17 14:01:59:

We can't gather AP-level or switch-level information for a collection of reasons, including that we don't actually run the APs and the switch fabric. (We're getting a free ride on the enterprise APs and switch infrastructure that the university has for the university-wide wireless.)

By James (trs80) at 2022-11-18 11:22:16:

You don't need to reinvent the wheel, there's a few open source projects (and plenty of commercial ones):

https://docs.google.com/spreadsheets/d/e/2PACX-1vSo0j6Rc2gKw_tlU2J5hUHhX8NbCdXP-lBLzn8pARKqq1_QVxPZNe5gw-rELxr9XCTrAVpylHWtqDHH/pubhtml

By Bob at 2022-11-19 11:12:26:

This seems like a good use for a Raspberry Pi Pico W, which is cheap, available even during the chip shortage, can have a cheap temperature sensor easily added and can be powered using a USB wall wart.

Written on 16 November 2022.
« It's useful to think about a 'ground up' recovery of your environment
Understanding how fast Ethernet really is (and in what units) »

Page tools: View Source, View Normal, Add Comment.
Search:
Login: Password:
Atom Syndication: Recent Comments.

Last modified: Wed Nov 16 23:20:37 2022
This dinky wiki is brought to you by the Insane Hackers Guild, Python sub-branch.