I’m using mrtg to keep an eye on our backhauls and customers, to watch for signal strength changes that might be a radio, cable, antenna, etc going bad. As I’d expect, I see daily variation in the signals based on sunrise/sunset, nice sinewaves for a few db change throughout the day.
However, I’ve also found sudden instant changes of 15-20db, sometimes at the same time a couple of days in a row, on both 2.4 and 5G. It looks exactly like somebody throwing a switch. My boss thinks the radio is lying to us, that something in the driver is getting confused.
I have 4 Prism and 1 Atheros interface on 2.4GHz. One day I have noticed similar problem with this Atheros. Signal strenght drop 20dB down, but only for clients in Atheros interface. This situation was for a couple of hours.
UniKyrn Can you describe how you make this signal monitoring with mrtg? Maybe some examples.
It’s rather a jury-rig. I’ve got a tcl/expect script that pulls the wireless stats from each radio and stuffs it into a file. Then I run mrtg with each graph configured to call an external program I wrote that pulls the stats for the link in question from the file the expect script left behind.
Oh yeah, I should probably mention that the 2.4 links are all prism based, the 5G links are all atheros based a/b cards. If it’s a driver problem, it’s common to both then.
No, the two backhaul links that show this behaviour are already throttled to 24Mbit, and they’re locked at that rate according to my graphs. I actually graph speed and signal strength on the same graph for each link.
There is certainly something odd going on, but I’m out of ideas to explain it.
Looking at what I wrote earlier, I think I should mention one last thing. These are INCREASES in signal strength, not decreases. A link that’s run for weeks at -72db will suddently increase in strength to -56db for a few hours, then drop back. I do see customer radios drop to weaker signals in the middle of the night sometimes as they go into power saving mode or something, I can explain that. Why a link would suddenly get better instantly for awhile and drop back is the confusing one.
My guess was a noise source vanishing for awhile that might have detuned the receiver, so when it vanished the signal jumped in power, but at both 2.4 and 5G at slightly different times and on some but not all of our links?
UniKyrn Can you describe how you make this signal monitoring with mrtg? Maybe some examples.