Are you graphing CPU utilization on your StarOS APs? That might help
provide a clue. If someone is getting DOSed or there is a broadcast
storm, you might see high CPU utilization before/during the problem.
Matt
David E. Smith wrote:
Okay, Scriv and I are stumped on this one.
Over the last couple of weeks, we've started seeing some very odd
oddness on a few of our 2.4GHz POPs. Not all, just some. Here's what
appears to be happening:
A couple times a day, usually during business hours, something somewhere
generates a massive amount of noise. Connections which report an RF
noise of -90 start showing noise levels of -60 (or worse in some cases),
as reported by our StarOS access point. If it really is RF noise, it's
very broad, as it's simultaneously hitting five or six POPs, some
several miles away, but all at the same time.
The towers are all running StarOS on Mikrotik RouterBoard hardware, with
a mix of Orinoco and Prism cards, some with amps, some not. Some have
sectored antennas (180 degrees), some have omnis. Between them, the
towers cover just about the entire 2.4 spectrum (obviously, one channel
per access point, but we're using at least channels 1, 4, 6, 8, and 11).
Those towers are basically identical to several other towers that aren't
affected.
The other really really weird part is the crazy latency. Pings to the
APs themselves are reliable, and our backhaul links (5.3 and 5.8 GHz)
don't seem to be affected. And pings to our end-customers don't seem to
get lost, they just take their sweet time getting there. While "the
event" is happening, I've seen pings that take in excess of twenty
seconds to complete their round trip.
64 bytes from 10.232.175.130: icmp_seq=7 ttl=62 time=27239 ms
(I think that's my record. In that particular test, there were no
packets lost, they just took a very long time to get there.)
I've checked or replaced just about everything I can think of in our
network that might cause something like this, and frankly, I'm stumped.
I don't think it's a network problem (traffic bursts or similar) because
of the weird bursts of RF noise. But that'd have to be one helluva burst
of noise to do what it's doing - affecting every channel across ten
miles at once.
I can go into more detail on any part of the network if you like, though
I think all the likely-relevant details are covered here.
Help!
David Smith
MVN.net
--
WISPA Wireless List: wireless@wispa.org
Subscribe/Unsubscribe:
http://lists.wispa.org/mailman/listinfo/wireless
Archives: http://lists.wispa.org/pipermail/wireless/
