I have two sites that are driving me nuts.

One site has 3 AP's, 2 x 5830 and 1 x 900. All run off the Trango supplied power supplies, which are provided 230V from an inverter run off 2 x 12V deep cycle batteries, which are in turn charged from mains supply.

The other site has 1 x AP and 1 x SU5800 in a similar power supply situation as above.

The second site has a 5800SU which backhauls to 1 of the AP's at site 1.

It is most commonly the same AP at each of these sites that stops responding, however it has also happened at one other site which is independant of either of these sites.

Today, the second site AP stopped responding for no apparent reason. The SU was still going, but the AP was not responsive.

About a half hour later, one of the AP's at site one also stopped responding.

I was able to go to one of the SU's hanging off the AP at site 1, and it was not able to see the AP or associate with it.

I power cycled both AP's, and everything has returned to normal.

I have another site which has 3 x 5830 AP's on it, and I have never had this problem (to date, AP 3 has only been up there about 3.5 weeks).

I use MRTG to graph stats and RSSI, and another program that pings each radio it can reach once every 5 minutes.

It seems that this happens approx every three weeks or so, sometimes it will happen 2 or 3 times in a couple of days, then all ok for about 3 weeks.

Each site has other equipment and radios that run off the same power source, and are unaffected at the time the AP's lock up, and it is always an AP, never an SU.

Firmware

Most guilty AP at site 1
[FPGA Version] 04092700 [Checksum] 53544B3E
[Firmware Version] AP 1p14H8002D05022301

AP at site 2
FPGA Version] 04092700 [Checksum] 53544B3E
[Firmware Version] AP 1p14H8002D05022301

AP1 at site with no issues
[FPGA Version] 02103000 [Checksum] 7ADD5AB6
[Firmware Version] AP 1p14H8002D05022301

AP2 at site with no issues
[FPGA Version] 04092700 [Checksum] 53544B3E
[Firmware Version] AP 1p13H8002D04093001

AP3 at site with no issues
[FPGA Version] 04092700 [Checksum] 53544B3E
[Firmware Version] AP 2p0a4H8002D05101303

Yes, I have Zyxel Zywall 5 routers/switches at Site 1, a 5 port switch at Site 2, I only ever have to power cycle the AP's to get the lot back up and running, restarting switches/routers does not help.

BTW the routers are completely set as eth to eth routing, no firewalling, NAT etc is used. This is a completely private, internal network, using private addressing schemes.

Yes, I have considered running straight off batteries, is 12v enough?

We want to put an AP on a non-powered high site using a battery and solar panel, is 12v ok for this?

If you drew the network on a piece of paper it would go (roughly):

My LAN -> ROUTER ->SU5800 -> AP5830 -> ROUTER ->SU5800 SWITCH AP5830 -> Fox 5800 -> ROUTER -> ROUTER -> Fox 5800 -> AP5830 -> ROUTER -> AP5830 -> SU5800 SWITCH AP5830 (a couple of SU5800's after that)

Each AP has SU's hanging off it, no SU2Su mode employed, where SU to SU was needed, we used a router behind the AP.

Reality is, connectivity all over the place is fine, it is just the random failure of the AP's that perplexes me.

Today, for example, 100% connectivity with no outages.

Probably be fine for another 3 to 4 weeks. Actually, last outage was 18 December 2005, so 3 to 4 weeks is the only pattern.

Yes, 12 Volts is enough. I have a purely enviornmentally charged site (Solar + Wind) and using 3 5830's plus a Link10 and a 2.4 802.11b/g radio all on the load of the 12 v. The most stable site i have...

12 Volts may not be enough.  How long are the ethernet from the POE injector to the APs? at both sites?  The APs work best when they have at least 12v at the end of the ethernet cable.  There might be too much DC voltage drop at the AP. The heater on the 5830s draws enough power that the old 20v power supplies can not give enough power to an 5830AP at the end of a 120 ft (35m) ethernet cable when the heater kicks on during the winter.  The first winter after the 5830 release we had some APs locking up due to the power drop and trango fixed it with firmware that added the temp and heater commands to adjust when that heater kicks on.  Trango also started to ship APs with 24v power supplies instead.  The old 5800 aps had a lower RF power output so it did not need the bigger 24v power supply that trango sells now.

Try shorter cables or more power and check the ethernet connection at the AP there might be a bad crimp.  The AP may have gone bad, we have replaced a few APs with those lockup problems and it seemed to solve the problem.

we graph rssi and bandwith for APs with up to 60 SUs, and so far have never had a single 5830AP problem of any kind. Did you change the default snmp community from public to something else? Or, are ...

I also graph APs, some with over 100 SU's, but is the link between your "graphing" computer and your troubled AP a Trango link?

This is when we had all the problems. We replaced all backbone links from Trango58xx<->SU to other types of radios and all these problems went away.

I believe that in some cases the trango gear actually corrupts packets, and so a corrupt snmp request to the AP locks it. We had 8 tower sites each with at least 2 5830 APs running fine for months. When we started the snmp polling/graphing every 5 minutes, the AP's started dropping like flies! had to go and reboot, tried replacing them etc .. was happening for almost 2 weeks before we figured it out. We stopped the polling from the office to the APs and all the AP lockups stopped.(Out office was connected to the network through a AP5800<->SU)

During this 2 week period, we installed remote power switches on the trango power supplies so we could remotely reboot the remote aps. Then we turned graphing back on and the APs started locking again.

Once we took out all Trango links from between our office and the towers (using trango only for ap's now, not backbone) we have never seen another lockup on an AP. Same network, over 1 year later....