Hi everyone!
I need help to understand a problem that is happening at least for more than one year! I will tell the story, is needed for understanding… Sorry for my poor english…
In the last 3 months of 2006, a problem began to happen in one of my POPs (I´m an ISP), in the center of my town. In that time, 70% of the clients (± 60) started to loose connection, the stability was lost. Normally, after 13:00 hs, the network was terrible. This was not a rule, problems occurred out of the afternoon, but the concentration was in that period. I changed everything (antennas, cables, radio, pig, etc) and the problem persisted. Since 2006, many things were changed, my network became good, but that ghost was always near. It happened again in march / april, august / september / october 2007, and now, as I write, again.
Ok, let´s go to the history of changes…
The equipaments always were: 1 omni (clients), 1 directional (backbone to link), cables, pigtails, radio. It happened too with 4 90º sectorials, but since 2006, is an Omni.
I have replaced ALL equipments, so it was not water, neither a connector, or an antenna… EVERYTHING was changed more than once, ever without good results. So, I started to change the RADIOS.
2006 - 1 Ovislink 5460 was used to link to the backbone, and an AP2000 with an Orinoco card was in the omni. After replace cards, antenna, cable, and the AP2000 itself, I started to use Mikrotik, and here is where I need your help!
In 2006, I´ve replaced the AP2000 to a PC (Intel 450 Mhz, 128 Mb RAM) with 2 PCMCIA Senao (Prism) 200mw with Mikrotik, and became worse. So, I came back with AP2000… I don´t know why, but after a few days, the network was OK with AP2000, the same original equipment…
Since that, all was fine for some months, when the problem came back… So I decided to replace Ap2000 again, and that time it worked ok with the same Mikrotik in a PC - it was in the tower yet. Two months later, problems again. One more time, I have replaced cables, antennas, pigs and the PC to a RB532A with 2 R52… ALL is new, but the problem is old…
In december, we replaced all backbone links (from ISP to towers) to 5.8 Ghz. This POP was the first one. We imagined that could be RF pollution, but we are in a small town, we are the only wireless ISP with 4 towers but I do not consider this place a polluted one… But we did it. With this action, 3 wireless links in 2.4 became 5.8, cleaning the spectrum… But the problem persists again…
I dedicated an entire day too to monitor the energy with a multimeter… No abnormal oscilations before or after the no-break witch feeds the RB…
Another thing we´ve done: reduced the number of clients in this AP, changing them to another towers. These were clients with bad signal or bad TX/RX (IE 11/1). It solved for a while, but now is bad again, even with less clients than other RBs…
That´s enough history. I need your help for the present!
The system now is a RB532A with a R52 (clients) and an EMP-8602 (backbone) - both Atheros AR5413. The R52 works in 2.4 with a 15 DBi omni antenna (Hyperlink) and the EMP-8602 works in 5.8 with a 24 dbi directional antenna (Zirok WLL-455). MikroTik is 2.9.38.
Now, what happens is that are some periods of time when all of my clients stops receiving packets, or the time goes too high. It affects ALL clients, and sometimes, even in interface WLAN1, wich is the 5.8 backbone. Is very intermitent, sometimes only some packets, other times hours of high times and loss… And other times perfect days…
In the image above, we see the ping to 3 clients (ip 248 is a D-link DWL-810+, 253 is an ovislink 5460AP and ip 5 is a RB133, all with few distance from AP and excelent signal). No queues are defined to these IPs. The ping to 172.16.x.1 is my backbone. Note the last 5 lines: From down to up, note the times so near (179, 290 and 298 ms) in the fifth line. The fourth line down to up is 548, 529 and 560. The last three, timeout. Is clear that is a general problem, not local. In that moment was high anyway, in normal situations it should be around 10 - 30 ms.
See the clients in registration table… Good signal, good TX/RX…
Informations about the interface wlan2 (clients):
[admin@apxxx] > interface wireless monitor wlan2
status: running-ap
band: 2.4ghz-b
frequency: 2427MHz
noise-floor: -100dBm
overall-tx-ccq: 89%
registered-clients: 46
authenticated-clients: 46
current-ack-timeout: 38
current-distance: 38
nstreme: no
current-tx-powers: 1Mbps:17,2Mbps:17,5.5Mbps:17,11Mbps:17
notify-external-fdb: no
To me appears to be ok, good ccq, noise-floor… Maybe someone can open my eyes for some detail I haven´t seen…
Well, another information that can help… I have others RB532A with more clients and traffic, but the CPU load in this POP with problem is always higher… Not 100% (something between 60 and 80%), but aways above the others. To lead me to insanity, it rainned yesterday…
Any help will be very appreciated… Thanks in advance to those who readed until the end…
´s
Denilson