I recently decommissioned our old unix PPPOE server in favour of Mikrotik OS because of ease of configuration.

We authenticate against a Freeradius server with a mysql backend.

The old solution never once gave issues, however, since we deployed the Mikrotik box it seemed some of our clients would
loose connectivity. The comms to the box is in good order 2ms response no packet loss.
The second tower also connected to the same box does not experience this.

It seems a client will connect, have comms, but as soon as data transfers the session will die.
However it does not drop away from active connections, simply no comms can occur through the tunnel anymore.
The client will disconnect and reconnect and will be working for a short while again.
Sessions don't last longer than 30secs.
After a longer peroid the log shows client disconnects "peer not responding".

Downgrade to v3.2 didn't fix this, however downgrade to 2.9.50 did , also recommissioning our old box also does.

All stable and been running since.

Any advice?

Infact here is the supout.rif file for those interested.
This is however of the system in working order but many features disabled, after downgrade to 2.9.50

http://www.thavinci.za.net/Downloads/mi ... supout.rif

The support file only can be read by the team mikrotik (support), send to support@mikrotik.com


Max
http://mikrotikexpert.com
http://maxid.com.ar

It seems infact both towers do the same.

No-one can download anymore.
Even a 1.6mb file causes it to bomb out.
However today inoticed that sometimes the tunnel DOES come back to life on its own, but takes so long that web-sites time out, downloads stop.

Now i normally would think its a problem with a connection to the box, but i keep confirming the links to the box is fast stable no packet loss , no higher than 2ms response time, and can infact be pinged by myself constantly if i put my pc on outside.


Looks like we going to have to go back to our old custom solution to avoid loosing all our customers, unless we can find the issue today.

More information on this one....
I have 4 Senaos as my access points connected in following fasion...

MT <------Senao
10.10.10.2/16 10.10.10.4/16
<------SenaoWDS<-------SenaoWDS<------SenaoWDS
10.10.10.5/16 10.10.10.6/16 10.10.10.7/16

Now the MT Does the PPPOE and uses 172.16.x.x for clients and 172.16.1.1 for itself.
The 10 range is simply to comms with the senao's.
Here is another odd thing that doesnt work.
The MT with NO firewall rules is unable to even ping any of the senao's! However it can arp ping them.
If i connect a pc on the same segment as the senaos connecting to the MT i am able to ping the MT and all AP's yet the MT can't ping the seanos.


The issues we are having at this stage is extremely unusual!

H(a)i,

It seems a client will connect, have comms, but as soon as data transfers the session will die.
However it does not drop away from active connections, simply no comms can occur through the tunnel anymore.
The client will disconnect and reconnect and will be working for a short while again.
Sessions don't last longer than 30secs.

disable MSCHAP1 and MSCHAP2 in PPPoE Server

Wow, my one customer that could stay on longer than 30 seconds seems to be on last 3 hours!


Will have to confirm this with everyone, but so far so good!


What is youre technical explaination behind this?

And thank you for input. Hope this continues working like this , if so ive found the solution!!!!

I'd also like to know what difference this makes and if I should do the same on my server.

you have some kind of packet loss. by turning off encryption it increased the performance

Well, I guess that's not really what I want to do then... I'd prefer to have encryption enabled for PPPoe over Wireless.

And to confirm , that did solve my problem.

And indeed it can be a packet loss related problem for the one client.
However it was strange to have the same issue 10metres away from the tower.


But all solved now!


Thanks

So, I have 3 RouterBoards... rb333, rb532a, rb133.

I intstalled ROS 3.3 on rb333... My clients Complained that after about 3+ minutes of login, They cant browse anymore. I went to one of my clients to check to see if that is true. Because I have setup everything inside my test room and tested it for one day and everything is fine before installing it to him. When I tested the PPPoE connection in his computer, I browsed for some 3 minutes. Then... Page Cannot be displayed anymore in IE or Firefox. So, I opened a cmd window and pinged continously the DNS Server. Then, I disconnected then reconnected again via pppoe... after three minutes of browsing ... Page cannot be Displayed again... The continous ping also stopped... Timed out... This Client is only about 50 meters from the AP.

I then Setup the rb532a routeros v3.3with the same setup as the rb333 ( I was thinking that because maybe the rb333 is still in a production stage when I bought it??? That is why it konked out!? I am one of the first rb333 buyers of wisp-router.com ). And still all clients, except the ones inside the test room experiences the 3 minute black out...

I then replaced the rb532a with the rb133 with the RouterOS 2.9.50, Everything went fine... This rb133 was the first I setup for PPPoE as a test unit. But because the rb532a and the rb333 is much much faster than the rb133, I opted to use the rb333 or the rb532a (that is why I bought it).

So I now downgraded the rb532a to ROS 2.9.50 and set it up again as before... Then replaced the rb133 with the rb532a and now everthing is working ok...

Now my Question:

you have some kind of packet loss. by turning off encryption it increased the performance

Normis, why can the rb133 handle the load with encryption using ROS 2.9.50, Against the rb532 ROS v3.3 can't handle the encryption? I have also tried to upgrade the rb133 to v3.3 during the morning hours... so most is offline. My mate called me to say that he is disconnected again after three minutes. But I can see him logged in on the rb133 with ROS 3.3

Please Normis, tell us if this is a bug and will be fixed...

If this is due to what you say like Performance issue and we need to encrypt those pppoe connection... I think I don't know what will I do with my rb333 (joke).

Or do we need a performance of something like a multicore Processor to handle the PPPoE encryption of ROSv3?

Thanks,

ColdHaze

PS.

I also tried ROS v3.0-3.3

Just an update, my setup was using a P3 PC 800mhz with 512mb SDRAM.
And i had the issue with 3.3 and 2.9.50 infact.

So i had lots of cpu "juice" in comparison to a rb, yet still had the problem.

I also did check packet loss and infact there was for the far away client, but close clients had NO packet loss.
And an average of 2-3ms pings.

It did seem worse on clients further away that DID infact have packet loss.

Ah ok... You have lots of cpu (juice/power)... Ok Maybe I should post another thread.

But why those things I experienced above??? Why would 2.9.50 work on my setup? Maybe Normis can answer this.

I am also using Ubuntu, Freeradius and mysql for authentication.

So maybe they can fix that packet loss and disconnection of PPPoE in the coming releases of ROS 3.

And hope that my rb333 will be happy again... (joke)

ColdHaze

Hi all,

I just want to say that i have the same issue and this is just a bug in MK i test many many things and confirm that. i hope MK solve this issue as soon as possible.

Best Regards

You replied to a thread that was 15 months old. MT have released 21 updated versions since then with lots of fixes. Make sure you're running the newest version.

You replied to a thread that was 15 months old. MT have released 21 updated versions since then with lots of fixes. Make sure you're running the newest version.

I am sure the problem still exist.