Is anyone else out there using MIkrotik to handle a large number of PPPoE clients? 300+ users?

Our router is a Supermicro motherboard with P4 and 512Mb ram. We actually have 2 PPPoE routers built like this and they both are giving us trouble.

We have mostly Canopy CPE modems. We use Mikrotik as the PPPoE server for users. On average it terminates 400 PPPoE users at the heaviest used one. The users have various speed profiles from 128k to 1500k. Occasionally I'll see a user assigned something like a 256k account runnning at 2-3mbps. I just had one 256k account download at 2mbps for a full 24 hours looking at the graphs and then in Winbox. Anyone else see this?

Another issue is after running for a month or so the router starts flashing in the logs PPPoE license limit exceded and will not accept anymore connections. It does this even though its licensed for 500 PPPoE connections and we are under that at the time. When it does this the entire router becomes sluggish and you cannot even bring up terminal in winbox. Rebooting clears it up but many times one must physically reboot it since winbox and other services stop responding. Anyone else see this?

Matt

Is anyone else out there using MIkrotik to handle a large number of PPPoE clients? 300+ users?

Our router is a Supermicro motherboard with P4 and 512Mb ram. We actually have 2 PPPoE routers built like this and they both are giving us trouble.

We have mostly Canopy CPE modems. We use Mikrotik as the PPPoE server for users. On average it terminates 400 PPPoE users at the heaviest used one. The users have various speed profiles from 128k to 1500k. Occasionally I'll see a user assigned something like a 256k account runnning at 2-3mbps. I just had one 256k account download at 2mbps for a full 24 hours looking at the graphs and then in Winbox. Anyone else see this?

Another issue is after running for a month or so the router starts flashing in the logs PPPoE license limit exceded and will not accept anymore connections. It does this even though its licensed for 500 PPPoE connections and we are under that at the time. When it does this the entire router becomes sluggish and you cannot even bring up terminal in winbox. Rebooting clears it up but many times one must physically reboot it since winbox and other services stop responding. Anyone else see this?

Matt

This is scary to say the least!

How many wireless interfaces/antenna's are you using? My first suggestion would be to use different users per antenna, and I would prefer not to run more than 80 PPPoE sessions per wireless card, if possible. Also, not more than 200 PPPoE sessions per router, although I have seen routers with more than 200 sessions working fine (high CPU, but still works fine).

Is anyone else out there using MIkrotik to handle a large number of PPPoE clients? 300+ users?

Our router is a Supermicro motherboard with P4 and 512Mb ram. We actually have 2 PPPoE routers built like this and they both are giving us trouble.

We have mostly Canopy CPE modems. We use Mikrotik as the PPPoE server for users. On average it terminates 400 PPPoE users at the heaviest used one. The users have various speed profiles from 128k to 1500k. Occasionally I'll see a user assigned something like a 256k account runnning at 2-3mbps. I just had one 256k account download at 2mbps for a full 24 hours looking at the graphs and then in Winbox. Anyone else see this?

We have been seeing it occasionally on one of our supermicro P4 (xeon) with 512mb memory. It handles concurrent 1150 users. The users have anywhere from 1500k -> 8000k subscriptions.
The problem, as i see it, is race conditions in the software which are triggered under heavy cpu load or severe bridging errors (loops). Normally, under no to low load, the server (any server) has no trouble keeping up with new pppoe connections. It will accept the connections, authenticate and add simple queues etc.
But under heavy cpu load there is the increased chance of failure in the whole process of connecting a user.
The potential results are many, the queues are not set up properly (missing or misconfigured), other sessions get stuck as well, "ghosts" (sessions that are properly discarded), sessions that are there but cannot be deleted, server crash etc.

The version that has worked best for me, under load, is 2.9.24. It still has problems but in 2.9.25 they changed the syncronization of the authentication process which gave everyone a big headache for a looong time. Unfortunately, as far as i can tell, they still haven't managed to get all the quirks out of the newer versions yet.

My advice is for you to;
Reduce any load on the server. Less than 80% at all times is sufficient. But if you, for other reasons, have large waves of users connecting/disconnecting you may still get these problems. Especially when you get a loop and the same authentication requests are slammed at the server, over and over. Make sure your pppoe clients are equipment you know or you control.
Make sure no two mac addresses log in with the same username (otherwise things go nuts, no queues etc).
*Optimize* your pppoe server. Use filter rules sparingly if at all, use policy routing instead. Use static mss adjustment rules (the dynamic ones are incorrect) and besides that is one more thing that adds to the login process.
Last off, for the long run, perhaps it is prudent to go for a different pppoe server solution? Especially if the clients are not under your control.
You might want to go for homegrown/cisco/juniper/nortel/whatever solution when you decide you want to sleep at night.
I haven't, because of the people who are meant to administer it and we now use a strategy where we terminate pppoe further up in the network.

Another issue is after running for a month or so the router starts flashing in the logs PPPoE license limit exceded and will not accept anymore connections. It does this even though its licensed for 500 PPPoE connections and we are under that at the time. When it does this the entire router becomes sluggish and you cannot even bring up terminal in winbox. Rebooting clears it up but many times one must physically reboot it since winbox and other services stop responding. Anyone else see this?

Maybe, maybe not, your descriptions are missing any real clues but it could indicate that you have "ghosts" or otherwise incomplete sessions.

Another issue is after running for a month or so the router starts flashing in the logs PPPoE license limit exceded and will not accept anymore connections. It does this even though its licensed for 500 PPPoE connections and we are under that at the time. When it does this the entire router becomes sluggish and you cannot even bring up terminal in winbox. Rebooting clears it up but many times one must physically reboot it since winbox and other services stop responding. Anyone else see this?

I have recently experienced the issue where the pppoe server believes the license threshold is reached (while in effect, it is not) on v2.9.42 but not v2.9.24.
I suspect it may have some relations with manually deleting pppoe sessions.
But this is just a suspicion since I've only come across this once.
The pppoe server had worked fine until after i had deleted some / ppp active users to fix an ip pool + radius assigned ip conflict.