I am facing this problem since january, and even tech support cant give me a solution.
I have 6 CCRs 12G 4S acting as PPP servers.
Each PPP server listening on vlans, working in pairs ( PPP1 and PPP2 with vlans 102 and 103, PPP3 and PPP4 with vlans 101,201 and 251, and PPP5 and 6 with vlans 221, 231, 22 and 252 )
These way, there are almost 1200/1300 users per CCR at a time, except when one goes down, where the other takes control until the reboot or recover.
Some firewalls rules to protect itself and customers ( like TR069 protocol block from outside ).
The problem is that -apparently- with traffic higher than 500 mbps, at some point the interface PPP side goes down, and up again. Not the WAN side.
This of course disconnects all users.
The ppp bandwidth control is with simple queues, most users dynamic IP, some others fixed public IP.
With versions like 6.29.1, the behaviour was suddenly unkown queues, IP address hooked by the ip pool, so must to reboot to release all and goes back to normal operation.
With 6.39.x to 6.40.2, latest, as told before, -same rules, same all- the interface flaps.
Tried disabling IPv6, no luck.
Disabled ppp compression, disabled ip neighbours, but nothings seems to help.
Seems crazy, no autosupout, nor log info, nothing.
Is like if someone unplugg/plug the PPP cable!!!
If someone can explain what to do, I'll be grateful.