Hello Guys, we are a ISP in Brazil, with more than 15k clients, and i use a lot of devices running RouterOS, we use PPPoe and recently i have a problem very strange, its very random, maybe 3 or 4 times per day the CCR 1036 goes to high CPU, like 80~90%, and a lot of clients are disconnected and sometimes the BGP Session gone to down.
I used to share all vlans in 3 or more PPPoe Servers(CCR 1036) and this problem was occured in different devices every day.
When the problem occurs i can see in Tools/Profile the option “firewall” and “queuing” go up, like 30~40% each one.
i think can be some client or CPE with problem, but i’m not able to discover, do you have some idea how can i debug this?
Do you use nat target masquerade? If so change to target srcnat. We had a similar problem with pppoe-server and high cpu. Since we changed all masquerade rules we hadn’t any problems.