Very high cpu usage with 20+ sstp clients

Hello, I am running some mikrotik routers with sstp server. I have two different situations, both with about 20-30 sstp clients (mikrotik 751); I get in both routers 100% cpu usage; this is not really a problem because it is responsive and gives good speed in sstp tunnels.

what do you need to debug this properly? I am not using an encryption-enabled profile.

edit: host is rb1100ah, I have also tried with x86 and inside metarouter, I get the same results.

Hello
what version of routeros you use?

Same here, 30 sstp gives 50% CPU on rb1200. That’s a lot of overhead. 5.12 currently and average throughput on all tunnels is well under 1Mb

What is the total encrypted traffic you are pushing? RB1000 with 200ish clients and about 40-50mbit of traffic pushes 25-30% cpu. I can see an RB1200 sweating with 1/2 that traffic.

Same problem here,
RB433, ROS 5.19, ~60 SSTP connections and constant 100% CPU usage. ROS responds quite well and almost no disconnecting was reported, though.
Clients are pushing about 500 bytes each 5 seconds, so there’s almost no traffic. With 5.12 from which I updated, CPU usage was around 15%.

I am using 5.19

any change about sstp in 5.20? I’d rather not upgrade a production router without a good reason.

Unfortunatelly have same problem with 5.20 - 100% cpu load mostly caused from ppp when sstp server active.
After restart sstp server it is ok until next sstp login.

y use this simple script:

/interface sstp-server server set enabled=no

:delay 5

/interface sstp-server server set enabled=yes

and set in scheduler with 1 hour period

poor solution, but works

I am having a similar issue with way less clients. 10 users total and maybe 3-4 sstp. RouterOS 5.20. RB2011.

I have the exact same issue on a RB1100AH with 6/7 clients. ROS 5.19 AND ROS 5.20

MT confirm its a bug and will be fixed next release. Can you get away with using L2TP/PPTP? We have to use SSTP because one of our cable providers blocks the other 2!

Best of luck

Jon

I tried also pptp but sstp is like skype: it works everytime, with static or dynamic ip and behind any kind of nat and firewall.

High CPU usage caused by SSTP will be fixed in v5.21

5.21 release date?

“when it’s done”? :stuck_out_tongue:

And any news from anyone? it’s really fixed in 5.21 version?

Bump, is this fixed in 5.21? The changelog mentions sstp but isn’t clear on it fixing this issue

it should be fixed. At least testing routers did not show this behaviour again. So please check this locally.

5.21 fixed this for me. I have since upgraded to 5.22 and have not seen this issue with it either.