PPPOE Issue

RouterOS General
1

Dears

I have 500 Users Connected via radius when 100 users disconnected RB Freezing and I faced high CPU more than 90%

I check profile i see the firewall most usage when users disconnected

I have more than 10 CCR all of theme same problem i change the version but still same think

Note : RB i used 1016 or 1036

Thanks

2

Are you using OSPF? See Most underused and overused RouterOS tools and features by Janis Megis search for “High CPU load on PPPoE server”.

3

Dears

I don’t use any dynamic route ,

4

Are you using radius? What’s the ROS/ firmware version on the CCRs?

Post an /ip export

5

Are you using masq or src-nat to nat your users ?
Don’t use just masq. When you get a many pppoe disconnecting the firewall has to remove all connections causing high CPU.


Sent from my iPhone using Tapatalk

6

https://youtu.be/BkZHRD6svQU


Sent from my iPhone using Tapatalk

7

That’s exactly why I asked about such export :smiley:

8

Yes i am using radius

Version: 6.3x. (stable) , I faced with more 5 router

9

in PPP i don’t using NAT , I know the problem causing remove connection

If i change close wait timeout or TCP close can it helpful

Thanks

10

I confirm absolutely same problems on my routers.
Heard about same problems even on 1072 with 1k PPPoE sessions.
I have even OSPF neighbors disconnects sometimes.

During my investigation, I found that problem appeared in ROS 6.33, when

*) ppp - added on-up & on-down scripts to ppp profile;

functionality was added to ROS. Made some tests with downgrade, CPU load was much lower.

I made a ticket Ticket#2017020822000267 with no success.

As always, guys from MT are ignoring some problems. Now I am experiencing big problems with my networks :frowning:(

I’ll try to downgrade to 6.32.2 this week.

11

@Ghaith93 and @rpra:

Symptoms you describe are typical from misconfigured PPPoE concentrators.

Trying to make Mikrotik to support you using an old ROS release will lead you nowhere, that’s why Mikrotik continuously upgrades and fixes ROS.

Unless you post exports of your config, it will be impossible to help you.

12

You are not right.
I’ve checked all the problems from https://mum.mikrotik.com/presentations/US17/presentation_4241_1496042977.pdf
Topic starter for example, do not use any dynamic routing at all, but still have issues.
Mikrotik team have seen my config in the ticket, but couldn’t help.

I’ll try downgrading and then report the results.

13

After downgrading to 6.32.2:

Router with 1500 active PPPoE, I’m disconnecting segment with ~500 devices.
I see high CPU load when they are disconnecting, but no packet loss and no OSPF neighbors disconnections.

So there IS a problem after 6.33. Please help me to solve it.

14

I will send config but will see there is a simple config

15

So there IS a problem after 6.33. Please help me to solve it.

You’re not right :smiley:

Having problems after 6.33 with a PPPoE AC doesn’t mean there are problems with pppoe specifically on a given ROS more recent version.

ROS evolves and there have been heaps of significant changes since 6.33 on all ROS areas, two years and a half has passed since then! Which was the more recent previous version you downgrade from?

Warnings to changes that require setup adjustments can be found on the changelog archive; the ROS 6.xx is released! threads may contain useful info too.

I confirm absolutely same problems on my routers.
Heard about same problems even on 1072 with 1k PPPoE sessions.
I have even OSPF neighbors disconnects sometimes.

What that may point to is unstable L2 on your network. Unless there’s pristine L2, you’ll experience PPPoE (and all sort) of related problems and unstability.

Did you watch Janis Megis presentation? because there’s exactly an interaction between non properly setup OSPF on a PPPoE AC that causes exactly what you describe.

I manage CCR1009’s with 500, 800, 1000 PPPoE users, average load rarely exceeds 20-30%. 6.39.2, 6.38.7… I advise using latest bugfix.

Unless you post relevant config, topology, scenario conditions it will be impossible to help you. I asked twice for ROS and Firmware versions and still, you don’t post those details, but demand help? No one has crystall balls here…

Clinging to obsolete ROS versions is not a wise choice. I think the fastest route to fixing your specific problem would be hiring a consultant to look at your live router.