PPPOE and random queue missing problem

RouterOS General
1

Hi Everyone!

We have a PPPOE server with 500-600 users. Sometimes the users queue won’t work, users have a 512k/128k but they can download 2-5 MB, if we close all the session after one-two days queue problem renews. - we use MT 3.7 , 3.9 and now I hope 3.10 will solve this problem…Any other suggestion? MT 3.3 works great? Or what is the last version that good for PPPOE server?

Thanks in advice!

Josh

Simple Queue on Intel PC 3.14 problem
2

Yesterday we upgrade to 3.10 but we have this the problem today, SO??? Any suggestion???

3

Hi,

Do you use ‘change TCP MSS’ feature?
What kind of queues do you use?

Regards
Lutz

4

i saw the same on 3.7

whe mangle the tcp mss and use default queu configured as pfifo with 100 paclet queue.

Regards
Ros

5

whe mangle the tcp mss

What do you mean here? Static Rule for all, or automaticly Rules per PPPoE-Login?

Regards
Lutz

6

static rule for all the subnets inside the ippool used by pppoe server

Regards
Rosario

7

I have the same problem. I see the queue simple without counting upload or download. If this happend i do an script to disable and enable this queue, after that the queue work fine.


Max
http://mikrotikexpert.com
http://maxid.com.ar

8

Any news about this problem?

I know a lot of people having it (including me) but no news about a fix.

9

Hi,

first off all make shure that you don’t use the automatic change MSS feature. You can use 2 static rules for all instead of 2 dynamic rules per PPPoE-User for that feature. Save 1.198 rules by 600 connected PPPoE-User.

Second, don’t use the rate limit per PPPoE-User or Profile, this also generate 2 simple queues per loged in PPPoE-User. Use PCQ-Queues instead against adress-lists or IP-Pools. This also save hundreds of simple-queues.

You maybe also use a little bit higher queue sizes (default = 10 if I remember correctly) try 100 here.

If you done this all, you should be able to run some more PPPoE-Clients without having that problem. But still remember also that the PPP-Stack in linux is not optimal for use as access concentrator. We have a system which is able to terminate > 50.000 concurrent sessions, but this is not linux based.


Regards
Lutz

10

Hi, I already made some tests on it disabling automatic MSS and use only two static rules. No performance change (maybe I don´t have too much users for that).

Use PCQ instead of dynamic rules is out of question. I lost too much control doing that and I have clients with too many different plans and speeds (a lot of clients have specific bandwidth). This is not a solution.

I don´t have too much online clients per MK router.

I my point of view it´s a primary MK feature (shape bandwidth for online clients) and it must work.



11

No news? MK support say it´s a bug but no more information if it will be fixed or when.

Continue using MK to control my users bandwidth is becoming useless, so I need to look for a professional solution. Anybody here hows a good and professional PPPoE Server ?

12

Hi,

I don’t belive that is really a bug. Linux is not suitable for terminating hundreds of PPP-Users. Regardless if it’s L2TP or PPPoE. Maybe they have some small problems, so you run a little bit earlier in that problem. But over all, if you need to terminate 500+ Users, I prefer non Linux based solutions. As I figured out, we have a system which terminate 50.000+ PPP-Sessions without any trouble. No it’s not CISCO or Juniper and their is no Linux PPP-Stack used.


Regards
Lutz

13

I think the problem is not related with PPPoE code, but with dynamic queue and PPPoE interface. It happens with much less than 500 online users (500 users is the max I have per MK router).

What solution you have? I need a better solution than Mikrotik (I need something that works).

14

We have our own system, complete developed by our technologie partner in Germany and tested and in production by very huge ISPs here since the last years. The system is based on a 16 core Cavium CPU with complete own routing software for OSPF/BGP4 (fully featured routing policies and IPv6), PPPoE & L2TP in and out, L2TP Tunnel-Switch, loadbalancing across several DSL-Uplinks. Tested with a load from > 50.000 L2TP- or PPPoE-Sessions. And lots more.

For more information please contact me per email directly.


Regards
Lutz

15

Hi everybody…

I am having this same problem. It used to happen with 2.9.x, and is happening with 3.x (we actually run 3.9 on 10 different Mikrotiks).

I have read the previous messages and to my surprise the only useful messages are talking about “moving to other solutions”. What Mikrotik support team thing about this ???

I really need a solution. My core Mikrotik usage is as PPPoE concentrator, so this problem is SERIOUS…

Thanks in advance.

Denio A. Lindo
Desktop Internet Services

16

I am here too.

I started to experience this issue something later then 2.9.3x.

The early 3.x seems not affected but now the problem is there too.

Support wrote me to pass to new feature present on 3.10 to use pcq…

Strange that all the major MT issues were about queue.
I think they should improve the area of the code.

Regards
Ros

17

They should be kidding that use PCQ is the solution. As I can see MK is not a serious solution.

18

Hello,

I think MT is a serious solution, but sometimes restricted to the Linux. If you need more in some special situations, e. g. a PPPoE-Server for more than roundabout 500 concurrent user, you may take a specialized solution for this topic, and not a full featured RouterOS System.

Regards
Lutz

19

Have been seeing this issue since 2.9.x as well. Would really like a solution. Do not really want to use PCQ for same reasons someone else already mentioned. Do not want to have to change a users IP range when they change rate plans.

We really need this issue fixed.

Matt

20

I have the same problem. I see the queue simple without counting upload or download. If this happend i do an script to disable and enable this queue, after that the queue work fine.

Can you share this script?

Matt