Hi.
sometimes (about 1 or 2 in a day) the winbox freezes and when i
disconnect it and again connect every windows of winbox is empty without
any data.
I tried ssh or telnet also . in the ssh or telnet print commands or any
command like that (which output some data) does not work and sometimes

action timed out - try again, if error continues contact MikroTik support and send a supout file (13)

this problem will be solved with a reboot
one time i wait to see what happen if i don't reboot . after about 1
hour the problem automatically solved ! but at that 1 hour any new user
could not connect with pppoe pptp etc .

Again problem occured about 10 hour ago. now i pluged monitor to it and see this message in all of the screen

unregister_netdevice: waiting for ppp61 to become free. Usage count = 51

i have contacted the support with the supout and they answer this nothing more

Unfortunately supout does not contain any information about crash. Disable all
unused packages like mpls-test, wireless, stpbridge-legacy etc. Write this
command in console:
/system hardware multi-cpu=no
and reboot router, it will load regular linux kernel.

If it is possible connect monitor and next time when router will crash it will
output info on a screen.

at the problem time my mikrotik goes in stage like this post http://forum.mikrotik.com/viewtopic.php ... 99#p101599 but the different is i have about 200 users active and cpu usage is about 7-10 % and everything was fine with this setting in 2.9.27

This is exactly what I mean FREEZE while using pppoe server with versions of ROS 3.5 and above. ROS 3.4 and below this is ok for nothing more than 100 simultaneous users and you may need ROS 2.9.45 if you think about nothing more than 750 users. But unfortunately, they said the same to us. We are still researching to know what exactly is happening and so far seems to be a random problem of instability envolving pppoe server only.

the support team have sent a new answer to me . just to share it with community i will write it here

Currently it looks like linux kernel bug, we will check it and inform you.

as i searched before the "unregister_netdevice: waiting for ppp61 to become free. Usage count = 51" is not the mikrotik problem and this problem have seen in other linux distros .

Hi . i found something about this problem and thought it is better to say.
i have radius server.
i disabled my radius server about 2 hours for some maintenance and about 100 pppoe users tried to connect via pppoe in mikrotik and mikrotik worked fine. i was waiting to see if mikrotik crashes i make a supout but mikrotik worked without any problem.

BUT exactly when i enabled my radius server after 5 or 10 seconds some users connected and then mikrotik again said

unregister_netdevice: waiting for ppp53 to become free. Usage count = 51

now i am sure that is problem occures when radius server answers to mikrotik and lot of ( approximately 15 ) pppoe users trying to connect at the same time.

another thing , Usage count = is always 51

What RADIUS server are you using?

i am using an accounting software which uses PYRAD as radius server.

We have the same problem view the topic http://forum.mikrotik.com/viewtopic.php?f=2&t=26079
But I use freeradius server. I do not think the problem is attached with radius server as yesterday I disabled it and my users where authenticated via local, but the router continues to crash no difference.

Regards
Karapet Aznavuryan

here is some of mikrotik support replies

Hello,

Currently it looks like linux kernel bug, we will check it and inform you.

Regards,
Maris

Hello,

Well, this error message is coming directly from the Linux Kernel and that
makes things complicated - we need to relay on Linux Kernel developers for the
fix.

I suggest to play with SMP setting try to enable/disable it. Also a faster
RADIUS server might solve the problem

Regards,
Janis M.

Do a /system reset-configuration
Then reconfigure manually.

Do a /system reset-configuration
Then reconfigure manually.

can you explain more please . i mean the reason of your advice . it is nearly impossible to reconfigure all of configuration manually AGAIN from first step. all of users will disconnected until config finish .

Do a /system reset-configuration
Then reconfigure manually.

I have done.
All the configuration I have done step-by-step for 2 days.
But there is no use.

Regards
Karapet Aznavuryan

Hello,


As I mentioned previously it is most likely kernel problem. Next version will
be with updated kernel so it might solve problem you are facing.


Regards,
Maris

i received this message about 10 days ago . now the question is

When the 3.14 will be released and my damn problem will be solved ?

i am under the heavy pressure from users and my boss. please understand.

interesting . mikrotik admins are very active !!!! in this topic. thanks for good answers , i am now relax because of your answers.

still me and somebody else have this problem and there is no solution from mikrotik guys.

please wait for the suggested version upgrade. it's still in testing. you don't want a rushed upgrade that doesn't fix your issue, do you?

hey mikrotik guys . I see in another post that you said the 3.14 will be released in next week . does this problem solved completely in 3.14 ? please say details also . thanks

i am waiting for a reply . it is very important .

Well all of us are looking to solve the same problem.
I tried everything that you say and nothing. Change hardware, reconfigured, change version of RouterOS, and the same.
As you said, I'm using a FreeRadius Server
Well we hope that its problem will be resolved in v 3.14
Good Luck MKT Team
Rafael Lore

Anyone know when this started? v3.? I'm thinking v3.10 + ?

It is difficult to be trying with clients working, but in turn it is the best way of having betatester forced jeje. Honestly not from when the problem is generated, but I suspect that it is from v3.0. I hope that in the close V3.14. the problem is solved. We all lean the team mt, but these they must give priority to the topic, in my case I am a Latin-American client with mas of 100 licenses you pay. We all expect anxious and desperate.

Regards

Anyone know when this started? v3.? I'm thinking v3.10 + ?

i dont think so . this is a linux kernel bug and since the version 3.10 to 3.13 use same kernel , so this problem existed in 3.x stable versions.

hey mikrotik guys . I see in another post that you said the 3.14 will be released in next week . does this problem solved completely in 3.14 ? please say details also . thanks

i am waiting for a reply . it is very important .

normis . still i am waiting for your reply

I talk with MTsupport already 2-3 month, but they do not see any problems ,

We have exactly same problems !

we have about 30 v2.9.x in network with pppoe, and work without any problems over 2-3 years,
and we have 2 v3.x with pppoe , and from all version of v3.x (we start with 3.6) problem are same,
but we have 6-7 v.3.x without pppoe and work normal !

We can not buy any more MT routerboard, because we can't install v2.9.51, ...
I must change this two v.3.x, because users complaints ...

I do not know what to say anymore about v.3.x

... give us 3.14 with solution of this problems

RouterOS v3.14rc1 released but nothing about this damn problem in changelog

I believe that for the present there is no any change. Momentarily i will use fixed directions in the problematic clients with private key even so much one finds the solution.

Regards

Unfortunately the problem still exist with no change in 3.14rc1 .
still this damn string shown

The freeze still there on ROS 3.14rc1... Oh my...

As it looks like it v3.14rc1 walks worse than v3.13. PPPoE still freeze. I wait that the
mt team solves the problem for the close delivery.

regards

Currently it looks like linux kernel bug, we will check it and inform you.

Sorry, but if this bug is on the linux kernel i think that the mikrotik team will not be able to fix it. They can only hope that the kernel maintainers fix it first.

Currently it looks like linux kernel bug, we will check it and inform you.

Sorry, but if this bug is on the linux kernel i think that the mikrotik team will not be able to fix it. They can only hope that the kernel maintainers fix it first.

i think this problem for kernel developers is not as important as for mikrotik developers . mikrotik developers should try more and solve this problem before think about adding new features

The work priorities is unrelated between linux kernel developers and routeros. Yes, mikrotik should to try another solution, but if the problem is really on the kernel the fix is more complicated.

it seems one of pppoe problems solved http://forum.mikrotik.com/viewtopic.php ... 42#p129242 . i hope this freeze problem solve also . i am tired of refreshing forum and checking my email to see solution .

Yes, how I see the problem stays, there was no solution.
I hope it will be fixed later.
I have upgraded My 2.9.51 to 3.14 after, but the PPPOE DIE problem was going on.
I had to downgrade to 2.9.51, in order do not have problems with my clients.
When it will be fixed? Who can tell us?

When it will be fixed? Who can tell us?

Guess it will be when MT do change the kernel. Googling around you may find that Fedora Core solved the "unregistered_netdevice" problem while other distros not. It is a bad bug and also it is very complicated for them to solve. But youre right, who knows? About 95% of our MT deployment are small pppoe servers to handle about 50 clients each and even ROS 3.4 that do less freezes (but some times it do) is a pain for our customers. Unfortunately, this minor amount of thousand people complaining about it does not take precedence on their priorities to solve problems. Who do even knows if it is on their "to do list"? Lets just hope and pray for this day to come asap...

Who do even knows if it is on their "to do list"?

this make us more nervous and they did not reply here

Googling around you may find that Fedora Core solved the "unregistered_netdevice" problem while other distros not

i heard this somewhere else . maybe it is better to write some of those links here for mikrotik developers .

in 3.14 the unregister_netdevice problem occures less (about 1 time per day) but still happens and still 3.14 is unstable

I had a similar problem here and, as a last desperate resource, exchanged hardware for a more powerful x86 platform (from Athlon64 to C2Duo) and the problem just disappeared. I just couldn't believe since the Athlon was running only about 20% of CPU, now the C2Duo is using SMP and running about 1-5% of CPU... Maybe it's a race condition in the kernel, who knows...

are you sure ? when you changed your cpu ? how long your mikrotik doesn not crashed ?

5 days, no crash.

But this is only my experience, YMMV. I don't know if everybody that is having this issue have a similar hardware setup.

Since RouterOS is bound to the disk only, worth a try...

I tried your suggest with Core 2 Quad . with 4 cores now the cpu is 0-5 % ( most 0% ) but the unregister_netdevice occures MORE . i am confused and disappointed completely

core 2, core 4, core 5000 ...
what about mikrotik RB433, mikrotik BR600, mikrotik RB 44444 ... etc...

that's is not solution of problem.

I tried your suggest with Core 2 Quad . with 4 cores now the cpu is 0-5 % ( most 0% ) but the unregister_netdevice occures MORE . i am confused and disappointed completely

Sorry to "hear" that.

Really I'm very unhappy with MKT, because there are a lot of people with this problem and we don't have any solution or something saying that they are working in its!!
At the moment I'm working with this problem in an extremelly unstable network, and I don't know what to say to my client.
Rafael Lore

I have gave mikrotik support team a FULL ACCESS at 30 september . no progress yet .

oh my god, almost two days, and still ... nothing !

please hold on, development takes time.

I am a developer also . you are not the only developers in the world
it is not just two days in this stuff . i think this problem started from alpha versions of 3.x . i said no progress because the dev team does not even installed the debug packages . does installing debug package want 2 DAYS or more ?

Good news . I have found the conditions which causes the unregister_netdevice and i have sent my researches to support . Now at least we can prevent those conditions to prevent problem occurs .

any short details? =)

I have found that this problem JUST and EXACTLY occures when few known users want to disconnect their pppoe connection.
for example i have a user with username john .
every times john wants to disconnect his pppoe connection -> the disconnect signal sends to mikrotik -> mikrotik receives the disconnect packet -> mikrotik is going to disconnect john but it can not -> UNREGISTER_NETDEVICE occures

now there are 2 problems
1. Why mikrotik can not disconnect these type of users .
2. Why a few users make this problem

we can not do anything to solve the problem 1 and it is something that linux kernel developers and mikrotik developers should try to solve.

but about 2 . temporary we can disable this few users and enable one of them ( to reducing the problem time in a day ) then sniff that users packets when he wants to disconnect and find what type of packets come from that user when he wants to disconnect. ( now i am actively doing this work )

note: if you want to know these problematic users , you should find the last disconnected user before UNREGISTER_NETDEVICE . this can be done simply with mikrotik logging.

Congratulations omidkosari, I am grateful for you very much for your investigation, it is of many help for all of us

omidkosari, please contact MikroTik support (support@mikrotik.com). Send us logs and support output file from the router.

I already doing this for several months . and you have FULL access to my router.

is there any change in 3.16 ?

Found that our "freezes" within PPPOE server stoped, even with ROS 3.16. We changed our setup concerning pppoe while using mangle and queue tree with PCQ to just dynamic simple queues. Seems that this freeze stuff happens while using pppoe together with PCQ/mangle/queue tree. PPPOE with dynamic simple queues does not shown any freeze. Try to remove all (if any) mangle/pcq/queue tree if you have this problem of freezes with PPPOE. If you dont have pppoe+mangle+pcq+queue tree and you got freezes, then it might be another issue.

Thanks
Ozelo

I found another new item related to unregister_netdevice .
I found a user which was in "Active Connections" but was not in "Interface" . I have watched that user and right after he wants to disconnect , unregister_netdevice occured .
I think the problem happens when a user connects and appears in "Active Connections" but Mikrotik does not allocate an interface to him so the user can not send or receive traffic and he should disconnect and connect again and at this time we have unregister_netdevice.

omidkosari,
did mt-team, try something with your router ?

did you used pool name in pppoe sever seting?
because, we have pool , but blank seting for pool name,
and attribut send pppoe in package, and mt find pool in pool list,
maybe this is problem in v3 , for some connections ?


btw, sergej,
what happend with 100 license for 100 routers (only work for v2.9 and v3), when you make v4, ... but we can not used v3 ever ?!
maybe we can not be used v4, also.

omidkosari,
did mt-team, try something with your router ?

did you used pool name in pppoe sever seting?
because, we have pool , but blank seting for pool name,
and attribut send pppoe in package, and mt find pool in pool list,
maybe this is problem in v3 , for some connections ?

They have FULL permission on my router but nothing done by them .

No ihave not pool name in my pppoe server . i am using radius server .

And just to share with others who have same problem this is last mail i received from support . unfortunately it is

Hello,


RouterOS can't set Active Connections if interface is not created. Most likely
interface disappeared after pppoe crash. Unfortunately we were not able to get
anything useful from supouts, yet. So problem is still unsolved.


Regards,
Maris

omidkosari,

we can not testing anymore, because we do downgrade to 2.9.51, on all routers....

but I look something and maybe this can help ...

users ap router "TP-Link", passtrought MAC from comp to mikrotik AP, and MT this MAC see like unkown_device ...
and this have happend on 2-3 sec , all time.

btw, TPLink router are set in apclient mode, and all work, router MAC are on mikrotik and router have pppoe connect, and users comp have 192.168.1.xxx ...

maybe this unknown_device are problems, can you test this ?

any news on the topic ? problem resolved ?

To mikrotik team . Any news about this problem ? The unsolved problems of mikrotik grows . please make a response and share the progress information . silence is not the solution and we don't forget the problem after some days because it occurs many times a day and we can just shame on users shouts .

omidkosari,

did you test only cable customer ...

maybe problem are only on wifi card (accesspoint), not on ethernet card.

can you test this ?

btw, you see this problem for some users, is this users wireless or cable user ?

They are not wifi or any kind of wireless . they are ADSL customers and connected with bridged Modem/Router which is a kind of cable.

To Mikrotik team , at least say the bugzilla or ticket id of this bug at kernel.org where you have pursue this bug.

problem is in Profile Remote Address seting in PPPoE, when this are empty, mikrotik have some bug with make interface.

sergejs, test this, and tell us what you think

I found this pages but all says that the problem solved in the recent kernel
http://bugzilla.kernel.org/show_bug.cgi?id=6197
http://bugzilla.kernel.org/show_bug.cgi?id=9778
https://bugzilla.redhat.com/show_bug.cgi?id=169502

problem is in Profile Remote Address seting in PPPoE, when this are empty, mikrotik have some bug with make interface.

sergejs, test this, and tell us what you think

And do you expect? Tunnel must have local-address and remote-address, you need to specify it ether in the tunnel or in the secred.

radius send attribut for this , ...
what if you have 5 pool, for 5 diferent groups , etc...

this work in 2.9.x, with no problems !

Good news . I received good results from my researches .
Now i have a temporary solution other than resetting (or rebooting) the routeros . The real problem causes by PCQ issue and when user want to disconnect the pppoe and connect again , unregister_netdevice occurs.
The OLD way to solve the unregister_netdevice was reboot but the NEW way is disable and enable all queues (simple queues or queue trees) and then disable and enable radius (if using radius)

We can do this job with a simple script but now the question is this . How we can determine that when unregister_netdevice occurs ? i mean determine unregister_netdevice by script , not by seeing the monitor or winbox . if a script can understand this problem right when the problem occured , it can do the job and problem will be solve .

Please suggest your good ideas.

@ omidkosair,

is there resolved problem in 3.17 ?

As i said before this problem caused by PCQ Issue and since this problem solved in 3.17 so i think the damn unregister_netdevice solved but mikrotik does not understand it

Anyway the problem solved .

I invite you to my new topic about Mikrotik PPPoE Timeout bug and share our experiences to solve another bad problem.
Thanks

omidkosari

there are too many cases when unregister_netdevice messages could appear. If you see this message all the time when something isn't working it doesn't mean that you faced the same problem again.

there are too many cases when unregister_netdevice messages could appear

I see you know many ways so why didn't post here before i say it solved ? you will help others just after the problem solved? You know what can be happen because of such problems in production ? how many shouts i heard from customers and i just could smile !

If you see this message all the time when something isn't working it doesn't mean that you faced the same problem again.

But in my case it does. I have said every details to mikrotik team.last months i exactly could say why it happened . i am a developer also . with such details if i wrote only a part of a program , i could detect the problem but they just listened and nothing. i was very angry because i did not ask just a new optional feature . the routeros was exactly unusable and very very unstable and mikrotik team was programming for new features. and now i am not enemy of mikrotik . i like mikrotik and i promote it but like should be in both direction. mikrotik also should like customers and respect them and respect their ideas and don't think that customers are idiots .

BTW as i said this problem solved . who knows how can i change this topics icon to SOLVED ?

anyone test the stability of the PPPoE server one v3.17? in the RB's which i have the pppoe server never uses PCQ tails and the problem of freezing is happening (im not probe 3.17 for now).
best regards

chapex,
do you have pool in Remote Address (pppoe seting in mikrtik),
or used redius attributs for poolname ?

i use pool for remote address asign ... (in ppp AAA: use-radius: yes, accounting: yes, interim-update: 0s)

regards

pd: pppoe server works well in 2.9.51

I found another new item related to unregister_netdevice .
I found a user which was in "Active Connections" but was not in "Interface" . I have watched that user and right after he wants to disconnect , unregister_netdevice occured .
I think the problem happens when a user connects and appears in "Active Connections" but Mikrotik does not allocate an interface to him so the user can not send or receive traffic and he should disconnect and connect again and at this time we have unregister_netdevice.

I have this problem but not with just one active connection, but maybe with 80 at one time, they appear in active connections but not in "ppp interface"

3.17 has this freeze crash still.
I had one PCQ for p2p I temporarily disabled.

I use radius so no remote address specified.


I never had this problem with only 64 PPPoE sessions, but now I tunnel in some PPPoE sessions from other routers so with 174 PPPoE sessions it freeze/crash

router doesnt crash.

How many of you have other PPPoE crashes and how many sessions do you have? Like I said no problem with 64 sessions, only problem now with 174 sessions. It goes 1-2 days then crash, before I had uptime of 154 days!

router with 5 connections, freeze too.

to all MT users: Can anyone who have not this problems , tell us what use for setting.

Is there any one who have radius and MT , and do not have problems ?!

Marko, please contact support department about your issue (support@mikrotik.com).

Unfortunately this problem still exists even in version v5.6 http://forum.mikrotik.com/viewtopic.php ... 55#p277155
seems PCQ bug will not solved

Since you seemed to not understand the zombie reference last time I'll elaborate.

While not specifically mentioned in rules it's generally considered bad forum etiquette to bump a very old thread. The proper etiquette is to create a new post with your issue and then include links to older threads as a reference.

An old thread is considered dead and is brought back to life or reanimated. The reanimated dead = zombie. Zombie + cute + humor = zombie kitten

The last post on this thread before yours was Wed Dec 10, 2008 5:57 am, nearly three years ago