PCQ issue

itfutura · Wed Oct 15, 2008 8:54 pm

I use RB433AH as a main router. I use PCQ to shape speed of my users. I use address-lists.

example of address-list:

 9   abo4096                                    10.0.15.1                      
10   abo512                                     10.0.15.2                      
11   abo512                                     10.0.15.3                      
145   abo192                                     10.0.19.3                    
146   abo192                                     10.0.19.4                    
147   abo256                                     10.0.20.1                    
148   abo256                                     10.0.20.2

example of mangle:

 6   ;;; Abo 128-up
     chain=prerouting action=mark-packet new-packet-mark=abo128-up passthrough=no src-address-list=abo128 in-interface=eth2-DMZ 

 7   ;;; Abo 128-down
     chain=postrouting action=mark-packet new-packet-mark=abo128-down passthrough=no dst-address-list=abo128 out-interface=eth2-DMZ 

 8   ;;; Abo 192-up
     chain=prerouting action=mark-packet new-packet-mark=abo192-up passthrough=no src-address-list=abo192 in-interface=eth2-DMZ 

 9   ;;; Abo 192-down
     chain=postrouting action=mark-packet new-packet-mark=abo192-down passthrough=no dst-address-list=abo192 out-interface=eth2-DMZ

example of queue types:

 7 name="p128d" kind=pcq pcq-rate=128000 pcq-limit=25 pcq-classifier=dst-address pcq-total-limit=5000 

 8 name="p128u" kind=pcq pcq-rate=128000 pcq-limit=30 pcq-classifier=src-address pcq-total-limit=12000 

 9 name="p192d" kind=pcq pcq-rate=192000 pcq-limit=40 pcq-classifier=dst-address pcq-total-limit=8000 

10 name="p2048d" kind=pcq pcq-rate=2048000 pcq-limit=130 pcq-classifier=dst-address pcq-total-limit=39000 

11 name="p256d" kind=pcq pcq-rate=256000 pcq-limit=40 pcq-classifier=dst-address pcq-total-limit=8000

and all queues are in queue tree:

 0   name="!Upload" parent=global-in packet-mark="" limit-at=0 queue=sfq priority=8 max-limit=9600000 burst-limit=0 burst-threshold=0 burst-time=0s 

 1   name="128up" parent=!Upload packet-mark=abo128-up limit-at=0 queue=p64u priority=1 max-limit=0 burst-limit=0 burst-threshold=0 burst-time=0s 

 2   name="192up" parent=!Upload packet-mark=abo192-up limit-at=0 queue=p64u priority=2 max-limit=0 burst-limit=0 burst-threshold=0 burst-time=0s 

 3   name="256up" parent=!Upload packet-mark=abo256-up limit-at=0 queue=p128u priority=3 max-limit=0 burst-limit=0 burst-threshold=0 burst-time=0s 

 4   name="384up" parent=!Upload packet-mark=abo384-up limit-at=0 queue=p128u priority=4 max-limit=0 burst-limit=0 burst-threshold=0 burst-time=0s 

 5   name="512up" parent=!Upload packet-mark=abo512-up limit-at=0 queue=p128u priority=5 max-limit=0 burst-limit=0 burst-threshold=0 burst-time=0s 

 6   name="768up" parent=!Upload packet-mark=abo768-up limit-at=0 queue=p128u priority=6 max-limit=0 burst-limit=0 burst-threshold=0 burst-time=0s 

 7   name="1024up" parent=!Upload packet-mark=abo1024-up limit-at=0 queue=p256u priority=7 max-limit=0 burst-limit=0 burst-threshold=0 burst-time=0s 

 8   name="2048up" parent=!Upload packet-mark=abo2048-up limit-at=0 queue=p256u priority=8 max-limit=0 burst-limit=0 burst-threshold=0 burst-time=0s 

 9   name="4096up" parent=!Upload packet-mark=abo4096-up limit-at=0 queue=p256u priority=8 max-limit=0 burst-limit=0 burst-threshold=0 burst-time=0s 

10   name="!Download" parent=eth2-DMZ packet-mark="" limit-at=0 queue=sfq priority=8 max-limit=9600000 burst-limit=0 burst-threshold=0 burst-time=0s 

11   name="128dn" parent=!Download packet-mark=abo128-down limit-at=0 queue=p128d priority=1 max-limit=0 burst-limit=0 burst-threshold=0 burst-time=0s 

12   name="192dn" parent=!Download packet-mark=abo192-down limit-at=0 queue=p192d priority=2 max-limit=0 burst-limit=0 burst-threshold=0 burst-time=0s 

13   name="256dn" parent=!Download packet-mark=abo256-down limit-at=0 queue=p256d priority=3 max-limit=0 burst-limit=0 burst-threshold=0 burst-time=0s 

14   name="384dn" parent=!Download packet-mark=abo384-down limit-at=0 queue=p384d priority=4 max-limit=0 burst-limit=0 burst-threshold=0 burst-time=0s 

15   name="512dn" parent=!Download packet-mark=abo512-down limit-at=0 queue=p512d priority=5 max-limit=0 burst-limit=0 burst-threshold=0 burst-time=0s 

16   name="768dn" parent=!Download packet-mark=abo768-down limit-at=0 queue=p768d priority=6 max-limit=0 burst-limit=0 burst-threshold=0 burst-time=0s 

17   name="1024dn" parent=!Download packet-mark=abo1024-down limit-at=0 queue=p1024d priority=7 max-limit=0 burst-limit=0 burst-threshold=0 burst-time=0s 

18   name="2048dn" parent=!Download packet-mark=abo2048-down limit-at=0 queue=p2048d priority=7 max-limit=0 burst-limit=0 burst-threshold=0 burst-time=0s 

19   name="4096dn" parent=!Download packet-mark=abo4096-down limit-at=0 queue=p4096d priority=8 max-limit=0 burst-limit=0 burst-threshold=0 burst-time=0s

And now about problem

From time to time I cannot define how often, PCQ queues freeze. Not all of them but those one without traffic.
For example I have small amount of users who has 128k UP - 64 Down. So PCQ quueue is not in use. When the user with this speed assigned, wants to get Internet traffic - cannot. When I try to ping your IP from RB - I cannot. When I try to ping the main route from your CPE - I cannot.

Solution - disable all queues including parents and enable it again. Voila it works. I can ping CPE I can ping main router from CPE.

My question is why I have to do that. Why PCQ queues freeze so often. Why can I do to repair my situation. Maybe it is a MT bug?

Chupaka · Fri Oct 17, 2008 2:38 am

what version do you use? as I remember, I had this kind of problem in about v3.7

itfutura · Fri Oct 17, 2008 9:54 am

I use 3.14.

Support, are any changes in ROS 3.15 in PCQ?

MiruK · Sun Oct 19, 2008 1:35 pm

I've the same problem on PC machine

itfutura · Sun Oct 19, 2008 5:23 pm

True. On PC I Have had the same. Then I have planned to change to RB. And... Nothing has changed. On RB PCQ queues still freeze.

Support HELP!

I see I'm not alone

When this will happen again I will send to support@mt suppout file. Will see.

hurley12345 · Mon Oct 20, 2008 12:43 pm

Hi, Running 3.14 and have the same issue with clients freezing and unable to ping. Disable and re-enable the queue and works again. Have been working with Mikrotik on this issue for quite a while and I am getting little or or feedback from them. I sent them a support.rif file while it was happening on 3.14 and heard nothing back from them. Its quite frustrating to have an issue and Mikrotik ignoring the issue.
E-mailed support on this issue on 6/6/08 "[Ticket#2008060666000537] Major Bug with queue" Have sent them numerous support files.
Just about to click send on this message forum and Mikrotik sent me an e-mail saying that if it is still happening on 3.15 I should send them login details of the AP.

Brendan

Danio · Mon Oct 20, 2008 3:45 pm

I have exactly the same problem.
I'm testing now 3.15 ... but it seems to have some addictional bugs.

Regards...

itfutura · Mon Oct 20, 2008 3:52 pm

Support! Are Ypu there? Why don't You do anything? Even simple answer about it?

MiruK · Mon Oct 20, 2008 5:50 pm

v3.15 seems to be OK, but I'm still testing

itfutura · Thu Oct 23, 2008 12:41 am

RB 433AH uptime: 5d8h39s and ...

I have no problems with PCQ so far

. I have turned off syn-cookies too. I don't know is any interrelation between PCQ and syn-cookies, but maybe .... MT Support can You confirm od deny between these two things?

rarbolay · Fri Oct 24, 2008 10:41 pm

itfutura, what did you do diferently? You stated that you were having issues and now say that you have have been running for X number of days without problems. What is the change?

itfutura · Sat Oct 25, 2008 12:03 am

When I started this post there was ROS version 3.14. Currently is 3.15. The change is ROS version

mietus · Sun Oct 26, 2008 9:47 pm

got the exacatly same problem on 3.15, after 6 days of working with no problem today it's hanged out

turn on and off these queue and working good

hurley12345 · Mon Nov 03, 2008 8:05 pm

Anyone got any update on this problem ?

Equis · Mon Nov 03, 2008 10:07 pm

I am also having this problem.
Has anyone got a fix?

Thanks

Equis · Mon Nov 03, 2008 10:46 pm

Are you running Masq/Nat on the routers with problems?

Equis · Mon Nov 03, 2008 10:50 pm

I wonder if this is a connection tracking problem.

itfutura · Mon Nov 03, 2008 11:11 pm

Yes, I use masq on the router. ROS 3.15 hasn't changed the issue

mietus · Wed Nov 05, 2008 1:27 pm

i don't have connection tracking and i still got the problem..
ros 3.16 is out maybe in this version it will be fixed

MiruK · Thu Nov 06, 2008 10:45 pm

v3.15 and I have still the same problem, from time to time PCQ queues freeze !!! Not all of them but those one without traffic:(
For me the last (PCQ) working version was v3.2 ( v2.9.xx works OK too)

SUPPORT do something or explain this problem, v3.3 - v3.15 have a PCQ bug !!!

....I found similar post: http://forum.mikrotik.com/viewtopic.php?f=2&t=27618 Support didn't answer

What is going on ???? It's a BIG bug !!!

itfutura · Fri Nov 07, 2008 1:52 am

Not all of them but those one without traffic:(

EXACTLY !!!
I have same PCQ behavior!

Equis · Fri Nov 07, 2008 2:00 am

Has anyone tried .16?

itfutura · Fri Nov 07, 2008 2:34 am

currently testing. also wrote a simple script to stimulate PCQ queues by ping.

EDIT: wrong idea ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

mietus · Tue Nov 18, 2008 1:04 pm

any ideas? workarounds?

hurley12345 · Thu Nov 20, 2008 2:00 pm

Can Mikrotik support update us as to what is happening with this issue ?

Thu Nov 20, 2008 3:03 pm

What MT support have to do with this forum??? For support you must write to support@mikrotik.com - Normis is saying that for years.

In here we can discuss our problems and possible solutions, try to find similarities in the problems and then all together report very specific bug so that they will be able to fix it with less problems.

BTW: I just got beta (untested, use on your own risk) version with address-list and PCQ fixes... so far so good.

itfutura · Thu Nov 20, 2008 3:08 pm

Me too. I got 3.17-beta version from support. After couple of e-mails with support and suppout files I got version that seems to work good. Probably there was a bug with address-lists.

mietus · Mon Nov 24, 2008 10:34 am

also got the 3.17 from support but the problem still exists

Mon Nov 24, 2008 10:41 am

also got the 3.17 from support but the problem still exists

I had 2 different versions, at first they give me one after few days other.

It looks like latest solved my problem

mietus · Mon Nov 24, 2008 10:51 am

also got two versions but the problem persists

could u post your mangle qtree and address list?

Mon Nov 24, 2008 11:58 am

IT will be to much

1) mangle chain forward marks by address-lists

2) queue tree on outgoing interfaces with PCQ type

for ~800 IPs on 4 different PCs

cartes · Tue Nov 25, 2008 9:06 am

Hi,

We also have a similar problem, but not exactly the same. On a random basis, some users would get no ping replies from this MT server. And, usually, these users would belong to a handful of subnets, say around 5 - 10 subnets would occassionally have users complaining.

To explain our network, we currently (still growing) have 4000+ users, divided over 80+ subnets of /24 each (division made based on region, speed limits etc). All subnets are connected to a dedicated MT RouterBoad ehernet port, so all are routed on Layer3. All the remote RBs are connected to the backbone and are using a central RouterOS as their GW. This ROS is the main BW Manager, the one where PCQ is being used, and it is running on a Core 2 Quad 2.33 GHz and 2 GB or RAM (MT using multi-cpu enabled). We've got this back to 3.7 as upgrading to 3.15 (first 3.13 then 3.14) gave stability issues where users would get low speeds which would match the data on the remote MT RBs, but the ROS would show full use of assigned PCQ limits. Even the x-fer numbers on the Interfaces would vary greatly than the numbers on the queues.

Some MT guy (probably normis) said a few weeks ago it was probably a clocking issue and MT is working on it. Going through this post, it seems 3.16 still has this issues.

Please help MT.

Regards
HASSAN

mietus · Sun Nov 30, 2008 11:36 pm

ok ;> i;ve got third version from mt and this one it's fully working with my pcq config

omidkosari · Mon Dec 01, 2008 1:37 pm

anyone has this problem with pppoe ? if yes , does disconnecting and connecting again solve the problem ?

Mon Dec 01, 2008 1:40 pm

anyone has this problem with pppoe ? if yes , does disconnecting and connecting again solve the problem ?

do you use PCQ with your PPPoE? if yes - it helps.

omidkosari · Mon Dec 01, 2008 1:47 pm

Yes for shaping them . but when they connect and they have no data transfer , right after disconnection the unregister_netdevice occurs and router should reboot or reset

cartes · Mon Dec 01, 2008 3:12 pm

In our case we're not doing PPPoE. We are just doing traffic shaping using PCQ. When the user gets this error, we simply disable and re-enable the simple queue. That solves the problem.

omidkosari · Mon Dec 01, 2008 9:06 pm

I think this issue only happens when the simple queue or queue tree does not have max-limit or limit-at value and queues with max-limit should not have this problem.

cartes · Mon Dec 01, 2008 9:33 pm

Thank you for this tip! Will try it out definitely and let you know the results.

On this note, which value do you suggest we try? max-limit? Or limit-at?

Tue Dec 02, 2008 8:56 am

I think this issue only happens when the simple queue or queue tree does not have max-limit or limit-at value and queues with max-limit should not have this problem.

AFAIK - you are wrong. i had both those option when the problem was there.

One way or another it is corrected in the 3.17.

omidkosari · Tue Aug 23, 2011 8:22 am

I know it old topic but anyone has this problem on ros 5.x ?
my router started to freeze frequently . pppoe users disconnected , snmp crashes, winbox freezes and disconnect , radius clients crash , ......

mikrotik supports did not detect the problem . i did some research and remembered this bug from old days . while changing the queue type from PCQ to DEFAULT everything works fine .

i am using Mikrotik-Address-List radius attribute to create dynamic address list for users .

omidkosari · Wed Aug 24, 2011 7:55 pm

Unfortunately support answer and even bringing router access to them does not help to solve the problem .
I wonder that just executing following line makes crash my router

/queue tree set [find where name="Users down"] queue="PCQ Download"

and doing another one solves the problem

/queue tree set [find where name="Users down"] queue="default"

and nobody understands the problem !!

Who is online