I set up QoS and PCQ user queues recently on the same router (RB433 v4.17). I tested it and it was working without any problems. I didn’t implement it though, because I wanted to allow bursting for individual users and so I had to upgrade to v5.x first.
But after the upgrade to v5.5 my PCQ user queues stopped limiting the traffic. The packets still get marked correctly and they flow through the correct queues in the tree, but their speed is no longer limited at all. i.e. each user has maximum speed.
I have not tested to see if QoS is still working.
What is the difference between PCQ queues on v4.x and v5.x?
Just added functionality. I took PCQ from 4.17 on RB1000s to 5.3 (I think?) on x86 via copy/paste and it worked fine.
Is it possible you added bursting in a fashion that makes it appear as if you’re not policing the rate because everyone is constantly bursting? Can you post your relevant configuration? It is kinda hard to guess at a root cause with only a vague description to go on.
I’ve been at it all day but I can’t get this to work.
I’ve disabled every mangle rule and queue, except for the ones posted above, and I still have the same problem. The only thing I’ve discovered is that the user does not actually have maximum d/l speed, but rather around 3 times the speed of what the queue type’s rate is set at. e.g. When pcq-rate is set at 256k, then user can download at between 600-800k. If I change pcq-rate to 64k, then user download speed will between 150-200k.
I can also confirm that the problem is definitely with the pcq queue type.
I just changed both the limit-at and max-limit values for the “basic_download” queue, to 256k. I tested again and found that it limited the user correctly. e.g.
maybe this is also tied to the 5.x versions showing incorrect totals for interfaces when queried via snmp. I have one router than shows 1/2 of the total, and another than shows about 1/4 of the total…