FlexQoS FlexQoS 1.2.4 - Flexible QoS Enhancement Script for Adaptive QoS

[Dinesh_kck]

I am having a bit of a weird issue, I upgraded my connection to ~220Mbps and QoS settings keeps reverting to 100 Mb/s
If i choose AutoBW it goes back to Manual setting after some time.
I have Factory reset the router but issue remains same.
Issue appears ASAP i install Flex QoS

Any assistance?

Thank you

 

[dave14305]

I am having a bit of a weird issue, I upgraded my connection to ~220Mbps and QoS settings keeps reverting to 100 Mb/s
If i choose AutoBW it goes back to Manual setting after some time.
I have Factory reset the router but issue remains same.
Issue appears ASAP i install Flex QoS

Any assistance?

Thank you

FlexQoS doesn’t set any of the bandwidth parameters. If you’re running spdMerlin it might be changing that for you.

 

[Kingp1n]

There is another update to the develop branch. Here are the updates currently in develop branch:

• Fix webui interpretation of iptables rules with multiple ports specified in both local and remote fields. More improvements were pushed today. If you use multiple ports (comma-separated or a range specified by colon), please test that the output is how you would expect.
• Improve locking mechanism when mounting the webui page to avoid conflicts and collisions with other addons during startup.

There are likely bugs in the first one still, but it only affects the Application display of a connection in the Tracked connections table.

I was already running the develop script 1.2.4. I ran the update and applied the hotfix. The correct develop script version is still remains at 1.2.4 correct?

Thanks for all your hardwork!

 

[dave14305]

I was already running the develop script 1.2.4. I ran the update and applied the hotfix. The correct develop script version is still remains at 1.2.4 correct?

Thanks for all your hardwork!

Yes, until 1.2.4 gets released to stable.

 

[Dinesh_kck]

FlexQoS doesn’t set any of the bandwidth parameters. If you’re running spdMerlin it might be changing that for you.

Thank you @dave14305 it was a spdMerlin setting.

 

[slidermike]

Dave (or anyone who knows),
I just switched to your excellent implementation of QoS.

Observation / Question
Clients running their own vpn (openvpn clients) (my work laptop to corporate LAN) appear to be in "Other" category unless I am mistaken.
My mind says this should be in the "Work from Home" category. Its quite possible Ive missed this topic already with the 40+ pages in this thread. (playing catch-up can be a bear)

The question I have is this.
Are the routers built in openvpn clients being classified by FlexQoS in the same class as host based client openvpn traffic hitting the router LAN?

Thank you

**EDIT**
I found the answer after posting.
On the FlexQoS page you click the "customize" button.
"Router/VPN Client Outbound Traffic Class" has a drop down to tell which class to put it in.

Having placed that into the "Learn-from-Home" class and applying I am surprised not to see any change in the graph.

I am pumping some 7MBS through the router vpn client 1 and so far the FlexQoS graph still shows the massive data moving in the "Other" class.

Maybe a reboot is needed.
I have to reboot soon anyway because I am working on nat-start variables.

 

[dave14305]

On the FlexQoS page you click the "customize" button.
"Router/VPN Client Outbound Traffic Class" has a drop down to tell which class to put it in.

Having placed that into the "Learn-from-Home" class and applying I am surprised not to see any change in the graph.

I am pumping some 7MBS through the router vpn client 1 and so far the FlexQoS graph still shows the massive data moving in the "Other" class.

Maybe a reboot is needed.

Check the counters on the iptables rule:

iptables -t mangle -nvL OUTPUT

It might not pick up the traffic until the connection is re-established.

 

[slidermike]

Check the counters on the iptables rule:

iptables -t mangle -nvL OUTPUT

It might not pick up the traffic until the connection is re-established.

Thank you Dave.
The re-establish notion makes sense.
I will come back to it in a bit and see if there is any change.

v125027@AX86U:/tmp/home/root# iptables -t mangle -nvL OUTPUT
Chain OUTPUT (policy ACCEPT 12108 packets, 5572K bytes)
 pkts bytes target     prot opt in     out     source               destination
 454K  129M MARK       udp  --  *      eth0    0.0.0.0/0            0.0.0.0/0            multiport dports  !53,123 MARK xset 0x403fffff/0xc03fffff
 237K   79M MARK       tcp  --  *      eth0    0.0.0.0/0            0.0.0.0/0            multiport dports  !53,123,853 MARK xset 0x403fffff/0xc03fffff
v125027@AX86U:/tmp/home/root#

 

[dave14305]

Thank you Dave.
The re-establish notion makes sense.
I will come back to it in a bit and see if there is any change.

v125027@AX86U:/tmp/home/root# iptables -t mangle -nvL OUTPUT
Chain OUTPUT (policy ACCEPT 12108 packets, 5572K bytes)
pkts bytes target     prot opt in     out     source               destination
454K  129M MARK       udp  --  *      eth0    0.0.0.0/0            0.0.0.0/0            multiport dports  !53,123 MARK xset 0x403fffff/0xc03fffff
237K   79M MARK       tcp  --  *      eth0    0.0.0.0/0            0.0.0.0/0            multiport dports  !53,123,853 MARK xset 0x403fffff/0xc03fffff
v125027@AX86U:/tmp/home/root#

Also include the output of:

tc filter show dev eth0 | grep -B1 0x403fffff

 

[slidermike]

Also include the output of:

tc filter show dev eth0 | grep -B1 0x403fffff

v125027@AX86U:/tmp/home/root# tc filter show dev eth0 | grep -B1 0x403fffff
filter parent 1: protocol all pref 5 u32 fh 827::807 order 2055 key ht 827 bkt 0 flowid 1:14
  mark 0x403fffff 0xc03fffff (success 149122)

 

[dave14305]

v125027@AX86U:/tmp/home/root# tc filter show dev eth0 | grep -B1 0x403fffff
filter parent 1: protocol all pref 5 u32 fh 827::807 order 2055 key ht 827 bkt 0 flowid 1:14
  mark 0x403fffff 0xc03fffff (success 149122)

Ok, so the final stats would come from:

tc -s class show dev eth0 parent 1: | grep -A2 "htb 1:14"

All 3 commands together should increment their counts in unison when picking up traffic from the iptables rule.

 

[cplay]

@dave14305 - just thought I’d say thanks for such a well thought out and functional qos - for me personally flexqos performs worlds above cake and the customisation is brilliant.

thank you for your hard work and dedication to this addon.

 

[dave14305]

@dave14305 - just thought I’d say thanks for such a well thought out and functional qos - for me personally flexqos performs worlds above cake and the customisation is brilliant.

thank you for your hard work and dedication to this addon.

Most of the credit for the functionality goes to FreshJR for the original script and the guys who maintain sqm-scripts at OpenWRT. Without their ideas, I wouldn’t have really known what to do.

Thanks for your nice words!

 

[Kingp1n]

Most of the credit for the functionality goes to FreshJR for the original script and the guys who maintain sqm-scripts at OpenWRT. Without their ideas, I wouldn’t have really known what to do.

Thanks for your nice words!

Thanks @dave14305, I'm also using FlexQoS over CAKE. I've seen a couple of posts in regards to CAKE & gaming not getting along to well with latency/random spikes etc.

One future request possibly, if it can be done or maybe it's not even necesary or possible at all, but,

I know you're not a gamer but I was wondering if a future develop script can be implemented for "gamers" specifically with adjusted burst/modify bandwidth to assist in latency, maybe??

Im not even sure if this would be possible or if it can even be implemented at all or if the current script setup works for all environments to include gamers by just simply setting up "gaming" as highest priority enough ?? Can anything else be adjusted?

I don't know the ins/outs of the scripts but I don't think anything else can be done for gaming traffic except for creating the available "gaming rule" which is already available within Flex correct?

Sorry for the noob questions and hopefully you can provide some insight.

Thanks again.

 

[dave14305]

I know you're not a gamer but I was wondering if a future develop script can be implemented for "gamers" specifically with adjusted burst/modify bandwidth to assist in latency, maybe??

What would you adjust? I need to know the theory to understand if I could do it.

 

[Kingp1n]

What would you adjust? I need to know the theory to understand if I could do it.

Great question....and unfortunately im not smart enough to give you any valuable input. I know @el pescador was trying to adjust burst values (I believe for gaming) but I'm not sure if he had any success. Maybe someone (smarter than me) here can provide some valuable info.

As always thanks for your response!

 

[DanNuke86]

What would you adjust? I need to know the theory to understand if I could do it.

Having tested FlexQoS and CAKE for a while now on my 500/35 asymmetrical link, I’ve found that FlexQoS suffers on upload and CAKE on download, although I get overall better results on CAKE it seems. May be due to download not being saturated. Always found low stable ping gives the best experience for FPS online play.

Found this article which ran a study in ping times and RTT effects. Still trying to get my head round it to understand but seems to be the right source of info for this discussion. Might be of some use to you Dave, if you don’t mind a bit of light reading

https://www.nas.ewi.tudelft.nl/people/Rob/telecom/conext06.pdf

 

[el pescador]

Great question....and unfortunately im not smart enough to give you any valuable input. I know @el pescador was trying to adjust burst values (I believe for gaming) but I'm not sure if he had any success. Maybe someone (smarter than me) here can provide some valuable info.

As always thanks for your response!

I had some success but it seemed to change quite a bit over time.
I've had way more success by simply blocking certain locations through Skynet. Really i block the testing servers and on some locations i have to block a group of servers too....so only about 4 or 5 entries in Skynet.

Finding the best located servers is a trial and error and also can change over the days...hours even.

IMO changing the burst rates can help a little more on top of this.

 

[el pescador]

Having tested FlexQoS and CAKE for a while now on my 500/35 asymmetrical link, I’ve found that FlexQoS suffers on upload and CAKE on download, although I get overall better results on CAKE it seems. May be due to download not being saturated. Always found low stable ping gives the best experience for FPS online play.

Found this article which ran a study in ping times and RTT effects. Still trying to get my head round it to understand but seems to be the right source of info for this discussion. Might be of some use to you Dave, if you don’t mind a bit of light reading

https://www.nas.ewi.tudelft.nl/people/Rob/telecom/conext06.pdf

I havent read this as i write....but if any games company is using RTT for latency then they aren't fit to be runnign an online game.
The routes to and from the servers are independent and lag comp should really be calculated on both independent streams.
Ive tested plenty to iD3 and there was always a difference between the routes on traceroute and reverse traceroutes...this will cause issues, although not massive ones if you have a small variance.

 

[el pescador]

The main issue with gaming these days is packet loss out of the home network....and how easy it is for gaming companies to hoodwink their customers on the issue.
Take FIFA for example.
The EA connection tool shows below 1% loss.
But fire up wireshark and you will see Packet Resends all over the place. They even admitted back in 2019 they were trying to get the resent packets to be sent quicker.
Yet they only show malformed packets as loss. Not the original lost packet which had to be resent.

Most other games dont use GVSP so resends wont show up. But you can presume the loss is way higher than what they say...Warzone for example.