There are two packet captures from fq with and without pacing here:

https://owncloud.proikt.com/index.php/s/KuXIl8h8bSFH1fM

The server (with fq pacing/nopacing) is 10.0.5.10 and is running a Apache2 webserver at port tcp port 443. The tcp client is nginx reverse proxy at 10.0.5.13 on the same subnet which again is proxying the connection from the Windows 10 client. 
- I did try to connect directly to the server with the client (via a linux gateway router) avoiding the nginx proxy and just using plain no-ssl http. That did not change anything. 
- I also tried stopping the eth0 interface to force the traffic to the eth1 interface in the LACP which changed nothing.
- I also pulled each of the cable on the switch to force the traffic to switch between interfaces in the LACP link between the client switch and the server switch.

The CPU is a 5-6 year old Intel Xeon X3430 CPU @ 4x2.40GHz on a SuperMicro platform. It is not very loaded and the results are always in the same ballpark with fq pacing on. 

top - 21:12:38 up 12 days, 11:08,  4 users,  load average: 0.56, 0.68, 0.77
Tasks: 1344 total,   1 running, 1343 sleeping,   0 stopped,   0 zombie
%Cpu0  :  0.0 us,  1.0 sy,  0.0 ni, 99.0 id,  0.0 wa,  0.0 hi,  0.0 si,  0.0 st
%Cpu1  :  0.0 us,  0.3 sy,  0.0 ni, 97.4 id,  2.0 wa,  0.0 hi,  0.3 si,  0.0 st
%Cpu2  :  0.0 us,  2.0 sy,  0.0 ni, 96.4 id,  1.3 wa,  0.0 hi,  0.3 si,  0.0 st
%Cpu3  :  0.7 us,  2.3 sy,  0.0 ni, 94.1 id,  3.0 wa,  0.0 hi,  0.0 si,  0.0 st
KiB Mem : 16427572 total,   173712 free,  9739976 used,  6513884 buff/cache
KiB Swap:  6369276 total,  6126736 free,   242540 used.  6224836 avail Mem

This seems OK to me. It does have 24 drives in 3 ZFS pools at 144TB raw storage in total with several SAS HBAs that is pretty much always poking the system in some way or the other.

There are around 32K interrupts when running @23 MB/s (as seen in chrome downloads) with pacing on and about 25K interrupts when running @105 MB/s with fq nopacing. Is that normal?

Hans-Kristian


On 26 January 2017 at 20:58, David Lang <david@lang.hm> wrote:
Is there any CPU bottleneck?

pacing causing this sort of problem makes me thing that the CPU either can't keep up or that something (Hz setting type of thing) is delaying when the CPU can get used.

It's not clear from the posts if the problem is with sending data or receiving data.

David Lang


On Thu, 26 Jan 2017, Eric Dumazet wrote:

Nothing jumps on my head.

We use FQ on links varying from 1Gbit to 100Gbit, and we have no such
issues.

You could probably check on the server the TCP various infos given by ss
command


ss -temoi dst <remoteip>


pacing rate is shown. You might have some issues, but it is hard to say.


On Thu, 2017-01-26 at 19:55 +0100, Hans-Kristian Bakke wrote:
After some more testing I see that if I disable fq pacing the
performance is restored to the expected levels: # for i in eth0 eth1; do tc qdisc replace dev $i root fq nopacing;
done


Is this expected behaviour? There is some background traffic, but only
in the sub 100 mbit/s on the switches and gateway between the server
and client.


The chain:
Windows 10 client -> 1000 mbit/s -> switch -> 2xgigabit LACP -> switch
-> 4 x gigabit LACP -> gw (fq_codel on all nics) -> 4 x gigabit LACP
(the same as in) -> switch -> 2 x lacp -> server (with misbehaving fq
pacing)



On 26 January 2017 at 19:38, Hans-Kristian Bakke <hkbakke@gmail.com>
wrote:
        I can add that this is without BBR, just plain old kernel 4.8
        cubic.

        On 26 January 2017 at 19:36, Hans-Kristian Bakke
        <hkbakke@gmail.com> wrote:
                Another day, another fq issue (or user error).


                I try to do the seeminlig simple task of downloading a
                single large file over local gigabit  LAN from a
                physical server running kernel 4.8 and sch_fq on intel
                server NICs.


                For some reason it wouldn't go past around 25 MB/s.
                After having replaced SSL with no SSL, replaced apache
                with nginx and verified that there is plenty of
                bandwith available between my client and the server I
                tried to change qdisc from fq to pfifo_fast. It
                instantly shot up to around the expected 85-90 MB/s.
                The same happened with fq_codel in place of fq.


                I then checked the statistics for fq and the throttled
                counter is increasing massively every second (eth0 and
                eth1 is LACPed using Linux bonding so both is seen
                here):


                qdisc fq 8007: root refcnt 2 limit 10000p flow_limit
                100p buckets 1024 orphan_mask 1023 quantum 3028
                initial_quantum 15140 refill_delay 40.0ms
                 Sent 787131797 bytes 520082 pkt (dropped 15,
                overlimits 0 requeues 0)
                 backlog 98410b 65p requeues 0
                  15 flows (14 inactive, 1 throttled)
                  0 gc, 2 highprio, 259920 throttled, 15 flows_plimit
                qdisc fq 8008: root refcnt 2 limit 10000p flow_limit
                100p buckets 1024 orphan_mask 1023 quantum 3028
                initial_quantum 15140 refill_delay 40.0ms
                 Sent 2533167 bytes 6731 pkt (dropped 0, overlimits 0
                requeues 0)
                 backlog 0b 0p requeues 0
                  24 flows (24 inactive, 0 throttled)
                  0 gc, 2 highprio, 397 throttled


                Do you have any suggestions?


                Regards,
                Hans-Kristian




_______________________________________________
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat


_______________________________________________
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat