ATT 2G fiber - upload is low

akhouston

michmoor

https://docs.netgate.com/pfsense/en/latest/troubleshooting/low-throughput.html

akhouston

@michmoor yeah, I poked around... the thing is definitely not CPU bound. Def something going on with the box but can't put my finger on what it is precisely.

stephenw10

Try setting the loader variable: hw.igc.max_interrupt_rate to something higher like 20000.
The default is 8000 and you might be hitting that.

Create the file /boot/loader.conf.local.
Add to it: hw.igc.max_interrupt_rate=20000
reboot.

Steve

Dobby_

@akhouston said in ATT 2G fiber - upload is low:

I have pfsense+ 23.05 installed on a box wtih intel i225 - so 4 2.5Gb ports.

Is PPPoE in the "game play" you will be nailed to one queue
at the WAN, if not you can use more queues, one CPU core
= one queue.
On top the same role with the LAN ports too. As an example:
Intel i211at are only using 2 queues and Intel i210at is
able to use 4 queues.

akhouston

@stephenw10 Thanks, I don't think it made a difference.

akhouston

@Dobby_ I am not sure pppoe is applicable to att fiber, but I am not sure either I understand your statement.
the box is basically celeron n5105 with i226 NIC and 32GB memory.

stephenw10

Your WAN is DHCP you can see the gateway name so it's not that.

Check the CPU usage when you're testing. The best way to do that is to run top -HaSP at the command line.

akhouston

@stephenw10 Actually this change cut down my download speed from about 1.4-1.5Gb to 600-700Mb. Removed it and the down speed is back to 1.4-1.5

akhouston

@stephenw10 do i just take a snap of that in the Command Line window when running a test? I am watching the cpu usage chart on the dashboard and at the peak up and down test (those blue and orange charts onthe right, it didn't go higher than 30%, hence load average

akhouston

So I ran iperf just to make sure nothing silly is going on. And it appears that pfSense sending is definitely heavier than pfSense receiving... Not sure where to dig from here, any ideas will be appreciated...

stephenw10

You set that rate limit to 20000 not 2000?
I'm surprised doing that could possibly reduce throughput.

Yes, run top at the CLI then hit Q to quit leaving the data fixed to copy/paste out of there. Make sure you're not getting hung up on one core at 100%.

Steve

akhouston

@stephenw10 I found the culprit, evidently the traffic shaper on the interface had a bandwidth limit of 1gbps - don't remember if and when I even set that, but whatever. Reset it and now i'm happy like a clam... Thanks everyone for your help.

stephenw10

Ha, that would do it. Nice.