Messages

1

24.7 Production Series / Re: "ping: sendto: No buffer space available" under heavy load using pfatt bypass

« on: September 01, 2024, 04:29:23 am »

Nevermind, it crashed after about 30 hours. I'm trying something else now: the obscure net.graph.maxdgram and net.graph.recvspace tunables. For some reason they were very small and while it works for most bypass applications crashes with Tor relays.

This forum post had some tunables and got me thinking, it must be netgraph buffers: https://freebsd-questions.freebsd.narkive.com/lPixO7df/ping-sendto-no-buffer-space-available

2

24.7 Production Series / Re: "ping: sendto: No buffer space available" under heavy load using pfatt bypass

« on: August 31, 2024, 08:21:44 pm »

I've found the real culprit (hopefully), apparently logging while using pfatt causes issues, that or the buffers were too small.

Tor relays open a lot of TCP connections and trying to log each session while running bypass for some reason kills my internet. pfSense+ does have the advantage of a built-in AT&T bypass but CE lacks it and Netgate no longer gives homelab licenses.

I've had 24 hours of uptime since I disabled logging while allowing ~150-250 Mbps concurrent Tor traffic.

3

24.7 Production Series / Re: "ping: sendto: No buffer space available" under heavy load using pfatt bypass

« on: August 30, 2024, 11:10:23 pm »

That's what I'm actually doing now, using the tunables section in the WebUI.

4

24.7 Production Series / Re: 24.7.x Dashboard slow/not-working Linux laptop; OK WIN 10 laptop -other problems

« on: August 30, 2024, 11:09:30 pm »

The 24.7 dashboard is slow on both Windows 11 (work Lenovo X1 Yoga 3rd) and Fedora 40 (home Dell XPS 14 9440). I normally use Firefox but it's slow on Chrome and Edge also. I haven't tested my MacBook yet.

I have an i9-13900F desktop PC with 96GB of RAM, it's still slow.

5

24.7 Production Series / "ping: sendto: No buffer space available" under heavy load using pfatt bypass

« on: August 30, 2024, 05:59:38 pm »

I have Minisforum MS-01 running OPNsense 24.7. This is connected to Frontier FiberOptic (ex-AT&T Fiber) using 802.1X with the AT&T Fiber bypass here: https://github.com/MonkWho/pfatt

My connection runs 6 Tor relays where each instance has 8000+ TCP sessions open at once.

I have a problem: when running my server for more than a couple hours, the WAN goes down and I get an "ping: sendto: No buffer space available" issue. Rebooting brings back the connection until this happens again. If I use the Minisforum i226 NICs, it happens within 1-2 hours. Using a Dell Broadcom BCM5720 NIC, it lasts longer but goes down around 12 hours.

Is there a way to solve this?

When I had the issues, I didn't have any tunables but I'm going to try this in /boot/loader.conf.local:

Code: [Select]

kern.ipc.maxsockbuf=68194304
kern.ipc.nmbclusters=32311168
kern.ipc.nmbjumbop=16155584