Howdy, folks!

Just making a personal shell script that acts as a reconnector for catgirl.

Basically, it runs catgirl in a infinite loop, and after it closes, sleeps the script for 5 seconds giving me time to Ctrl+C out of the script. Afterwards it runs catgirl again.

termbin.com/0jku my bsdinstall installerconfig is getting nondeterministic network init time. any solution?

it's like it takes a few seconds for the network to warm up or something then traffic starts passing normally

but elsewhere when i run this it always works immediately

termbin.com/b264 is clearer

sleep 5, and it works. maybe i need to set the if to "up" somewhere?

common reasons to get "solaris: notice: cannot find the pool label for zroot" after an install?

seems i got a prob with my geli encrypted zfs

100py: forums.freebsd.org/threads/how-to-c…nce-of-a-host-in-shell-script.69582

last post best post

when i run ifconfig after service netif restart it says status: no carrier. maybe i need to set it to UP?

where'd the i915kms kernel driver go?

i used to use it for my mac mini to run freebsd

but now it says no such file or dir

ah nvm

if an interface has status: no carrier, what command makes it status: active?

ifconfig if up will do it

i guess the error im running into is the lag between when i bring an if up with service netif restart, and when its status changes from no carrier to active

maybe i have to embed a sleep 5 in there to deal with that

ok so now i just need to figure out why my geli encrypted zfs isn't booting right

JRiver 29 running on FreeBSD 14.1-RELEASE in Wine 9!!

trying to install grub for multibooting purposes, doing `sudo grub-install --target=x86_64-efi --efi-directory=/boot/efi --bootloader-id=GRUB` gives me `grub-install: error: relocation 0x4 is not implemented yet.`

yabobay: did you verify that the efi partition is mounted? i don't have it in my fstab, so i have to mount it if i want to monkey around with it

scoobybejesus: it's mounted

i got my grub from the `grub2-efi` port, i hope that's the right one

dang, i tried. i don't have any more ideas atm

That looks like a repetitive bug for grub... linux/freebsd alike that has been re-appearing for several years

(in my initoal few minutes of searching)

there's a relocation patch to fix it on the linux side

not sure if that was added in the FreeBSD port

huh. so just like wait a bit for some upgrades to come out?

cause if so then i'm not in any particular hurry so i'm ok with that

yabobay: problem is who knows if its been reported/acknowledged/taken to work on

Tenkawa: is there any way to get freebsd's regular bootloader to recognize windows?

That I canot say.. I haven't run windows in 10 years

lucky you haha

I barely ran windows then heheh

also this isn't a support question but while i was digging around in /boot i realized a lot of the bootloader stuff is written in forth which is kinda nuts

why? forth is sturdy and stable like cobol and assembly... it doesn't change

(I am from the cobol and assembly era)

i'm from the right now era. was the cobol and assembly era better when it came to developing stuff?

It was different..

bbl... time for lunch.. cheers all

anybody here using prometheus to monitor services on freebsd? if so, what exporter are you using for that? not seeing anything related with node_exporter.

phryk: Not sure how related, but I'm using node_exporter on pfSense. Seems to be exposing counters for a lot of the network stuff going on. Haven't explored "true" freebsd exporters.

services that I write myself, i usually put a /metrics endpoint into a/the http.ServeMux, and expose service-specific stuff there. Not sure how a more generic exporter is supposed to know what counters I'm interrested in, in a custom piece of sw

oh, exposing my language of choice, there. :)

mml: yeah, i'm not interested in the network counters (i mean i am, but for different things). what i want is a "service is running" state timeline, i.e. what 'service x [one]status' shows. on linux+systemd machines there's the `node_systemd_unit_state` metric for that – essentially i'm looking for something giving me the freebsd equivalent of that…

i can of course hack up my own exporter that enumerates all rc files and calls status on them but that sounds really inefficient. wondering if there's some C API that'll make this less awful, but given that rc scripts are just shell scripts… i don't think so?

i general, i keep coming back to the idea of writing a "libfreebsd" that directly uses any C APIs to expose freebsd-specific stuff like package states and similar info.

TIL libpkg is a thing 🤔

but, how'd you expose metrics of services running? a prom-metric is a number, with associated labels. You could of course attach a label with servicename, but that would pollute your series, heavily.

upservice="servicename"} would explode pretty quickly

up{service="servicename"} would explode pretty quickly

and, probably use another seriesname than "up" :)

are there any non-obvious power settings to consider? i have a server board with a Xeon-E and 1 nvmeplugged in. If I boot any recent flavor of Linux (defaults, no config), it idles as 29W. Boot FreeBSD, and that is around 40W. I've noticed the same with FreeBSD bhyve instances too. I can't figure out if there's something I can tweak, or if the FreeBSD kernel is just less concerned with power

efficiency. i've tried hwpstate_intel(4) tweaks, but nothing gets the idle as low as linux.

markmcb: Is fan control handled by the OS or by IPMI?

Xeon-E is… vague. Assume post KabyLake?

duncan: IPMI. Xeon E-2388G.

OK, I'd check if there's any difference by monitoring IPMI, but you presumably have a good idea of it assuming it's rackmount and loud

the fans power down for both FreeBSD and Linux after the boot process, so I don't think it's that

Do you have CPU frequency scaling set up, or perhaps NVMe is at a high power state? 11W is plausible in terms of combination of a few things.

also is this average or observed at specific points?

FWIW, I have X12STH-SYS and the 28W idle reported by IPMI is not accurate. More like 35-40W at the wall.

it's average, as reported by the BMC, checked periodically with the same results. I have a X12STH-F, so could be a reporting issues, but I poll it for both FreeBSD and Linux and see the difference.

haven't done anything on CPU scaling, just speedshift tweaks, e.g., dev.hwpstate_intel.0.epp=100

I'd graph it over 20m with a similar workload and compare. I would guess Linux for relatively new hardware like this has a power subsystem with buy-in from the manufacturer.

My box runs NetBSD HEAD though

ipmitool should report it

I think sdr power is the command

yeah, i've monitored it over several house on both OSes with no workload other than being on. 29W and 40W are the steady idle power readings I get

fair enough. that sounds fairly robust

and i've been using ipmitool dcmi power reading

I wouldn't expect much higher idle usage without frequency scaling, but I wouldn't be surprised if it was consequential either

mml: nah, that's pretty standard practice. the real problem is having stuff in labels that you wouldn't feasibly put into an enum. like URLs visited on a site. a couple dozen or a couple hundred isn't much of a problem.

node_exporter does it for systemd units and windows_exporter does it for, well, windows services.

node_exporter is directly published by the prometheus team and windows_exporter is published by the "prometheus-community" account on GH, so I consider this canonically correct and officially supported use of prometheus.

Howdy, folks!

duncan C-states? maybe.

maybe the timecounter too man.freebsd.org/cgi/man.cgi?query=timecounters&sektion=4

because depending on what you have set it might prevent C-states

phryk: Well, might just be that I haven't been exposed (!) to a big(ger) setup, so that my threshold for big and messy is somewhat different :)

mml: windows_exporter routinely exports like 100+ services. i think the most i've seen was around 300… :F

at the end of unattended bsdinstall, the debug log pops up and it says at the bottom "cannot unmount /mnt/tmp: pool or dataset is busy" so if i want to try zfs unmount -f to see if that fixes it, would i pass it zfs unmount -f /mnt/tmp or zfs unmount -f zroot/tmp?

yeah! clang/llvm 19 :-)

in current

nice

llvm20 when

i don't know, v20 isn't on the llvm releases page yet. I'm happy to have v19 as it makes it easier for me to build zig from source

thanks to dim@ and whoever helps on that

oh wow current llvm is 15 so ya prolly pretty big upgrade

looks like a busy day tomorrow, 3x sec advs

9for me anyway, i have lots of bhyve)

i wonder if any woulda been prevented with rust

problem with rust, at least on freebsd, is that it seemingly cant be held to a stable api

$thing updates? rebuild rust!!

curl updates? rebuild rust!!

ripgrep updates? rebuild rust!!

takes 45 mins-1hr 30 each time. a tremendous amount of energy could be saved if rust was held to a point

not built ports for 60 hrs? rebuild rust!!!

before oct23rd, llvm on freebsd-current was llvm/clang v18.1.x, now it is llvmorg-19.1.2-0-g7ba7d8e2f7b6

llvm doesnt need to be rebuilt all the time though. the problem with llvm is that there are 57 different versions and each port wants its own particular one

yeah, after freebsd switched from gcc to clang, buildworld times got pretty long

i stick with openbsd to my important, personal machines. at least it holds to a version, then when it needs to upgrade, everything happens at once

grafana port pretty old right? v10 but v11 launched 6 months ago

wha a beautiful 7/24 alive community this is felt like installing freebsd :) ive used years ago on my lenovo t430 for few years