yuripv: are you willing to handle MFC of this and taking it to 14.0, or would you prefer that i deal with it?

I think, from looking at the bug reports, that this is bad enough that we should probably EN it to 13.2

kevans: please do if you don't mind

yuripv: yup, no worries

hah, thwarted again by aarch64's default-unsigned char

couldn't figure out why the scenario in the PR for this regex(3) bug wasn't failing for a couple minutes, until it hit me again that it's default unigned for the second time this quarter

if I run slackpkg but then the screen scrolls by, is there any way to go back and get a log which packages were still installed (or skipped, whatever happens) with bad md5sum?

oops, wrong channel. Sorry

Hi all. I'm using 13.2-RELEASE on my x220 with zfs rootfs. I've noticed that time to time the zfskern consume the entire CPU resource and the machinee become too slow.

I'm wondering if a small value for ARC can make the system going on this

lognull, I am typing this on my own x220, however mine is running a Linux kernel, but my x220 often heats to the point of thermal throttling.

I am wondering if perhaps that is what is happening to your x220. That it is heating up until the Intel cpu thermal throttles itself. Which causes things to become very slow.

For my Linux kernel system I see this logged to the system log. I would hope and expect to see the FreeBSD kernel log it to the /var/log/messages system log.

rwp: its strange... The only thinh I'm runing here is a xorg+fluxbox and a firefx with one youtube tab

I mostly notice it when surfing reddit with all of the videos auto-playing. That's when it is most noticable on my machine.

And then things become very slow and chunky.

The reason I think it might be thermal is that my x220 has 8GB of RAM. But I have three other systems running large-ish ZFS arrays, two of which have only 4GB of RAM which most people would consider small-ish and have no such slowdowns or chunkyness on those.

Therefore I don't think it is ZFS that is having a RAM resource problem since those machines are small RAM systems.

Hmm, maybe. You got a point there.

Runing "top" would show the state of the caches in the top header lines.

I totally forgot to check this.

It is possible though that ZFS is a cause of cpu use and therefore might be contributing to heating to thermal throttling.

The X series was designed for small light weight travel use and not for heavy cpu use. I have tried to improve mine but have so far been ineffective at really making it cooler.

And by that I mean I have replaced the fan and replaced the thermal paste.

It's interesting that I just bought an x270 which is about 10 years newer than my x220 and I rather find it to be about the same exact machine for most practical purposes.

Which is why I am typing on the x220 right now. The x270 has the same exact screen size so no improvement there. It has the island style keyboard which is just personal preference now. It has 16GB of RAM where my old had 8GB and that is the only improvement but I apparently don't need it. Battery life on the new with the standard two pretty good batteries is about the same as on the old with its one

pretty good battery.

those machines are such a bless

I have this x220 with FreeBSD13.2 and other T60 with OpenBSD. No regrets.

I also love my T60p! Due to size it has become my coffee table machine. But still running strong.

I'll check the memory usage related to ZFS. Maybe I forgot to limit the RAM usage

why does tar xvf -C dir FreeBSD-13.2-RELEASE-amd64-memstick.img fail saying error opening archive failed to open -C? it's in man bsdinstall!

a -C dir has to exist

dunno what that means i thought -C was an option

ya it is

polyex, use dd if you want create an boot usb stick

it's for dir

no

i'm copy pasting a command from a man page and it errors so isn't that a bug in the docs?

i don't know tar good enough to know

oh is tar xvf -C not right to extract a .img file?

maybe that's the mess up im making

tars f option expects the filename as next parameter. the space is optional

so the man page does have a bug?

the tar-manpage? which line?

no bsdinstall man page

tar xvf -C dir FreeBSD-13.2-RELEASE-amd64-mem... line

oh sorry they use a different name for dir, so the xvf -C line

i've found the line.

ya i really feel like that's a bug. just tested it with an .iso file

looks like that to mee, yes

will you or someone like meena pls report?

you found it, so the credits should go to you :)

lol i don't care about credit just freebsd being the best!

so what's the right way to extract an .img file?

not iso like the buggy line but .img

i've no idea if tar ist capable of extracting it, but reordering the parameter might work by fixing the tar command

is there a way to tell the *actual* clock speed of a running cpu?

specifically, cpu-monkey.com/en/cpu-intel_xeon_e_2378g is 2.8GHz base, but can boost up to 5.1GHz on a single core

ya the order fixed it

this will be a huge help for build from src if the boost is actually occuring

lol on a hunch i looked at newer versions of man page for bsdinstall, it's fixed!

just 13.2 isn't backported

`sysctl dev.cpu | grep freq:` looks promising but so far we are sitting firmly on 2801 still

.img isn't recognized by tar psycorama

polyex: thought so.

dch: do you have any process running to limit your cpu clock speeds? (powerd, or similiar)

psycorama: not, not atm

so anyone know what the util is for .img like tar is to extract .iso pls pls? :)

polyex: I'd try 7zip from ports its pretty good at unpacking most stuff

is turbo-boosting set in the BIOS/EFI?

nothing in base for .img!?!?

psycorama: I don't have direct acccess for that

polyex: recapping as I missed above, you have FreeBSD-13.2-RELEASE-amd64-memstick.img and want to extract what from it?

dch: you could try to set a minimum frequency with powerd (or similiar)

just extract it like i extract a disc1.iso of the same installer

so i can put a file in /etc/installerconfig for unattended bsdinstall

polyex: ok, you want to modify it.

ya

then repack it as .img again

switch to root

mdconfig /path/to/.img

mount /mnt /dev/md0

now you have a r/w UFS partition

actually I should check this, that path is almost certainly wrong

is there any tool to do it in place edits of .img files?

polyex: here you go

Title: Snippet | IRCCloud

so this is the "quick" way

if you want to build custom FreeBSD install images today, I would recommend poudriere-devel and the image format

Title: poudriere image.8 devel · freebsd/poudriere Wiki · GitHub

it looks 1000x more complicated than it actually is

what's the benefit of poudriere-devel? and ya i need to automate generating unattended install/scripted bsdinstall images to then deploy. that make sense?

polyex: the mdconfig ... mount ... umount ... trick above is manual, not repeatable

polyex: poudriere image automates that

explain high level how it works pls?

- input: git for src & ports of freebsd if needed, and an overlay directory

- output: your freebsd install image, with the overlay bits in the right place

basically your /etc/installerconfig (and whatever else you want) ends up inside the image

can the output be in different formats? just like how freebsd installs come in like disc1.iso, minimal.iso, memstick.img different variants

or just 1 or?

polyex: it will produce whatever flavour you want

incredible

ya the dream is to do like i said, automate generating custom freebsd install images that can deploy, start up once and install/configure everything, reboot once and be in service

ok I updated the URL above with copying a file in, then umounting and removing the md device

after that your modified installer is complete

wow tyvm!

polyex: theres also my oldish notes on poudriere, docs.skunkwerks.at/s/3SL9taN8s to help you get started there

Title: pkg and poudriere - a lightning trip - HedgeDoc

ty!

so what's the zfs dataset thing you do?

i run zfs on host, and am planning to start running bhyve vms on zfs too. not ufs

(ufs on zfs)

so the scripted bsdinstall will need to expect zfs backed storage

found that in the man page for bsdinstall so shouldn't be too hard

zfs in poudriere?

it just uses zfs to keep things nice and tidy, ensuring builds are not dependent on host system

well pou will run on a box and that'll be running zfs ya? maybe i'm missing something

ya that sounds good

yep poudriere runs fine on ufs as well, but it can do smarter stuff on zfs

you said you deploy zfs datasets on top.. on top of what?

aah ok like this

Title: GitHub - KlaraSystems/sync-be: Configuration management for ZFS Boot Environments

this might be a bit further on than you're ready for but in a nutshell:

- use poudriere-devel image to create a full install image

- blat that on by arcane means not covered here

some kinda raw disk imaging?

- periodically push a new Boot Environment (zroot/ROOT/os-20230926 ) for example

man that sounds amazing

is a zfs boot env same as a freebsd boot env?

- sync-be transfers a few config files and dirs from the current BE to the next one

why do you want that?

polyex: people.freebsd.org/~dch/talks/euroBSDcon2023/immutable.pdf the video for the talk isn't up yet sadly

if you have openoffice you can see the speaker's notes people.freebsd.org/~dch/talks/euroBSDcon2023/immutable.odp

polyex: the image is the same across all servers

but things like /etc/{rc.conf,hostid,ssh/*} are different

ya this is exactly what i'm getting into. cutting out CM by deploying preconfigured images

all immutable

damn this is sick

also moving to heavy use of jails in addition to bhyve

the servers are installed via a custom cloudinit hook that does DD over the top

damn, nice

oh. dch always has some nice howtos!

yes, that magic is from Equinix Deploy.

what's equinix deploy?

metal device create ... --userdata '#cloud-config

#image_uri=...img.gz'

equinix has a hosting service as well as networking (they bought Packet.net)

jails linked with ipv6 mesh network niiiiiice

ahhh

equinix provides BGP and similar stuff as part of this

Equinix is one of the largest datacenter presences on earth, not just a hosting service.

for scale

and the sysutils/equinix-metal-cli port allows creating systems from a custom image

there are a few rules, I forget the precise requirements, but its about how the partitions need to be named

winternull: yes they are a veritable leviathan

14 STABLE WEN?

equinix sells vps?

no VPS, just bare metal. such sweet bare metal.

and fix the CDN please I have to do pkg add xfce like 20 times and it starts working with no reason

psionic: this sounds like you could do a BZ report with slightly more specific info

equinix does dedicated servers? that's what you mean by bare metal?

yep

what makes it sweet?

polyex: all the flexibility of "cloud" but its actually a full machine

Title: Bare Metal Standard Gen 3 Servers - Automated Configurations by Equinix Metal

full mgmt incl provisioning via API

the server says network 2 x 25gbps then the price of like $50/mo... that doesn't INCLUDE 2x25gbps of bw tho right?

that would be way too good a deal

dch deploy.equinix.com/product/servers/m3-small

Title: m3.small.x86 - Equinix Metal Intel® Xeon®

ya looks like bandwidth is extra

Title: Directly connect your infrastructure to thousands of destinations – Equinix Deploy

50 Mbps $0.15 /hr  is cheapest = $108/mo

so like $160/mo for a good little server and fair bit of bw. seems like an ok deal?

if they had generous and powerful ddos protection to rival cloudflare monopolists that would be tops

amazes me that potential/inevitable DDoS victims will host systems and applications with the very provider that's protecting many of the DDoS actors on their own networks

^ this

equinix networking is not surprisingly very good

better aggregator to xconnect to and just use equinix hw?

winternull i don't know network world inside and out, what can i do? how can i find good upstreams?

depends on the requirements. there's genereally not a one-size-fits-all solution.

:/

how are you defining Good or Sufficient in this case?

what's your cost limit, ingress/egress p95, do you need geo-redundancy or ability to rapidly scale nodes or instances, do you use containers or 'serverless' compute, etc etc

k

there's a range of standard questions for scoping a deployment and locating providers, so it depends on the level of infrastructure provider involvement and expectations from the customer to facilitate a solution

dch isn't it possible to skip the step where i mount the installer img? can't i just git clone the latest release branch of the source, add the /etc/installerconfig file, then generate an img from that with the /usr/src/*/release scripts?

im trying to run /usr/src/release/amd64/make-memstick.sh and it says /boot/loader.efi no such file or dir. i cloned the freebsd/freebsd-src git repo releng/13.2 branch

polyex: there are an infinite amount of ways to build FreeBSD from source, but start off with poudriere-devel and an overlay dir, it will do all the things you need

make memstick has prior dependencies which you're not meeting (buildworld, buildkernel, etc)

ohhh damn. but .iso doesn't?

ah .iso depends on it too

so i need to dl existing isos or imgs, can't use a git clone/pull

unless i use pou for the make build world stuff

there's nothing wrong with nutting this all out yourself, but for getting from zero to the finishline, poudriere-devel is the way to go

the #1 advantage of poudriere is reproducible builds in a clean environment

but if i can just take disc1.iso extracts, then copy my overlay dir right over the extract copy, then run the iso maker script, why do i need any more?

ya from that same disc1.iso extract, i can use release scripts to repackage as both .iso and .img!!

polyex: did you figure out what i meant with: "07:25 <meena> a -C dir has to exist"

ya the man page syntax was buggy. i found the answer, and the -current man page has the bug fixed lol

polyex: for building installers, I'd err on a clean toolchain and reproducibility myself

yuripv: while you're on a roll reviewing my phabricator contribs, do you also wanna take a look at: freebsd/freebsd-src #827 ?

Title: document pmap_kextract/vtophys by igalic · Pull Request #827 · freebsd/freebsd-src · GitHub

has anyone in here 'emby-server' up and running in a jail?

emaste: just wondering if there is any smbfs2/3 work in progress? (looking at hackmd.io/@jhb/ByWrxQmr2)

hi, I want to format a SD card, I see somewhere that gpart needs to be used, but gpart is not installed, is there anything newer to be used instead?

gpart is part of the base system

oops

I was ssh'd into a linux box

is there a feature similar to "zzz -Z" in openbsd? i.e., hibernate. apropos lists nothing for me.

zzz(8) says it simply invokes acpiconf -s <state>, try specifying the state via acpiconf?

and how do I delete partitions from the sd card, I see them with gpart show, but how can I delete everything and restart from scratch

(format as FAT32)

# gpart destroy da0

gpart: Device busy

if there are existing partitions, you have to delete them first, or use -F

or use dd :-D

thanks RhodiumToad

once I create the partition with gpart add -t fat32 da0

do I have to "format" it? (I am sure it is a dumb question, but better ask)

mkfs -t msdos ?

newfs_msdos

newfs_msdos -F32 ...

thanks

is there a way of checking cpu degradation? i suspect this happening on my thinkpad w520. having massive CPU0: local APIC error 0x80 dmesg message

Use a pastebin.

Welp, nevermind. Wasn't scrolled down.

any idea what happened to the forums ?

they look ok from here?

CyberCr33p: that's not the first time you are asking, and forums don't seem to have any problems. If you are using vpn, you are most likely blocked.

err, i mean public vpn services, of course

I'll die a happy man if I don't have to deal with ghc tooling >:(

okay, so I've got a Debian jail & a FreeBSD host sharing some media files over nullfs

I need a "media" group on both systems

isn't that weird to have user dbs kinda blur together like that?

well, I'm just trying to figure out what GID would work on both systems

there's very little in /etc/group on the standard FreeBSD install, wondering if daemons add some lower number groups afterwards?

veg: create a dedicated user/group with the same UID/GID on both system to share the date.

thanks meena

polyex: no, not weird at all