so how do we get wireguard connection set up as an rc service in 14? do we still gotta use wireguard-tools pkg? i thought it was just in kernel now

freebsd 14 unattended bsdinstall throws up a "rebooting" gui when it's done. 13 didn't do that. can disable?

because now every bsdinstall takes 10 seconds longer out of nowhere

rennj: being anti-capitalism. Also rennj: paying for hotels and Starbucks.

freebsd 14 unattended bsdinstall throws up a "rebooting" gui when it's done. 13 didn't do that. how can disable? because now every bsdinstall takes 10 seconds longer out of nowhere

it's not documented in man bsdinstall

what's the right way to zfs snapshot my home dir? i do a fresh install, then zfs create zroot/home/polyex, then zfs snapshot zroot/home/polyex@first. seems to work but then i can't sftp into my account (???)

oh, the create command deleted my home dir

ouch lol

if you destroy it, the old content comes back. similar to mounting over something

destroy, rename, create, rsync

how can i make this zfs dataset just adopt what my home dir already is?

as if i created then moved files in

because the OS creates my home dir by default so i just wanna work with that

ya

but i don't want that

there's no zfs create adoption type flag, you'll just have to move it away then move the contents back

yikes

recent adduser(8) can create the dataset for you

ok so i create the dataset before creating the account and then the home dir will end up being in the dataset?

ya

correct

ok i'll try that now

tried zfs create -o mountpoint=/home/polyex zroot/home/polyex and it worked kevans!

hm no it didn't wtf

happy freebsd day

you too

polyex: did the dataset actually get mounted?

no clue

mount(8) knows all sees all

seems the only way to do this is zfs create THEN move the home dir files in

you can not create over something existing. have you checked `man adduser` if they way described there would work for you?

(or rather, you can create, but the content will not transfer over)

hi, i've upgraded my 13.2 homeserver to 14.0. my pf ruleset that seemed to work on 13.2 does not work on 14.0 anymore. the relevant parts of my pf.conf pastebin.com/raw/Nm3jBsXL

the last line does not work anymore on freebsd 14. a tcpdump -eni pflog0 looks like this pastebin.com/raw/zeqqA80N

the weird thing is, the traffic going back does get blocked. seems like keep state does not work or am i missing something?

Grauwolf: I'm a ipfw user, so I can't immediately spot what's wrong with your pf config - but if you have spare CPU cycles and can set up a testing environment, one thing that can help you narrow down what changed is git-bisect(1).

It helps a lot if you also use BSD make's meta-mode as described on wiki.freebsd.org/MetaMode

Title: MetaMode - FreeBSD Wiki

In my experience, it's pretty consistent in narrowing down issues that changed between versions, if you don't know what caused them and can reliably test them, regardless of what the rootcause is - it just requires a lot of patience.

this is probably what i'll have to do in the end. but maybe someone knows what the problem is

Definitely worth asking, yeah.

freebsd 14 unattended bsdinstall throws up a "rebooting" gui when it's done. 13 didn't do that. how can disable? because now every bsdinstall takes 10 seconds longer out of nowhere. not documented in bsdinstall man page

dvl thanks. I ended up using your configs and some more to get my new time machine working.

Does anyone know if FreeBSD has any common issues with VLAN performance?

I used to get 999Mbps with my old setup, but now I get half of that on VLANs

altho to be fair, I did change the switch

maybe it's a switch issue

I am able to get 1Gbps with iperf -P2, tho

Is it possible to have jail's path be a jailed dataset? How would I go about mounting devfs in that case?

I'd like to emulate separate machines using zfs on root without going full-vm route.

dautor8518050867 what do you mean?

dautor8518050867 the jail command is running on the host, so it *can't* be jailed

dautor8518050867 however, if you want to keep some kind of persistent data, you can have a jailed dataset nested under the jail's dataset

antranigv: I'd like to be able to create snapshots of the whole (jail) system from within jail. To allow that, I'd need to jail the dataset and zfs-jail it. But if you jail a dataset it is hidden from prison0.

The problem is that there is no way to jail a dataset and create a jail rooted at that dataset at the same time.

antranigv: good. Isn't it great how well all those components play together. Remember to do snapshots on the time machine filesystems.

Anyone have any idea what this file might be?: bpa.st/JXSA

Title: View paste JXSA

Title: FreeBSD Day 2024 - Ed Maste - YouTube

live now :P

CrtxReavr: is it a picture of a castle ?

daurnimator you can use fstat I think to pass the ZFS to the jail, so the jail can control it. would you like an example?

I think I can get a dataset jailed and controlled from within a jail. The thing I can't is to have the dataset be mounted at the jail's /.

dautor8518050867 sure you can.

:O could you share the process, please?

happy 31st birthday freebsd!

i wish i were turning 31

part

dautor8518050867 sorry, I was AFK. so you wanted a Jail on a dataset, and a dataset that's mounted in the jail?

anyways I will repeat my query and try to be more clear, zfs pool encrypted with geli (which I believe is the default the installation media uses), the entire disk I believe is encrypted under zfs, so using an unencrypted home directory still has the FDE encryption protecting it, correct?

aka if I say no to zfs encryption, geli still will encrypt the dataset?

polarian you're not the only one young here :P

antranigv: younger than 20?

polarian okay you win. I am younger than FreeBSD, but not 20

:)

im sure there is that one prodigy sitting in this channel whos like 14 and got like 22 patches in the FreeBSD src tree :P

polarian altho my pre-20s were good, hitchhiking from city to city, I had a shitty laptop, had to go to a public library to chat with friends on IRC

polarian by any chance you're also Armenian? did I ask this before? -ian is the most common Armenian suffix in last names.

polarian to answer your question, yes, the home directory still has FDE with GELI, and ZFS would be running on//inside GELI

antranigv: no I am a Londoner lol

and thanks for the answer, just wanted to make sure... its just a server user... I am not going to store anything in the home directory (or well not much), its just FDE disks are easier to destroy when they die :P

less hassle destroying the platters

ah destroying disks, aka the hobby of btrfs

freebsd 14 unattended bsdinstall throws up a "rebooting" gui when it's done. 13 didn't do that. how can disable? because now every bsdinstall takes 10 seconds longer out of nowhere. not documented in bsdinstall man page

polyex does it install tho?

ya it works

gives 10 second countdown

nice feature for users but bad for unattended installs

in 13 you had to issue a reboot command in the installerconfig script. made more sense

antranigv those two datasets should be the same dataset

10sec times number of computers...500 computer would 83minutes more!

to push image out to computer network..

how fast is our grep? Or even more pointedly, how fast is our libc regexps? Context: several times this month I've gotten the feeling that it doesn't perform well.

regex(3) isn't great, but grep makes the problem worse

the former has some optimizations that grep could take better advantage of if we didn't insist on line scanning unless we really needed it

I'm watching grep parsing a 320G log file --- the end of a poudriere run --- and it's 100% cpu ... which I don't really expect should happen.

sigh.

help welcome

I'm ENOTIME to screw with it (it's not a fun problem to try and solve) and it's one that gets lots of complaints but nobody willing to fund the work necessary to make it suck less

yeah... putting on the list of ideas. Right now the top of my list is finding out why math/pspp is generating 300+ gig logfiles of crap, then looking at a new hanging problem on risc-v ... then finishing my new jobd/job pair (to enhance poudriere) ...

oh! you're the person from the hacker lounge

got a link to that work, btw?

I hear you. I said that I put it on my list. I like optimization. I just need to solve pressing issues first.

I don't use github unless I have to.

and I'm still working on it actively. It's not in sharable state yet.

fair enough

i try to do the same, run my own git forge and mirror to github for projects that someone else may want to interact with because I don't allow registration on my source of truth

Been distracted by the paying-job stuff ... like moving datacenters and migrating customers.

respect. I don't like some of github's terms. I also trust my own process. But sharing is good. I do offer logins on my riscv host --- soon to be upgraded as soon as the 550 comes out next month.

why moving datacenters?

do you use quarterly or latest PKG repos?

i use latest

probs happen in both but atleast in latest you arent stuck with it for 3 months

ye, I have one machine on quarterly, the other on latest

if /home/foo is "not a dir" (created as a symlink) why can i touch /home/foo/bar.txt and that works, but i can't run cp -RpP /tmp/foo/ /home/foo/? says /home/foo: not a directory

antranigv: you hate btrfs I assume?

also the urge to slap a bunch of platters in a plant pot and then thermite it is overwhelming :P

Btrfs is basically the GPL people saying "fuck ZFS and its CDDA license!".

No fuck, CDDL license.

Still early morning here.

funny enough I have thought about getting a furnace... dump the platters in it and melt them... free scrap matel!

remiliascarlet: GNU supporters, will always be, jealous of what non-gnu supporters have :P

btrfs has some major problems iirc... its why I refused to use it when I built my first server 3-4 years ago

btrfs is developed (since 2007) by the well known GNU supporters at Oracle :)

I am actually running ext4... yes I have no data integrity, it is RAID 1 but any corrupted files will remain corrupted

I am swapping my old server out for a new server I am working on as we speak :)

which will use a better OS (why do you think I am in this channel? :P) and use zfs....

although I do think most of the zfs features are overkill... I just need the data integrity :P

GPL is such a meme license. I know the intentions were good, but the consequence is that innovation is getting stagnant, nobody in free software apart from GNU wants to have anything to do with GPL licensed code, and the GNU people want to have nothing to do with BSD, CDDL, MIT, Apache, Zlib, and whatever other license that's not GPL.

I do wish there was a middleground filesystem

like UFS but with checksumming

Middleground filesystem, you mean FFS2?

FFS2 doesn't have checksumming still...

If you want guaranteed data corruption, just go with NTFS.

xD

Speaking of which, you might find this article interesting: unixdigest.com/articles/battle-test…adb-on-ffs-ufs-ext-xfs-and-zfs.html

Title: Battle testing PHP fopen, SQLite, PostgreSQL and MariaDB on FFS2, UFS, ext4, XFS and ZFS

polarian: does it maybe try to set permissions on /home/foo?

because of -P?

I still find it hilarious how old NTFS is and Microsoft has no plans to replace it in the future (or I haven't heard of any plans if they have)

eeh -p

wrong person

yourfate: yes wrong person :P

polyex: does it maybe try to set permissions on /home/foo? because of `-p`?

remiliascarlet: it would be nice if there was checksumming on FFS2... however from tests I have read people conduct, it is quite rare to even get failed checksums on small installs... bit rot isn't *too* common

Almost had the opposite happening to me yesterday: replying to polyex, but accidentally tagging polarian. But I noticed on time, so it didn't happen.

I believe an error happens once or twice a year... according to the sources I have read

anyone who runs a zfs nas can back that up or disprove it?

how often does a scrub fail?

by fail I mean, find checksum mismatch

The OpenBSD people have already said "no" to ZFS, but I do hope they will adopt HAMMER2 in the future, which looks like a better ZFS.

Filesystem support on OpenBSD is really horrible, the only way to actually get files transferred across operating systems is by having a FreeBSD-based NAS in the middle, and have all computers connect to it over an NFS mount.

remiliascarlet: I think ffs2 is much much more likely to happen on OpenBSD.

Which works, until you put Linux in the mix.

armin: No, it's not going to happen on OpenBSD, it already happened long time ago.

Since OpenBSD 4.2 to be precise.

My concern though is this line from the readme

> "This repository will be abandoned once Linux or FreeBSD is stabilized with write support. OpenBSD is not the main area of interest."

And it's not in OpenBSD so far from what I've seen.

openbsd needs zfs or hammer2

after going zfs i'll never go back to fs that doesn't autoheal

I guess my FS wants are: compression, snapshotting, checksumming

What sucks is that if OpenBSD crashes, the command you have to run on bootup is `fsck_ffs`, the default layout is US, and the location where the 「_」 key is on JP keyboard doesn't exist on US, so pressing it does nothing. And always have to look for where that key is on the US layout, which really slows me down.

remiliascarlet: AFAIK HAMMER2 is meant to be amazing, just unstable...

lol

Linux has that exact problem >:D

also I get why OpenBSD dislikes zfs... they believe in small maintainable codebases, zfs is a monolithic codebase

So autohealing is indeed a great thing to have.

amazing but unstable for a filesystem

ya i get it too, but they need SOMETHING

they need an answer for zfs

is like "this car is great, it just catches fire from time to time"

maybe hammer2

polarian: I think so yea, the OpenBSD spirit usually emphases much much much more simple things.

ppl buy porches

polyex: autoheal == more code, openbsd doesn't like that idea :P

I think FFS3 with checksumming and basic data rigidity would be a good compromise

also FFS is a good acronym

And perhaps FFS9 will finally have snapshots.

yourfate: xD great way to describe openbsd

remiliascarlet: never used snapshots in my life

my very short life but still :)

we have an internal project at work called the "waveform test framework". its abbreviated exactly as you imagine it would be.

polarian: I make snapshots automatically once every 6 hours on my FreeBSD NAS.

I actually love simple things. Software I can reasonably understand before I will die is good software.

And then auto delete them 48 hours after each snapshot was made avoid it from bloating up.

Me too, and simple programs are far less bug prone too.

I use snapshots for backing up the rootfs of my freebsd server every night

don't forget to rewrite ALL THE THINGS in rust

ofc

Here's a good 1 hour long presentation about how software these days is in decline, which I agree with quite a lot: invidious.jing.rocks/watch?v=ZSRHeXYDLko

Title: Preventing the Collapse of Civilization / Jonathan Blow (Thekla, Inc) - Invidious

I never wrote a single line in Rust.

I am C's waifu.

polyex: saving money. rationalizing.

in unattended bsdinstall installerconfig im trying to integrate zfs for my user to learn. having a bug pls: termbin.com/c3ba

what about fs dedupe and encryption

before WINE there was sun WABI, i had to give all my users a copy of winblows 3.11 workgroups nfs drive,...dedupe would have saved space i think.

sun WABI begat WINE

and grep awk and such probably faster before the utf8/unicode foobar, back in the ascii/ansi days.

c++ 2011 unicode support i want to say...

perhaps benchmark grep/awk with different "locale charmap"

in unattended bsdinstall installerconfig im trying to integrate zfs for my user to learn. having a bug pls: termbin.com/c3ba

why you using -P when default for -R is same.

-P No symbolic links are followed. This is the default if the -R option is specified.

you already gave the -R

so you dont need -P

it worked in another place where i did a similar thing to this. but that was /jails, instead of /home/polyex. not sure why it's acting differently

ill try without it

perhaps just rsync the foo

no clue

need help getting zfs and unattended bsdinstall working together pls termbin.com/mj5o

i guess it fails because the destination of the cp can include a symlink in its path, but it can't be the last item. weird