where we go: shrinking a zroot zpool by sending it to smaller drives. dan.langille.org/2025/11/22/test-ru…ot-using-zfs-snapshot-and-send-recv

FreeBSe 14.3

heh

JetpackJackson: looking at ports I found prusaslicer

based off of sli3r

same shit orcaslicer is based off

Yeah I guess I can use that as a base for my port

Hmm they have instructions for how to build in docker

i got it to build past boost

With docker? Or the scripts

cmake build

Oh duh how'd I miss that

Thanks

i fixed the boost issue by removing system from the list

i install tbb

but still fails to find it

Hmm

just have to play with where cmake finds things

linux expects evertything isn /usr/

we expect everythig in /usr/local

Had to fix my VM cause it ran out of room

So Happy that zfs can resize

I kinda wanna get mindustry to build too

zip: I love ed and still use it a lot. I use rlwrap enable vi-mode line editing. I only resort to vi for extensive source code editing as a last resort.

Then again, my shell (and libedit itself) is set to use vi mode.

I wrote/updated a large part of the vi-mode functionality in tmux because I use it so much. tmux+rlwrap makes ed a lot more usable.

it's a powerful and elegant program

topcat001: you using a custom plugin or do you use the vim-tmux-navigator one?

no, my tmux setup is fairly simple. I `set -g mode-keys vi` basically. The vi keys are most handy in copy-mode and : cmd line editing.

one could argue that I cheat by running ed inside tmux and rlwrap :)

I am excited about version 15 and that's so strange to me, normally I'm not excited about system updates lol

15.0-release will be the holiday gift to the FreeBSD community

ya im excited

big moves in freebsd world. pkg base and oci support

kerneldove_: what is pkg base

instead of managing the base system via freebsd-update and other tools, there is a pkg repo for stuff in the base system and you can decide more fine grained what parts of base you want

splitting the OS itself up into individual pieces that can be independently versioned and upgraded

Oooh

That sounds cool

kerneldove_: that sounds like something my native boot proposal'd be able to help out with

87.2T 68.9T 18.3T - - 3% 79%

starting to get dangerous heh

is the 80% thing still a thing?

Upgraded to 90% or so

oh really?

is that when i'll start seeing massive performance loss?

You should still have an upgrade (or data deletion) path planned by now

yeah. i'm already working it out in my head. i mean it's a personal pool so it isn't THAT serious ... but i was going to make use of raidz expansion and add 4 8TB drives as another vdev and grow it from there

(as raidz2 on a separate vdev)

i'm just going to wait on RELEASE before i try doing that just to make sure fbsd 15 is settled

i'll be a little limited once i upgrade the pool

You could also just build a new server with new disks and have the current one as a new kickass backup server

This

This "move zroot from larger devices to smaller devices" is now ready for production: dan.langille.org/2025/11/23/test-ru…ot-using-zfs-snapshot-and-send-recv

isn't this easier to just change openzfs to allow that mode

to avoid people doing stupid hacks

ketas: what do you mean by "that mode"?

shrink zpool while replacing devices

ketas: It does support it, see what I did here: dan.langille.org/2025/11/22/moving-…maller-drives-with-uefi-zfs-freebsd

that leaves that questionable map

ketas: two problems I have with that approach: 1 - it mangles the gpart settings for the old drives, meaning you can't just boot from them again. 2 - there is a slight, and I mean trivial, cost associated with a table to redirect from the old stuff to the new stuff.

ketas: I did several runs with both approaches. The send | recv approach sat best with me.

how come there's no "table" when you put larger device there

also why any outside of zfs thing matters

any gpt etc

ok, how does zfs replace disk anyway?

it only copies data, i understand

ketas: moving to larger devices usually involves `zfs replace` - shrinking is different, you add another mirror, in effect you start striping. Then you remove the original mirror, zfs evacuates all the date to the new drives.

ketas: My specific use case is the zroot pool, which involves partitions.

so if i have 100m data on 1g pool and want to replace 1g device with 500m one, and i can't then why?

does it write it to the +800m mark or what

ketas: who says you can't?

on new device

ketas: In my tests, I went from 240G to 30G

and from 240G to 128G

the table in question assumes that it has some immovable data there

or i don't know eh

ketas: I can't help you there, I just know the zfs folks seem to know what they're doing.

"the table" ?

rtprio: openzfs.github.io/openzfs-docs/man/master/8/zpool-remove.8.html mentions a mapping table.

See "Removal of vdev 0 copied 452K in 0h0m" at dan.langille.org/2025/11/22/moving-…maller-drives-with-uefi-zfs-freebsd

dvl: I can see it worked, but I just cant comprehend that zfs can allow removing a vdev from a stripe... Or am I misunderstanding something?

I am planing to learn FreeBSD.. B$$ companies are using it.. Have to find out that whats include in it which Linux doesnt have .. Freebsd and Linux kerenel is Same ?

no

hmm.. so Linus Torvalds kernel work is only for Linux.. well some one game a very good example.. Freebsd is a Complete Car and Linux is only an Engine.. which need body parts for making it a complete car.. and specially freedom from License point of view.. thats intereseting..

Tingo: you can find a lot of info online on the differences

yes.. Reading it .. and its interesting to know more about Freebsd.. well, Thanks!!

You can think of the Linux kernel development as an autocracy while FreeBSD development is more democratic :P

kenrap: it gets old having to explain the differences when it can all be done with a few google searches

cpet: yeah. I thought I would throw in some "fresher perspectives" rather than "Linux is just a kernel and single component while FreeBSD is a complete and cohesive OS"

kenrap: pretty much what it is

i am working with a dataset storage/postgres which contains a database.. that is running in a FreeBSD jail.. when I make changes to the zfs dataset, say zfs set primarycache=metdata storage/postgres do the options on the server "trickle" down to the freebsd jail?

i have been looking over the internet and it talks about how the dataset will take the options.. but i am not seeing anything if i have to worry about the FreeBSD jail aspect, my gut is telling me no.

voy4g3r2: normally it will take on the option of the main dataset so if you have /var and you create /var/slaps and you dont add any options it will copy over the ones form /var now if you set options to /var/slap they wont revert back to /var

is that makes any sense :P

that is the inherit option

correct?

i think it does, if my statements are correct :)

voy4g3r2: yeap

okay cool.. so my changes to that dataset and not pool are working.. it is saying local when i am making modifications

voy4g3r2: are you making changes to that zfs or its parent?

damn default webfonts dont have Time new roman

rtprio: to the dataset itself

another example is making sure the blocksizes are 8k :)

reddit.com/r/freebsd/comments/1iven5k/reflinks_in_freebsd oooh so now cp will reflink by default if pool supports block cloning

im running into an issue here. I am using 32 bit netmask addresses for vnet jails and am pretty sure I have all of my static routes set up. On the host side I can ping and traceroute to the jail. However, on the jail side I seemingly cant traceroute, but I can ping the host. the jail also cant traceroute its loopback address, which is very strange. any ideas?

I should note, I have this same setup on another box and it works perfectly as expected.

are you able to paste the setup somewhere.. so we could take a gander?

like a pastebin type thing

yeah sure, what specifically do you think you'd wanna look at? I can post a few things

do you have pf enabled by any chance? anything in the logs?

the whole thing

i've been meaning to implement a proper pf ruleset for a server + jails for a few years but have not made the time yet but its on my radar so this sounds very relevent :)

nope, no firewalls should be enabled. this is a fresh freebsd install im working on. gathing stuff

allow.raw_sockets=1 ?