Macer: It will work, but it will almost certainly cause a need for rebuild/scrub (depending on if you're using those identities for ZFS inclusion.)

Shouldn't cause any problems with the disk(s) or data, though.

Macer, I don't know about the path property but I have changed the paths by importing the array using the desired path. For example by booting install media, getting a shell, and then "zpool import -d /dev/gpt" so that it uses the /dev/gpt paths to the persistent gpt labels rather than the not persistent /dev/ada0 or /dev/da0 type of paths. And then once imported and changed then I have rebooted and the change persisted that way.

I'll make a note that there is a zpool path property, though I can't find my information about it, and try it the next time I am in need of making that type of change.

Macer, Alan Jude says that setting path works and that's authoritative: youtube.com/watch?v=Yy8fhyrBV-Y&t=155s

rwp: yeah i was about to say

there is a way to do it on the fly. i've done it before :)

i mean it isn't like the disk disappears. the label does. zfs probalby just finds it with the tag and renames it to something and then you can just rename it later

no scrub/rebuild necessary

i think it was when i had imported a pool from elsewhere and it used some other way to name them. but i vaguely remember being able to change the gpt/label on the fly for the pool

rwp: i'll test it on the pool i'm going to dismantle and see how that works

Macer: how did your operations go? did recovery work?

o\

Macer, The FreeBSD kernel does something that other kernels don't. As soon as any of the /dev device paths is opened to a device then all of the other alias paths to the same device are hidden.

Which means that if /dev/ada0p3 and /dev/gpt/mygptlabel point to the same partition then if/when zfs opens /dev/ada0p3 then /dev/gpt/mygptlabel is hidden and no longer visible. It actually causes more confusion than it helps in my not so humble opinion.

Setting the zpool path doesn't actually change what is open now. It sets the path that zfs uses to find that partition when it is next imported. So right now setting it to a different path won't actually find the newly set path because it is hidden. But upon reboot and booting up then all of those paths will be visible and it will open the newly set path and again the other aliased paths will then be hidden.

Hope that description makes some sense of things.

SponiX: if you’re asking about me expanding the raidz2 vdev… it went well once I didn’t have sketchy disks. I expanded twice already.

And I guess it’s good to know that if there is a failure during expansion then zfs handles it well.

Macer: well congratulations on the successful operations :)

Macer: so did you have to replace faulty drive first and then resume the expansion, or just replacing the drive was enough to make the whole expansion sucessfull ?

has anyone successfully booted the recently released alpine 3.23 (linux 6.18) with bhyve? I'm getting a hang at "Loading initial ramdisk...". Alpine 3.22 (linux 6.12) works fine.

i used to run: tmux new-session -s sess 'mybin > /home/me/app.log 2>&1; exec $SHELL'. but then i wanted to make sure all errors were redirected to app.log so i changed it to: tmux new-session -s sess 'exec > /home/me/app.log 2>&1; mybin'.

but now the problem is when mybin terminates, the tmux session doesn't terminate. so how can i have all errors redirect to the app.log, and also have tmux session terminate when mybin terminates?

makes me wonder whether you need an interactive terminal session or whether you can just daemonize it and have daemon redirect the logs however you want

:)

eventually but for now i'm gonna keep using tmux

mzar: I had to replace both of them.

The expansion just halts. Even on 1 read error.

And the read error just wouldn’t clear. It’s kind of weird because I figured after trying to clear it a couple times that the read error would disappear and the drive would mark it as uncorrectable. Instead after I tried to clear it a few times the drive just fell off the pool and was kicked out.

uncorrectable is not very easy to get set unless you use MHDD or the like

thanks for the report Macer

installing stable drm gives different versions see if any help

removing the modesetting driver makes the screen not tear up

what would the difference be between: tmux new-session -d -s sess 'mybin &> /home/me/app.log' and: tmux new-session -d -s sess 'exec mybin &> /home/me/app.log'?

kerneldove: isn't that nohup with extra steps?

don't know

kerneldove: that's probably beyond of the support capabilities of this channel, please try to inquire for more details at #tmux

Remilia: oh. I figured drive fw would try a couple of times and give up.

kerneldove: when you run something with exec the parent process dies and whatever exec is executing becomes the parent, using no exec runs that program as a normal shell command

so if ps youll see sh bleh vs just bleh

Macer: generally the firmware does not mark anything uncorrectable while the drive is ‘online’

they are called ‘offline uncorrectable’

I do not remember how exactly it works but iirc software like MHDD uses ATA/SCSI commands to directly tell the drive to mark a sector, after verifying that read times on that sector exceed 300-500 ms

cpet, ah ok so i should use the version with exec if i want mybin exiting to terminate the tmux session, right

?

Yes, a

hard drive's firmware automatically detects and marks bad sectors as they occur, remapping them to spare sectors to keep them hidden from the operating system, a process called reallocation. This happens during write operations when error correction detects failures, and the drive uses its built-in reserve pool to replace the faulty physical location, preventing data loss and OS errors until that spare pool

runs out

kerneldove: any shell command that exits will cause tmux to temrinate that session

everytime HD bad sectors are mentioned it reminds me of the badsect program we had long ago

i haven't seen that behavior *shrug*

kerneldove: i run fossil in tmux if it shuts down the session shuts down

i have never seen it go back to a shell prompt

and the scripts isn't that complex

tmux new fossil server --baseurl fossil.unix.beer --nojail --repolist /var/www/fossil.unix.beer/

kerneldove: I remeber one time when we had a long discussions about running exec x in .xinitrc vs without

Remilia: wonder how well that would work for an smr drive. I guess those read just as fast as cmr. It’s the writing that is problematic.

So… terminate on exit?

Macer: no

it's exec() vs spawn()

exec() means the process that called it exits immediately and then the new process is started, possibly with FDs and environment etc. passed

if you `exec ls` your login shell terminates and once ls finishes you're back to login

(if you do that from your login shell that is)

if you have a screen or tmux session open, open a new window and `exec ls` and watch it close immediately

🤔

I’ll have to try that out.

How does one get UCL output from FreeBSD commands? I ask because I'm told pkg has UCL, but nothing in the man pages.

dvl: im not sure it outputs ucl, but pkg supports ucl in pkg-message, etc. files

zi: Oh yes, that'

s where Ive see it

cpet: yeah, found that page when searching for freebsd pkg ucl

the other idea is using lua, built int.

the only output commands can do is xo

cpet: You refer to pkg-message there? I need pkg-repositories

local ucl = require("ucl")

local parser = ucl.parser()

parser:parse_string("key = value; num = 100") -- Parse UCL

local obj = parser:get_object()

local json_string = ucl.to_format(obj, 'json') -- Convert to JSON

print(json_string) -- Prints: {"key": "value", "num": 100}

theres lua to json

er ucl to json

so you have the ucl parser that then concerts that to json

for lua

there ya go bill in the mail

you know that you will get quicker calls for help if you offer some $$$ heh

Thank you. I had something similar in that URL.

I need to modify it to read from STDIN and output to STDOUT. So I put out a call: bsd.network/web/@dvl/115708352504899727

I may be able to research this over the weekend. I've done similar with Python code, but I'd like to not add Python to the dependencies of this project.

libucl offers pretty much all that

Good!

General syntax sugar

heh

I find no docs there. For example, I want to know more about parser:parse_string() because that seems to be the function call I need to modify

cpet: I got it, bin.langille.org/?f884b1be1566e139#…grip9JLWgH88KaZFpEYdd695mqP4ANqUFFn

weirdo

So from there, where?

bora bora

Uncle Osama?

uncle sam?

Sam Drucker

I have so many trails, I'm going to review them another day.