zpool is the volume manager (and software RAID) part of it, and zfs is the filesystem part of it.

V_PauAmma_V: where I learn it?

can peole not google any more ?

man zfsconcepts

the whole zfs documentation

you can ask your phone to do it for you heh

hey google what is the difference in zppol and zfs

openzfs docs need a lot of work

cpet: I want to learn, not be converted...

converted to what ?

tell me what is the difference in ask google than asking in here ?

Unlike Google now, I'm in the business of providing accurate (and if at all possible usable) information.

first result was a link to oracles docs

probably more accurate

See above, "usable".

V_pauAmma_V: thanks for the man zfsconcepts I didnt know: there is no stupid question!

Chip1972: apropos x will give you every man page with x in it

(I don't think snapping at someone for not using Google is productive.)

if you dont want to google

"(I don't think snapping at someone for not using Google is productive.)" I agree

arch users are migrating to fbsd?

why does it matter ?

How are you cpet :)

i have no beer for turkey day

oh, broke?

no people dont want to work holidays any more

haha!

so I didnt know rouses closes at 2pm today

Chip1972: where did you ear such thing ?

since searching google is bad, i do not know as I cant google

:(

i like it when people move form linux to freebsd they complain that wayland doesnt work right :)

again not googling at the fatc that wayland support is there but behind a bit

wonder when that will be available

wow. long smart tests still going on these disks

yeap it takes a while

i think when it started one of them said 1000 minutes

zfs in itself does a pretty good job at detecting issues and fixing them

we used to have a program called badsec which would remap bad sectors

yeah but i haven't added these to the pool yet. i wanted to make sure they'd pass the long smart test before adding the 2nd vdev

heh

now the firmware does that

yeah i thought hard drives do that on their own

i would of just added it in :(

lol

if smart says the drive is good

then hell if im going to wait 14 hours

well... smart hasn't said it's good yet heh

they're down to 20% left .. so i can wait it out

smart does its own small little tests to see if the drive is good or not

i need to yank them to make sure i have the serial numbers and positions right in the spreadsheet anyways

based on what the results are

yeah I wouldnt last that long

added drives, add drives to spread sheet goto break room and dirnk coffee

heh

i'm not really sure what it does for teh short testing

but i figured the long test scans every sector

in one sitting end to end

i had a NAS that said a drive was bad i rebooted it it then said it passed

so i dont trust that smart stuff much

once i get this vdev going in can start ordering an 8TB drive a month or something and expand it until it gets 12 wide like the other vdev

hopefully without destroying the pool lol

attaching a drive should cause any data loss

most places i'm seeing say 4 disks are required to start a raidz2 vdev.. i always thought it was 3

but i guess that would be awkward

4

i hope you mean shouldn't :)

what mnakes sense to you ?

should or shouldn't ?

attaching a drive shouldn't cause any data loss

yeap

well for now i'm going to have mismatched vdevs .. not sure if that's a problem with zfs. i'd think not but there may be science in it

nope

1 vdev will be 12 8TB raidz2 and the other will (for now) be 4 8TB raidz2

yeap the minimum

so i have something to scale off of .. hopefully larger drive prices come down

i'll have 12 bays to work with so i could just make a 20TB x 12 pool or something and get rid of all these old disks

9 Power_On_Hours 0x0012 087 087 000 Old_age Always - 93576

the 4TB drives i'm running are troopers

there is a ZPOOL and ZFS tutorial for noobs?

There's a chapter on that topic in the user handbook. docs.freebsd.org/en/books/handbook/zfs

# 1 Extended offline Completed: read failure 30% 1255 -

hah

so it failed

one of them did at least

hmmm I should prolly start replacing my crappy old SAS drives

though it's a RAID-Z2 and has a hot spare so meh

what is difference between mount ZPOOL and mount ZFS?

As far as I know, you can't mount a zpool, only a zfs (which can be the root filesystem of a zpool). Why do you ask?

according to 'ZFS for Dummies'

Use the -R flag to mount the pool to an alternate root location

# zpool import -R /mnt/tank2 tank

This is actually mounting the root filesystem of the pool.

(in addition to importing the pool)

what is difference between mounting and importing?

cpet: you know what people are more inclined to do than searching on google for answers? Asking ChatGPT. :P

Importing (and exporting) is what lets you move pools between computers or VMs. Mounting is what makes a zfs (which again, may be the root zfs of the pool) visible somewhere in the directory tree.

can I mount whithout and importing?

You need to import the zpool in order to mount the zpool's datasets.

(datasets being what I not-quite-correctly called "filesystems" above.)

how fstab import the zpool?

I think that on multiuser startup, some script in rc.d/ will automatically import all zpools that the computer or VM has access to.

So you wouldn't need to touch fstab, if that automatic importing is what you want.

shame on zfs docs...

That "ZFS for dummies" is 5 years old, judging by the URL. It may have been correct then, but OpenZFS changed meanwhile.

I'm reading the handbook page, is that up to date?

is like trying to build a wall, and only find documentation to build a skyscraper

JetpackJackson: I seem to remember that most of that chapter dates from 2022 or 2023. So nearly everything in it should still apply.

Alright cool

Makes me wish I had paid more attention in my vm install with zfs cause I just picked the first one (stripe, no mirror, I think?) lol

Gonna see if I can install on one disk, add more disks, and then change the stripe-y stuff

That's very easy to do in VMs. I've added disks (well, partitions) to a stripe several times when my first disk turned out to be too small.

Ah ok. Want to be able to add disks freely to my upcoming media server setup so I wanna prepare by figuring out how it all works

# 1 Extended offline Completed: read failure 10% 31134 -

hm.... 2 of the 4 failed the smart test

makes me wonder if there is something else wrong

Macer: you do know that zfs will detect those bad sectors and mark them so ?

zfs is a pretty resilient fs

i've had it break several times, it's quite fragile. but it can detect bad sectors and refuses to knowingly return incorrect data

i guess it depends on the drive as well

Macer: is it procedure to scan the drives first or are you doing it as a good employee ?

expand: expansion of raidz2-0 in progress since Thu Nov 27 22:35:46 2025

ah well.. 2 of the 4 disks failed the long smart test so i'll just pass on those two and just expand the existing vdev 4 more disks and work on making another 16 disk vdev later

i considered making the vdev 24 disks wide but that is kind of pushing it

LxGHTNxNG: ive done some weird things with zfs and never had it break but then again i install the OS and leave it alone

so cool you can ecpand them now

rtj: freebsdfoundation.org/blog/openzfs-…on-a-new-era-in-storage-flexibility

took half my turkey meal and froze it for xmas and new years

cook once eat on 3 occasions

cpet: yes, thats what i was talking about

rtj: the handbook says not to make pools bigger than 9 drives

good to know. i cant afford that many. no need to worry here. ;)

yeap and that is why my zfs pool include a mix of nve ssd and HD

heh

not smart not efficient but it works

15R coming soon!

soon !

i keep refreshing download.freebsd.org/releases/amd64/amd64/ISO-IMAGES/15.0

lo

l

maybe you should go an eep

hernan604: you do realize that the actual release date is set to Dec 2 ?

and that too can change

releng/15.0 was created so you can upgrade now using src to 15.0-release

cpet: yes, but the iso always land there earlier

(than the expected date)

hernan604: if you really wanted it you would just compile the src

4.03T / 69.1T copied at 223M/s, 5.83% done, 3 days 12:59:26 to go

uhm....

so it was going rather well... now it almost seems like it completely stopped its expansion reflow

cpet: nah, im just looking for fun

uh oh

this can't be good

it's mounting the local filesystem after i decided to reboot it and now it's just sitting there

ok. it finally started up. that would have sucked

not the end of the world but still sucked.. but yeah i'm guessing because this drive sucks and isn't a nas drive that it's just super slow

its a HDD the max it will go is SATA 6 speeds

which is 600 mbits/s

4.03T / 69.1T copied at 223M/s

that's like hdd maximum

pretty much

lets all add in terabytes of storage the complain when it takes to long at the max speed of the sata interface

there is no 70t hdd so raid i guess

of 2 disks :p

i love how far we have come

bro that's pretty much the up limit of a new SATA3 disk ketas

cpet: oh. no. it was actually not going that fast.. that speed was slowly declining and zpool iostat was showing very little activity

maybe i should have been more clear

200+MB/s something

the equivalent of 2GbE basically

Macer: what are you transfering?

i'm expanding a raidz2 vdev

yeah it's pretty sata hdd max

esp if random access too

maybe it was the random access when it really got slow. zpool iostat was showing like 2MB/s for the drives

yeah

after rebooting and having it resume it seems to be going faster again

i bet it's random anyway

that's weird zfs behaviour maybe

not sure... i guess i'll just wait it out

my personal favourite of raid is 0 + 1

but it's not like you keep rebooting during disk ops to get +10mb/s extra oomph

parallels in series basically

yeah lol

that wouldn't be practical. i did it because i figured something was wrong.

unsure if i would even want 0 part of it

scottpedia: that pool is for raw storage. speed isn't that serious.

but in some cases you need it

if you want to write down 100t file it doesn't fit

:p

ideally speaking they should have some kind of a statistical analysis of how often a drive goes haywire

what were benefits of raidz?

speed?

i'd say space

aka what kind of importance constitutes the need for a raid 1 setup

space+redundancy

oh maybe

read about apple's fusion drive some time ago

here i see myself doing 2 hdd mirrors

they say it's kind of good cause it's a tiered storage system

dunno how well that could perform in practice

managed to crapperize myself a bit with singles

in my case i'm going to expand this vdev to 16x8TB then move onto making another 16x8TB vdev (or larger maybe) later ... i'd be a little nervous making a raidz2 vdev > 16 drives

so don't do that

or, do, if it's ok

lol. at least with singles if you really screw up you can just mirror on the vdev to work your way out of that even though you have to go 1:1

i've seen people who accidentally added that single disk vdev to a pool lol

you can mirror it out

actually zfs needs more general features

the only time it probably matters a lot is when people try to add slog or l2arc and do it wrong and add a 128GB nvme or ssd to the pool lol

one dreaded one is to replace device with new smaller device

does zfs even let you do that?

it has... hacks

you'd think it would spit out an error if you try to replace with something like half the size of everything else

but no

i think they have that new thing where you can add different sized disks .. not sure if that's in 2.4 or not though

i have my few pools too large and sane idea is to replace with larger disks eh

that's probably something i would never attempt. i guess i can see the use case of throw anything you can at the storage due to financial constraints

different size disks always worked

yeah but you'd have to wait until you replace them all with larger disks for it to expand.. i guess it's easier with mirror vdevs

in mirrors at least

yeah. depending on yow you do it you just have to replace 2 at a time to expand the space .. that's a lot of space to give up though. i'd probably take that route for super important production storage

and have 3 way mirror vdevs lol

yeah why not

someone told it i was wtf but

it's just a disk

lots of read speed at least :)

do zfs mirrors take speed of the drive into consideration? like if you had a 4TB ssd and a 4TB platter would it just favor the ssd and let the platter trot along?

i never tried

apparently it's mixed result it seems

i have a 4TB SSD sitting on a table and a 4TB platter .. i should try that

just to see what happens lol

need guard space i guess

i'm curious if zfs adjusts for that

i missed it on disks

damn

i highly doubt 4t hdd is byte equal to 4t ssd

that can be adjusted with partitioning

yeah

but it should be the same (in theory)

i don't think they use a different standard for sizing

never know eh

i'd guess that zfs would adjust to the slightly smaller drive size regardless

i have 4 same hdd's here, but also diffrent ones

i had brainfart and forgot the part size

:p

ah. that's why i just copy the partition tables from the other disks

so now that pool stays that

then gpt label them

back in the day when i used freenas figuring out where broken disks were was a mess .. they identified them like 3 different ways .. zfs would add them one way but the web ui would show them differently

the best thing i ever did was just install vanilla fbsd on the nas

unsure why all 4 12t hdd's from 3 makers are byte equal tho

and what's max diff?

they should be

in old ones i see 200m diffs on 160g

maybe there is some sort of requirement for it nowadays?

could use public smart db's to do stats :p

like it has to be x +/- 1MB

well there are no hdd standards iirc

on size

or maybe everyone moaned

and they were fffs ook

there sort of is... like 1000GB = 1TB :)

so if they market it wrong then i'm sure that would be a big issue

1000 vs 1024 is another fun too

kind of like how WD didn't let people know they were pawning off SMR drives

yeah. i was a little ticked off about that change.

change?

disks were always sold by 1000

i mean.. 1024 USED to be the thing.. probably not with TB but earlier disks

it was?

like if it was a 1GB disk then it was 1024MB

as far as i can remember.. yes .. other than what you lose formatting

at least i don't recall anything like that >=1999

and people have always moaned how disk is too small in os

:p

yeah. they started doing that in the 2000s or something when disks started becoming larger because if they change 1024 to 1000 then they can market it with more space

I always get mixed up when different programs show GB vs GiB

ounces make pounds i guess

gb / gib is fuckup

it's great confusion

at many times it doesn't even matter

remnant of binary

well.. i THINK most filesystems still use the 1024 method whereas disks use the 1000

if 78g didn't fit but 72g did it's out of space anyway

i'd have to check on that though

df -h / df -H examples

:p

and i'm sure changing that would involve a major undertaking because i doubt that filesystems can come away from using binary :)

i'm sure when quantum computers become the norm it this will all be moot and filesystems can just be in all states at one time

shrodinger's filesystem

you can cat from that

:p

*schrodinger's

the files can be deleted and exist at the same time!

wait... i think they do that already

Schrödinger's Scheiße

ah well.. let me start doing stuff. i have 2 days and counting until this raidz expansion is done lol .. guess i just need to keep the thing running

on a side note. is RELEASE still slated for today?

oh seems so. the 15.0 release process says they started building RELEASE today

hmm

hopefully it

it's all fixed now

I thought it was for December

i mean lagging release using is ok

because apparently devving testing isn't enough and going prod is like wait what

Ah

JetpackJackson: i'm just going off the 15.0 release page ... the RC builds were available the same day or a day or so after

Ah

Cool

i think the announcement is for dec 2

but i'm not sure if that's also when they let it out into the wild

9.65T / 69.1T copied at 281M/s, 13.96% done, 2 days 13:40:35 to go

it seems to be picking up speed

you have collected a lot of data

87.2T 69.1T 18.2T - - 3% 79% 1.00x ONLINE - i guess the expansion doesn't show until it's done 'reflowing' the data?

this has to be pretty intense for the disks.. lots of r/w action across the entire pool

oh yeah !!! what's up freebies

Hi

:)

great. had a drive die while expanding lol

go figure

it broke so badly that zfs just kicked it out the vdev

12.2T / 69.1T copied at 259M/s, 17.65% done, paused for resilver or clear

Macer, I don't know if I should feel sad for you or not. You were just running a testing system exercising the pathways, right? This was a testing system and not an actually in use system, right?

sup

the system

15-RELEASE: download.freebsd.org/releases/amd64…BSD-15.0-RELEASE-amd64-memstick.img

so just replace the vdev, what's the big deal?

hernan604: Nothing's official until the announcement comes out!

hernan604: it's not "fully" official because there can be some last minute rebuilds with fixes until the announcement. Think of it as a "beta" to the release until it gets the green flag.

^