can freebsd/ufs handle bad sectors?

hmm

I should just say what is happening instead of guessing

my hdd was dirty, I do not know why, I separated /home and made it its own slice, now I reboot and its dirty

reboot in single user mode and fsck the partition i would assume.

should have chosen zfs..no fsck needed..

fsck 100's drives from crash is long boot process. whole point of zfs was to do away with fsck foobar, and CRC checksums.

bad harddrive bios, bad raid firmware..yeah..zfs

why jbod and not hardware raid

it says lower ram than 8gb to not use zfs rennj

ECC memory. This isn’t really a requirement, but it’s highly recommended.

8GB+ of memory for the best performance. It’s perfectly possible to run with 2GB or less (and people do), but you’ll need more if using deduplication.

Without ECC memory rare random bit flips caused by cosmic rays or by faulty memory can go undetected. If this were to occur OpenZFS (or any other filesystem) will write the damaged data to disk and be unable to automatically detect the corruption.

go up to mountains of colorado get more bit flips

internet is filled with stupid now, best thing to do is test it yourself..

en.wikipedia.org/wiki/Sun_Fire_X4500 thumper

you know back in 2006, i got thumper vm here in vmware

i also had solaris 10 beta machine..what would i know..

imgur.com/p95rKOf solaris 10 beta tv card working. even had tv-card working

and the 90's was more compatible..haha compiling software was easier then today...but keep arguing with me. channel.

ports no shit

heh

hp-ux irix solaris linux freebsd ..was more compat back then

but keep telling me how things are better today

UFS has worked for a long time for me. it seems I don't have journaled soft-updates on this slice...

maybe that's it

yeah you should enable it

nothing wrong with ufs

if you had big old disk array then i would say something..

1hour boot time on mission critical stuff cause fsck all the drives...

255 drives in fc-al loop

dell emc gear

and linux in 2004 was joke on that gear compared sun or hp...

megaraid driver..or whatever..damn you to hell

linux is lucky ibm dumped 1billion into it in 2000 and 2013

fight off m$ stupid

$2billion, and now ibm owns redhat

zfs zil for writes,l2arc for reads

hybrid storage array with ssd and spinning rust and lots of ram...

that was thumper

i.imgur.com/fVNpK.jpg i.imgur.com/BJGGz.jpg i.imgur.com/px0V8.jpg, thumper vm, which virtualbox could not run, but vmware player no problem

and sun bought virtualbox..and could not get it to run on it

haha

but vmware could run it

i have the old links to blog post about that very issue...

2011

core2duo with 8GB of ram

you want compatible netbsd pkgsrc

thats the best we got right now

lets check repology.org

nix (nixpkgs unstable) - 97110 vs FreeBSD Ports - 30986

currently containing over 13000 packages. pkgsrc

pkgsrc.org currently containing over 26,000 packages. yeah i think that a lie

smartos people need to fix that error

pkgsrc currently contains over 22,000 packages and includes most popular open-source software. It is the native package manager on NetBSD, SmartOS and MINIX 3, and is portable across 23 different operating systems, including AIX, various BSD derivatives, HP-UX, IRIX, Linux,[4] macOS,[5] Solaris, and QNX.

wikipedia

but fanboys, zealots will argue..

uskerine: zfs has many nice features, like checksumming everything, transparent compression, cheap snapshots, ...

check summing without ECC is cosmic ray foo

better then nothing but still error prone

Reed–Solomon error correction

while having ECC ram would be nice, not having it has the same risks on all filesystems and if a random bit-flip on read happens, zfs detects that because of the checksum

i posted the openzfs page notes

Without ECC memory rare random bit flips caused by cosmic rays or by faulty memory can go undetected. If this were to occur OpenZFS (or any other filesystem) will write the damaged data to disk and be unable to automatically detect the corruption.

see openzfs page

and DDR5 jedec has error correction but you still need ECC ram parity chip/data line

perhaps you need to read the openzfs page

cause you are spouting lies

rennj: yeah, that is the write case, were stuff can go unnoticed, but in the read case zfs reads some data and its checksum and if they don't match the read error counter gets increased

or probably the checksum error counter, read is probably disk errors

Holy wall of text.

what's the use case for `reboot -r`?

AllanJude: Hi, I was wondering is there is any work underway to use ZRAID expansion with more than 1 drive. Can't remember if or what the reason was for only 1 drive at a time.

mostly complexity. and the use case doesn't usually involve more than 1 drive at a time

like, if you have 4 drives, and are moving to 6+, you might be better off making a new pool instead

makr: to change your root filesystem without actually rebooting (because the BIOS etc part of boot is the slow part)

but no, there are no plans to make RAID-Z expansion support more than 1 disk at a time

yeah, but in my case for example, I have a -Z2 with 6 drives and want to add 2 more.

Just my use case, I know :-)

I use Toshiba N300 drivers. They support 8 max in a bay. So can't even add 6 more in one go.

s/drivers/drives

AllanJude: thanks

weust: that limit is only re: vibration, so if you have good caddies it is likely not an issue

but yeah, you can just expand twice

Oh, storage talk?

weust: I like the M09 Toshibas

SAS

Guess they are actually MG09

AllanJude: They are in a Dell R730xd. I has 12 bays in the front, so I would assume it's OK....

SponiX: I already have 6 N300's so will stick with them for now.

Also, since mine are SATA I can't mix with SAS

weust: Yeah, those are good Units (R730XD). And those drives generally do well also

My current drives are 5 to 6 years old now. All fine

weust: double check that, my disk shelves and a lot of machines have SAS that can take both, even mixed together

Not saying you have to go SAS. I just often find them cheaper on eBay than SATA

But, I only store remuxed movies and series on them, so write is very low. Read not much either.

Dell specifies with their controller it's one or the other.

And I have two 2.5" SSD's in the back too, for my music and other data. Also SATA, on the same controller.

Sounds like a good setup for what you are doing

Just a home NAS :-) I got the server for free at my last job. I was being decommisioned. Was only used for some light backup for VM transfers between clusters.

So I was lucky there.

cool

The drives are 12TB each, but it's getting full. Those 4K UHD discs eat disk space for breakfast.

I have 3-5 or so N300 drives - They have been doing well for YEARS

I took the 12TB specifically because of the low Wattage in idle

and low noise too in idle.

it's in a room in my house, so needs to be fairly quiet.

My system is in the bedroom

Even with ipmitool to limit fan speed that wouldn't be enough hhaa

I have a fairly large fan blowing across the stacks right now. It makes enough noise the drowned their sound out

And you sleep with that noise too?

Yes

my home HVAC system is louder than all of it combined lol

That is wrong on multiple levels

System usercontent.irccloud-cdn.com/file/Vcl3vCLF/IMG_1726.JPG

Client: HexChat 2.16.2 • OS: Fedora release 41 (Rawhide) • CPU: Intel(R) Xeon(R) CPU E5-2696 v4 @ 2.20GHz (1.20GHz) • Memory: 245.7 GiB Total (203.8 GiB Free) • Storage: 371.3 TB / 515.4 TB (144.2 TB Free) • VGA: NVIDIA Corporation GA106 [GeForce RTX 3060 Lite Hash Rate] @ Intel Corporation Xeon E7 v4/Xeon E5 v4/Xeon E3 v4/Xeon D DMI2 • Uptime: 1d 19h 54m 6s

weust: there is a technique to making the disk shelves quieter. You run them with only one PSU and one IOM6 controller

reduces the noise quite a bit

I do run only one PSU. You're is a HP server?

No, all 3 of those disk shelves are NetApp DS4246 and my computer system running them is a custom built X99 Tower

oh, right

wow your screenshots are vintage, rennj

thanks, old man yells at cloud.

hmm when you doas <command> and pipe to less it messes up and doesn't work properly

any workarounds?

how does it not work?

you could `doas sh` and then run your command

you could run it in a subshell

well, stuff like doas <some cmd> | less has the problem, that stdin for the password prompt gets redirected

or you could configure doas to not prompt for the password

its completely fine to not use a keyfile when using geli right?

I know it adds additional security... but this is a server... it SHOULD never be stolen, the encryption is just to make destruction easier...

rtprio: hm I *could* but rather not

prevents accidentally privilege escalation

project gutenberg gutenberg.org/ebooks/2680 keyfile

hehe

key it secret!

grr keep

yes, completely fine, just depends on your security needs

you could use /etc/services or /etc/hosts only you would know.

but make sure it doesnt change

services could change

use a usb stick with a picture of your sweetie DCIM_14212.jpg

rtprio: ok cool thx

its just a second layer isn't it?

random data stored on another device plus passphrase

iirc it's a composite key. like one really big passphrase

ah

interesting...

well boot .iso.., ram-os/in-memory, i dont want nothing changing

checksum all the files..tripwire

if the software is always changing, how am i going to know if it software fault or hardware fault. static os, i know is good. means its hardware.

last laptop i got 7years of use 2016-2023, power on hours, and it was el crapo hp pavilion 4core/4thread amd, 16GB ram, 512GB sata ssd

the power switch is what failed..

little plastic nob with metal contact on, the metal contact came off

smartctl data power on hours of sata ssd i did like 22TB on that 512GB drive

i have a drive with 97000 hours, wonder if it will hit 100k or die before

hm so geli stores metadata backups in /var/backups/ however the inital geli encryption made in the installation media doesn't... (I think its called bsdinstall?)

the metadata I assume stores the hashed passphrase and info about the encryption (such as alg and bitsize), I assume this is something you should backup as part of your system backup?

is the metadata sensitive?

i dont think so

to which, it being sensitive or needing to back it up

correct me if i'm wrong.. i think the big problem with hardware RAID controllers is that they lie. they say something is done (e.g. flush to disk), but they're really saying "we're taking care of it, don't worry." but ZFS wants to control the disks because it relies on not being lied to

more or less

so I got two HDDs and an SSD, the SSD is the boot disk which has a single disk pool which was geli encrypted within bsdinstall, the two other HDDs will be a zmirror pool but have geli underneath... if I stick these in the geli_devices in rc.conf will zfs automatically assemble the pool after booting into userspace and being prompted to decrypt them?

try it and find out; that's a very specific question

lol guess trail and error time :)