SponiX: i always figured the (RENEWED) drives were smr lol

Macer: I've eBay purchased lots of refurbished, renewed, and straight up USED drives without getting anything SMR. I do always research them to be sure though. And yeah, most of them I got at $10/TB or Under

I may have to try that some day.

I mean I have 3TB drives that have literally like 12 years of run time.

I miss actual hitachi drives. :(

Macer: IBM deskstar's :)

the ones that you dropped once the glass platters would goto shit :D

but I think they are now rebranded hitachi drives ?

i know Lenovo bought IBM's computer biz, and some other company bought there storage

I'm contemplating removing the HDD from the pool

my pool has 3 NVE's 1 SSD and 1 HDD

heh

it may soon be just 1 NVE as I may install freebsd on this mini PC as I want to play games again :(

*nvme

emvn

always thats one person

i have a ryzen system with freebsd on it that's just sitting in cold storage because i don't really need it right now and it only has 12 hotswap bays

For my personal use for larger storage I have been buying used SAS drives from datacenter recyclers. Those have seen what I could call gentle use in a rack and not bouncing around someone's basement. So far I have yet to have any of those used SAS drives fail. And I have redundancy and backups when inevitably one does.

i'd really like to get that thing going but i'd want to put those 16 2.5" hotswap bays in it and just start scaling up SSDs in it

SSDs are not perfect. And in many ways spinning disks can be more reliable and have been more reliable. In total numbers I have seen a lot of SSD failures. But those were all lessor vendors. But SSDs are not without their failures too.

like i'd want to get 5 of those

rwp: yeah i'm sure anything can break but at least seek times would be lower lol

One of the SSD soft points is long term data archive storage. If the data is not active for a long time it will degrade. This is actually a strong point when coupled with zfs because if there is a localized failure due to data decay then zfs will heal it from one of the redundant blocks.

cold storage right?

i guess it has to stay charged up

I have experienced the problem of NAND storage decay on laptops. At least two laptops that were powered off for a couple of years and then would not boot due to non-system disk failure. Did a rescue read-write pass over the NAND flash and after the complete rewrite of all data then it recovered and appeared perfectly functional again.

i've had that happen with usb thumb drives i left in drawers.. they wouldn't function at all though

but i probably wouldn't use SSDs for cold storage.. if that were the case i'd probably consider something like tape

And then in a datacenter setting an Intel SSD with what was probably data completely static for probably several years then turned up with a read failure on a few blocks. Again doing a full read-write pass and then it appeared fully functional again.

i'm running 4 SSDs in my proxmox server raidz ... i'm waiting to see if those ever give me problems

i have all the containers and VMs backed up on platters in 2 separate locations

Mag tape is pretty good for long term storage but it still requires maintenance. I forget the specs now but every three months the spools need to be rotated a quarter turn to help balance out the stretch due to sag while on the spool. And every two years need to be spun off and on again to ensure that the material doesn't become sticky and peel off. Or something similar to that maintenance. It's been a while.

i'm still waiting for the price of 8TB SSDs to drop before i switch

evoprox 14.5T 1.50T 13.0T - - 26% 10% 1.00x ONLINE -

after trimming...

evoprox 14.5T 1.50T 13.0T - - 13% 10% 1.00x ONLINE -

Last I looked the knee of the cost curve for new was between 4TB and 6TB spinning disks.

kind of crazy how trimming reduced fragmentation by 13%

i have a 5th one sitting on a table because the 1u i bought only had 4 hotswap bays

I am a little confused by that fragmentation number too. Because that's the fragmentation zfs sees and trimming or not trimming should not be affecting its fragmentation.

rwp: yeah. i trimmed and it dipped to 13%

Weird!

i was a little confused about that too lol

i mean it's free space fragmentation right?

so i guess trimming just freed blocks so the free space is no longer fragmented?

like regular trim /dev/device? Not zpool/zfs scrub ?

zpool trim

but free space is free space, regardless of whether it's trimmed

Oh, I've never did a zpool trim, I've only had zfs on spinners lol

ivy: not sure... pastebin.com/6sEbEJiS

that's what happened though

SponiX: i typically only run containers

but i guess with VMs you also have to trim within the VM

what ever happened to HAMR drives?

i thought that was going to be the next great breakthrough in higher density disks?

probably just as groundbreaking as smr :/

Macer: i thought they appeared a few months back, aren't Seagate's latest 24TB+ disks HAMR?

oh are they? i didn't see them listed anywhere. but given how well these seagate smr drives worked for me...

the actual real-world increase in capacity seems to be less than the marketing claims would have led you to believe though

HAMR isn't like SMR, the performance characteristics should be the same as CMR, aiui

ah, 30TB: arstechnica.com/gadgets/2025/07/sea…are-now-available-for-anyone-to-buy

the article specifically notes they are CMR, not SMR

superheated cmr though :)

i wonder if i could achieve the same results by letting my drives run at 200C

i'm thinking not, but if you try, please record video

lol

who needs science when you can use alchemy?

this ... Seagate’s press release is focused mostly on the large drives’ suitability for AI-related data storage

🤦🏼‍♂️

hi.

Matt|home, do you have a question? Custom here is to ask directly and wait for someone who can answer.

nope.

I should make a bug report tomorrow about that WiFi driver thing

I'm nervous though

I've never done a mailing list before

I'm triple checking to make sure no one has posted it yet but im still worried

Sounds like you've done your due diligence and can open a report, just make sure it is clear how to reproduce the bug and include as much logs as you can

I need an account for the bugzilla right?

Yep

Yes.

And the way I see it, if someone posted about the same thing before and you couldn't find it, the symptoms or related information are different enough that it's worth to post it anyway.

This was a helpful video that I found. I never actually submitted the report but it might help someone else. youtube.com/watch?v=6fwfRoDt9bQ

anyone run into problems trying to boot the amd64 14.3 disc1 installer on amd epyc servers? i see boot menu, autoboot starts, then when it starts booting installer it says panic: madt_parse_apics: double apic id 241

Hi! Has anyone figured out how to update a (fresh) 15.0-RELEASE pkgbase installation to stable/15? I've modified /usr/local/etc/pkg/repos/FreeBSD.conf and 'pkg repositories' show the correct urls. pkg always reports "Repository FreeBSD-base has a wrong packagesite, need to re-create database".

zapata: show the contents of /etc/pkg/FreeBSD.conf, any files in /usr/local/etc/pkg/repoes, and the full output of 'pkg update'

kerneldove: i'd try updating bios on that server, that tables comes from acpi and error seems to be non-recoverable

ok ty i'll try that

ivy: privatebin OK? privatebin.net/?64733e3f2943dd52#GX…8x3EE8fkYFdKJYFCHjVmDMuyLNBCvoKKHw2

ivy: pkg now shows: No trusted public keys found.

zapata: change the signing keys for FreeBSD-base to /usr/share/keys/pkg, the same as FreeBSD-ports. releases uses a different signing key to stable

ivy: Thanks alot. The error is gone. 'pkg upgrade -f' seems to work now.

ok tried cyric and we're already on latest bios

is hint.acpi.0.disabled="1" bad to use?

V_PauAmma_V: alright thanks

im having problems upgrading my jails, root@:/home/sopparus # freebsd-version -u

14.3-RELEASE-p2

root@:/home/sopparus # freebsd-update -r 15.0-RELEASE upgrade

src component not installed, skipped

freebsd-update: Cannot upgrade from 15.0-RELEASE to itself

root@:/home/sopparus #

-b on host have the same problem

only the host is 15.0 not the jail!

are you running that in the jail? just run it from the host with -j

sopparus: also is your host running 15?

jails are upper bound by the Host.. so if host is on 14.3 jails can be no higher than that version

Amazon failed me yesterday. My 2 hard drives did not arrive and I got the "we're sorry" email.

Hopefully they arrive today.

same thing happened to me last week and when it finally arrived it was just a box of incense..

Macer: are you a prime member? if you are a prime member.. you can request a deduction in costs..

voy4g3r2: yes host is 15.00

.0

isley: on jail

run it on the host with -j jid

yes, what isley said

freebsd-update -j <jailname> -r 15.0-RELEASE upgrade

ah yes and if you have multiple jails, don't run freebsd-update in parallel for them :p

or you'll at least need to define different working directories for each I suppose

cd -

err, oops :D

hmm weird

so I copied /usr/obj over to a jail to update it

but then I get:

cc: error: no such file or directory: 'ERROR-tried-to-rebuild-during-make-install'

the commit built and the commit using make installworld from are the same

polarian: because the timestamps make it think it needs to rebuild things

kevans: the fix?

for local system? I'd probably just null-mount both src and obj trees

admittedly i personally just rely on binary upgrades for jails, while hosts remain source-based (but pkgbase distributed from my local infra)

kevans: nullfs for jails would be annoying tho :/

ohhhhh..

I finally understand why nfs

nfs doesnt modify the mtime

so when you mount it the makefile doesnt get confused

er, why would nullfs for jails be annoying?

kevans: means editing the config :)

im lazy

is there no flag to just "skip-rebuild"

or "install-anyway" :P

--i-am-not-an-idiot-please-install-anyways-and-ignore-mtime :)

--enable=true --disable=c:\windows\twain_32.dll

anwyays I guess nullfs is better anyways, no need to move obj around :P

also I just realised I only need the nullfs mount when I am updating, which is manual anyways so I can just mount it when needed

no need to edit the jail config to have a persistent mount

polarian: nah, such an option would fundamentally break how make(1) operates and people would get angry

kevans: what exactly is make doing though, I know it checks the time of the last build in the metadata, when it checks the mtime and sees it doesn't matches so then rebuilds?

it's an easy comparison between mtimes in the object tree and mtimes of the files it depends on in the src tree to determine if the objects are outdated

also while talking about the makefile, is there ever a reason to cleanworld and then rebuild (apart from development reasons), is using the cached objects and only compiling changes always the better option?

kevans: ohhh right, thats obvious xD

im stupid

some src.conf(5) options break ABI and would necessitate it, but we've been moving towards a world where everyone does incremental by default because it's ssafe

wait so... doesn't that technically mean that not using git then would cause obj rebuild?

wait git will still have different mtimes

so I need to mount the host /usr/src too

yeah, git doesn't track mtimes like that

yea I am aware, it was oversight :P

or rsync, cp -p or something to sync times

well I dont believe there is a guide on "deploying your own src builds to other systems" in the handbook

so cant blame me for being stoopid :P

i typically try not to assume baseline knowledge in channels like this, and i also try not to be an ass when i'm not assuming baseline knowledge to avoid offending (but it doesn't always work)

i mean, i know i've seen you around for quite a while, but i also don't pay much attention these days

lol

just because I have been around doesnt mean I cant be an idiot :P

hmm

install: rename: /lib/INS@muiMNy to /lib/libc.so.7: Operation not permitted

hah :-)

still doesnt work ugh

its root, I see no reason why it cant

ohhh you can do it from the host...

I guess the jail might not have all the permissions needed to modify itself?

can't believe the 14.3 installer panicked on startup on this epyc server. like wtf

lol

kerneldove: why not 15.0?

if you are doing a fresh install surely its best to use the latest release, and not have to update to 15.x by mid-2026

it was the newest iso in the library at the hosting company

otherwise i woulda for sure

i got 15 running everywhere else

I have found the distributeworld task in build(7) so lets see if this lets me update the damn jail

kerneldove: what company (just curious)

polarian: why not just use DESTDIR from the host?

Koston: yeah I have figured it out after reading the man page more

whats distributeworld for then?

ah for releases, although I assume I can distributeworld, then tarball it and then copy this to another server and extract it into /

although then I would also need to clean up all the old files still which the tarball cant do hmmm

So I replaced SSLH with HA-Proxy due to stability reasons but an unexpected side benefit has been that the response times are much faster with ha-proxy. When I did an SSH connection from my ipad to my home via cellular, with sslh in the middle it would take a good 10-11 seconds before the password prompt popped up whereas with ha-proxy this is instant. I had always thought that it was due to cloudflare

handling DNS for me and doing "stuff" in the middle, but turns out it was SSLH. Something to investigate.

just an fyi for folks :-)

voy4g3r2: yes I am. I should see if that’s an option lol.

It was put for delivery at 5PM. And had a latest delivery at 10PM… and never showed up. lol

anybody running 15.0-RELEASE and signal-desktop ? mine crashes on startup :-(

"Render process is gone [...] exit code 134"

i'm so annoyed with signal :P

polarian: when installing world, some files/directories are given special flags with chflags(1). this can fail if your install target doesn't support those special attributes. just FYI - it's not relevant if you do things by the book, but something people regularly run into when they get creative

it's the least bad messenger i can recommend to end users, but the outrageous electron app needing a full gui to just the desktop app is stupid

this is why open protocols are superior to platforms

+1

IIRC there's a TUI for signal now in ports

i use xmpp when i can, but it drives me nuts i can't use irssi with signal

dch: yeah, scli. i'm on 14.3 and it doesnt' work

it didn't work in 13 either

libsignal keeps changing

and it's a hack

Demosthenex: this one freshports.org/net-im/gurk-rs

I didn't try it yet but maybe I need to

hrm. gurk? that's new

yeah

hrm, that's promising

looks like profanity

so what to do

dch: as to signal desktop, no idea what a render process is

hm i lost my byhve usb passthourgh with 15.0, can i go back to 14.3 without reinstall? i cant turn off my lamps atm! :)

Ot only the kernel if possible

You can't have 15.0 userland and a 14.3 kernel, I think. (The reverse is possible.)

:(

I've just packbasify my 14.3p6 install with the lua script, now I would like to upgrade to 15.0 I couldn't find any info on the release notes for the upgrade. Should I follow the steps from the pkgbase wiki? wiki.freebsd.org/action/show/pkgbas…rect=PkgBase#Major_version_upgrades

aww. gruk signal CLI is actually a TUI. :(

Hrm... I managed to botch something up during an upgrade.

gurk*

s2r: Yes, but I would for sure make backups of anything important prior to attempting that

sopparus, Boot Environments? You could boot the previous 14 environment and that would take all of kernel, base, and /usr/local ports back to before the upgrade?

rwp: That's what I'm struggling with at the moment. Even previous boot environments aren't booting.

It's really strange. Still digging.

Huh. All of a sudden the wifi is working

But still spits errors

Thank goodness for live usb!

ek, Previous Boot Environments are not booting? Has anyone reported that the new 15R bootcode won't boot 14? That would a problem for me for certain.

Hello. What's is the best way to do a diskless setup for freebsd? I've set up a nfs in /stranger/diskless_bsd, and trying to boot it with loader.efi but it hangs while saying it is using "net0". I'd like to know if it is possible to do this using iPXE isntead of my dhcp server. But it seems like as loader.efi sends a DHCP packet it will ignore iPXE's set root-path and set dhcp.root-path

saya99: i use ipxe, both on a bhyve vm and hardware

Do you have a diskless setup?

Can you send your configuration files, or a guide?

i followed mostly `man diskless`

let me boot the master and check some things

And the wifi is broken again. This is weird

I have followed diskless(8), indeed, seems like my problem is at the DHCP server/loader.efi

is your autoexec.ipxe loading?

Sure, but I have it configured to load menu.ipxe instead of autoexec.ipxe. environment reasons

the iPXE server works perfectly, I can load linux distros in iSCSIs target perfectly

so i `chain tftp:// /boot/pxeboot || chain tftp:// /boot/loader.efi` in mine

and of course copy /boot into /tftpdir

AH. i see. I'll try doing that, thanks!

bhyve could load loader.efi directly, my intel card it was too big, so that's why i did the pxeboot first

Nope, seems like it isn't working yet. Do I have to make changes to my DHCP server configuration?

my dhcpd.conf: bsd.ac/muvx21w

Okay, I changed the set-root at the beggining for tftp and the loader worked. which is gnostic. Now I have problems with NFS ^^;

yeah, i had `filename "tftp:// /autoexec.ipxe" in dhcpd.conf

you can probably get it from here; i had a couple scripts to jail into /diskless/hostname to do things like install packages or installworld

rwp: Yep. Super strange. If it weren't such a detrimental system, I'd have gathered more info for a possible bug report.

But, choosing a BE during boot in order to "rollback" wouldn't work. Tons of signal 12 messages.

Booting from a live USB and chroot'ing in and activating an old BE worked fine, though (thankfully).

rtprio: I could make the diskless freebsd boot. Thanks for your Enlightment

Now it seems I cannot lock databases... lockd and statd are running in my server

zfs set sharenfs=-alldirs,-maproot=root,-network=192.168.0.0/16 stranger/diskless_bsd

wavefunction: good luck :D

:) me too!

Thinking of upgrading my VPS as well

o/

just checked, after upgrade to 15.0-REL I don't actually have even access to previous boot environments via the bootloader

menu says default is "3 of 5" but pressing 4 or 5 doesn't do anything at all lol

only previous one available to boot via the menu is an intermediate 15.0-REL one which I assume only has 15.0 kernel but 14.3 userland

not a problem as such, this server is not yet in production and 15.0 is working fine so far. just got curious reading the channel backlog and wanted to test

Koston, I assume that is after updating efi/gptzfs

presumably? I'm not very well educated on what freebsd-update(8) does under the hood

all I did was standard freebsd-update and pkg-static upgrade -f business, nothing else

AFAICT it doesn't update the boot partitions

on zfs?

yep

so I installed windows on the desktop and played doom3 for like a coulple hours cause why not. then I added a nvem drive the mni PC, installed xorg, firefox, kde screen would turn pink, messing with it some more screen would turn pink, I google and intel said it could be cause by driver or cable issue

so I decided to change out the cable

no more pink screen

never had a cable go bad but i guess first time for all the things

this box was originally a 14.3-REL install, only updated to latest -p5/-p6 before 15.0. haven't updated bootcode or zfs since install.

i have never updated boot code n any install

I have, but only when either instructed by UPDATING or other issues forcing to

.oO(what is the UPDATING I'm supposed to read when using freebsd-update anyway)

jesus one of then .oO people heh

I'm part of a movement, man

i started using IRC when you had to telnet into it

but now we have modern clients with mnemonics. what a time to be alive

what are thought bubbles

On my NAS at home I've been upgrading this machine since 8.0... so I decided to do a fresh install. Time to clean up some cruft.

well, here we go. 14.3-REL boot environment starts up fine, just needed to boot it up manually via loader

I guess I should file a bug about it tomorrow. bed now ->

upgrading to 15-REL takes a bit of planning here, 14 jails, upgrading all the packages

I've done quite a few, just kind of winging it, it's been fine.

Koston, As I recall (would need to check) the 'B' or the number key is hit repeatedly to cycle through the list of Boot Environments. There is no shortcut number to get to boot environments beyond it. Hitting the BE number repeatedly cycles through the list of them.

freebsd works for amazon now, great

yes, it works basicaly for everyone who accepts its work

I am using FreeBSD and have set an interface to be "DHCP" in my rc.conf, how can I also pass flags to the ifconfig at the same time as using "DHCP"?

My beginner approach of 'ifconfig_igc0="DHCP -lro"' did not work

I don't know but I was curious so I looked into it, and it appears to be handled in /etc/network.subr around L167 and looks like it should work - did you try to place the options first ? (Before DHCP)

Many thanks, I will try

I don't know either and was also curious and have been looking but I think DHCP is a pseudo argument to switch to starting dhclient on the interface. And so I think the solution is to have dhclient do it.

dhclient has dhclient.conf but no facility to add optional flags to ifconfig. But it calls dhclient-script which sources /etc/dhclient-enter-hooks if it exists.

I think I would try calling "ifconfig igc0 -lro" in /etc/dhclient-enter-hooks and seeing if that does what you want.

@o0x1eef Unfortunately this approach did not work, but thanks for looking

@rwp Using the dhclient-script approach worked, many thanks

That's great to hear :)

Sorry for the disconnect/reconnects, was rebooting the router

I guess I could use opnsense/pfsense, but thought it would be more fun to hand build something and use kea + bind.

That was my approach and reasoning, too. I bought a pfSense router, then wiped it with stock OpenBSD.

I have this cheap mini PC with an Intel N150 CPU in it and 4x i226-V, I have played with a few different OS on it, and for some strange reason, Windows Server with routing and NAT enabled, gives the fastest internet speed. Tinkering with FreeBSD on it to see if I can match it. Not sure if its something to do with proprietary drivers.

I noticed on Windows and Linux I was able to set tx and rx ring buffers for the network adapters, is this something that can be done in FreeBSD.

nicks, Yay! Glad that worked. It was a new problem.

nice. new drives showed. and it's amazing that the CMR drive is going 1GiB/s and will be done in hours vs the SMR drive that took 5 days to resilver

what a world we live in

How did you end up with the SMR drive in the first place?

i had old barracudas in a drawer and a drive died .. although i still have some of those in the pool and never really had problems with them

Ah

didn't realize it was smr lol

I've had some Barracudas in the past, but they were made before they came up with the SMR nonsense

it has become quite the hindrance for me

i was just expanding the raidz2 vdev and 2 of them just died on me during reflow. i think i'm going to be at the point where they'll all die off and i'll wind up replacing them all one by one and i think i still have 6 left

17.1T / 69.2T copied at 30.6M/s, 24.71% done, paused for resilver or clear <- it only made it to 25%

i'm hoping that the one in the drawer (which failed right after it resilvered) just had issues and the rest can grind through

that probably explains why it was in the drawer :)