thorongil, It sounds like your array is failing beyond the redundancy available and has therefore degraded to failed. Your OS was running and so continues to run.

It's not like you have many options but upon reboot you might need to be resigned to finding that you have lost the array if you can't get disks back online.

Meanwhile... I have a Seagate here that glitches out every so often, requires a power cycle, then is okay for a day, week, month, then it glitches again. So maybe...

Also just a reminder that if you do get the array back online but zroot/ROOT/default is corrupted remember that Boot Environments at boot are available.

I successfully recovered a problem on one of my very recently by booting back to an older BE at boot time.

Anyone on 13.1 without virtualbox-ose-kmod installed able to pastebin `pkg install -n emulators/virtualbox-ose-kmod` output?

got my usb microphone

pcm3: <USB audio> (rec)

:P

thanks jmnbtslsQE & rwp. i am definitely resigned to losing the current drive contents---fortunately, they are super easy to recreate. (it's a one-drive pool, fwiw)

How do i use my usb microphone ? shows up in /dev/sndstat

hw.snd.default_unit=1 < -- speakers with pcm1

if/when i go through the reinstall process, is there any reason to choose BIOS vs EFI?

Everything made in the last decade or more supports UEFI, and it'll be the best-tested configuration.

i guess i figured if there wasn't a compelling use-case for UEFI over BIOS that many people would still be using the BIOS option and so it would remain pretty well-tested even so

skered, pkg install -n emulators/virtualbox-ose-kmod on my system. Anticlimatic. bsd.to/5t5O/raw

Title: 5t5O

thorongil: Legacy boot is still the most common for VMs.

rwp: Any chance mpfr and/or sccache-overlay installed?

thorongil, The one thing I hint is GPT partition types FTW for sure.

Mm. GPT labels are great. But you can use GPT with either Legacy or UEFI.

skered, pkg info output bsd.to/9RvC/raw

Title: 9RvC

hmm. do GPT labels really do much for me if all my disks are given over to ZFS pools?

rwp: If you run `pkg autoremove -n` does that try to remove anything?

(I swear I'm not trying to be contrarian, just trying to understand)

skered, pkg autoremove -n, yes just two things bsd.to/qVGH/raw

Title: qVGH

rwp: Ok thanks

skered, Perhaps "pkg info -d emulators/virtualbox-ose-kmod" is what you are looking for?

rwp: It should be nothing.

thorongil: Yes, you want your ZFS to live inside partitions, and GPT is a great way to label those.

I think its working, i had to chagne the settings in audacity and its reocrding. wonder why it dosen't work in chromium

mason: I don't want to allocate the entire disk to zfs? or are you suggesting that i create a single GPT partition that comprises the whole disk and point ZFS at it? if so, what does GPT buy me in that case?

in freebsd you want the pool to be inside a partition of time freebsd-zfs, but it can be the only partition if you want

of type

thorongil: A single partition would be better. More controllable if you have to add a new disk into the mix in future, for instance.

thorongil: Example, your single-disk pool becomes a mirror, but your second disk is bigger.

Or let's say you need to migrate to another similarly-sized-but-not-quite-the-same disk.

I had this as a surprise today for instance. I'm migrating my mirror to SSDs, and lo and behold, the 1TB SSDs are 20G bigger than my 1TB spinning rust.

But since I use partitions, this didn't matter.

I guess it's personal preference, but partitions are nicely hygienic.

works dosen't work, hmm

Works in chromium

mason: interesting, thanks! can you say a bit more (or point me to something I can read) about how a partition makes the zfs pool more controllable and/or how using partitions made it easier to transition your mirrored pool to different drives?

Why don't i have a root parition in fstab ?

using zroot i think the defualts from the installer

zfs filesystems aren't mounted via fstab

thorongil: one element is that a mirror needs the providers to have the exact same size and sometimes physical drives have slightly different sizes

i also think that using the entire disk can interfere with freebsd's recognition of the disk, but i don't know the details about that. i do remember having some issues orannoyances when i tried it

i think there are platforms where using the entire disk is recommended instead of a partition, but on freebsd it is the inverse

you will also not be able to use labels or do anything else with gpt if you use an entire disk for zfs instead of a partition

thorongil: Just saw that. wiki.freebsd.org/RootOnZFS

Title: RootOnZFS - FreeBSD Wiki

i'm still confused. if the partition takes up the whole disk, i don't see how that saves me from the requirement that both disks (partitions) be the same size in a mirror.

i do see the need for that requirement, just not how partitions alleviates the pain of it

if the new disk in the mirror is larger, than the original, then the new one will need to be on a partition of specified size (size of the original)

if the original is a specified size in GB, then sometimes there will be a few MB or KB at the end left, which i guess helps in the case that the size of a new disk is slightly different

for me, the real benefit is being able to use labels properly

i thought, in that case, zfs makes available space equal to the minimum of the two drives and that, if the smaller drive is replaced, it magically increase the available size to equal the minimum of the new array

for example it means i can replace an old drive with a new drive that has the same label

well, i was referging to the case where you had a mirror later on

so you have N drives in your mirror (possibly N=1), then add N+1 th drive that is slightly larger...that larger drive will surely need a partition to match the size of the current mirror volumes

in my experience, that is not the case

zfs just doesn't use the space that exceeds the size of the smallest drive

but if the smallest drive dies and gets replaced with a larger drive, the *new* smallest drive now determines the usable space

this seems like an argument to NOT use partitions tbh

thorongil: you are correct, i just tested it. i think the issue would arise when the new mirror disk is smaller than the existing ones, not larger

not during replacement, but when adding a device to the mirror

what do you do in that case, then? can you resize the zpool to accommodate the new, smaller mirror drive? i haven't heard of that feature.

as far as i know, it cannot be resized like that

if the physical disks are specified to be the same size, then surely the difference is very small, possibly in bytes, so if your mirror disks partition leaves a few MB or KB at the end, it is safe to add the new smaller disk

in the case where the new mirror disk is smaller than the existing ones, would zfs even allow you to add it to the pool? i would have assumed not.

right, you cannot add it

so the idea is that your original mirror partitions are slightly smaller than entire disk ; and so the new mirror disk is not too small

ah, okay. so you plan ahead and create 990G partitions on your 1T drives, so that you can later replace with other drives that claim to be 1T but might be a little less.

that makes sense!

i specify their sizes in GB, and often there are several KB left over

but i'm just an ordinary user with a relatively insignificnat infrastructure

most of my drives are physical 3TB drives, and looking at them, i do see that several have a few KB left over, but several do not

for me it really is about being able to abstracting from the physical identity of the drive using labels

gotcha, thanks

this has been an interesting discussion

i haven't heard of the behvavior you mentioned about the mirror zpool size changing to match the smallest drive when the old smallest is replaced, it sounds suspicious but i agree it would be a good way to be flexible and be able to increase the size of your pool

i have heard of that behavior in the context of RAIDZ. i sort of assumed it held for mirrors, as well, but i don't know for sure.

actually, i've tested that behavior on RAIDZ pools myself

i see

that sounds even more suspicious, but i only know enough about zfs to keep my systems running, along with whatever i have learned in times of crisis

i was not able to reproduce that behavior for raidz in a test with memory disks

interesting

however, i'm on 12.3, so maybe it's different on the zfs that's been used since 13

the feature i'm thinking of predates the inclusion of ZFS in FreeBSD

OK

anyone using gnome and have the extenstion app installed? if so how did you do it?

No updates needed to update system to 13.1-RELEASE-p2

-p2 ?

jb1277976: what 'extension app' ?

rtprio: Like dash to dock and stuff for gnome

my workstation just stopped reaching the kde desktop

sddm login is ok then i get the kde gears but then just a black screen with mouse cursor

windows boots ok

fedora recently broke their nvidia 390 drivers so that doesn't boot to kde atm

freebsd 13.1 on an oldish hp z400 xeon workstatiion

doesn't seem to be crashing just either hanging or extremely slow

something had wriiten some crap in ~/.config possible kmail that I started accidentally yesterday

still extremely slow

maybe, create another user and log in with it

such way you can isolate whether something to do plasma/usr configs

I tried renaming .config to .config.bak

ok, also possible.

sddm is such a finicky thing

sddm seems OK it looks like Xorg or plasma has become glacially slow

after getting back to desktop 'baloo_file' was killing perfromance

.oO( Of course! )

OK restored my .config and turned off baloo seem to be back to normal

The issue as I understand is that baloo is used to index files; it uses kqueue|kevent, on FreeBSD, for notification of changes; which results in all the files of interest being kept open; that causes slow down. See bugs.freebsd.org/bugzilla/show_bug.cgi?id=230726 & bugs.freebsd.org/bugzilla/show_bug.cgi?id=256269

Title: 230726 – sysutils/kf5-baloo: OS performance suffers – sometimes, as if frozen – when reaching the virtual memory vnode limit

One could create a baloo configuration file to index only certain files (how or what format? I do not know). See twitter.com/probonopd/status/1583724940867825666 & helloSystem/hello #33

Title: probono on Twitter: "Is anyone using @kdecommunity's baloo full text indexing on #FreeBSD? Does this need to be configured in special ways in order not to open too many files? t.co/fPW8dTaPSC" / Twitter

I never use it so I'll leave it off

Works

Guys it's mixer1 mixer2 dsp1 dsp2 all represent pcm1 pcm2 ?

jb1277976: I don't understand your question

You can get a list of audio devices with "cat /dev/sndstat" and get more info out of it after "hw.snd.verbose=1"

You can also use 2 for verbosity, but it might be too much

meka: Sorry. Does mixer0 and mixer1 represent pcm0 and pcm1 ?

they usually do, but I didn't read the code to claim it can not be different

Hi everyone, for some reason I've been getting this message while trying to install a package to a jail bsd.to/RjAt

Title: dpaste/RjAt (Plain Text)

If I understan correctly, my kernel is behind the package?

monkey_: what does `freebsd-version -ku` say?

probably 13.0 right?

On the host it prints 13.1-RELEASE-p2 twice

I was thinking that maybe I had 13.0, but that's not the case, and the jails is a fresh install

13.0 install?

the cvs-jail, that is

Originally it was a 13.0 install, but I upgrade my system to 13.1 a while ago

hmmm you sure the jail's upgraded too?

Right know I'm working on a new install

I just did a bsdinstall jail /path/to/jail

have a look at freebsd-upgrade'ing the jail then

something somewhere is b0rked ;)

Looks like there is a 13.1-RELEASE-p2 how do i see what is in the -p2? i googled it but nothing really came up. is it on the freebsd main website?

have a look at the security advisories/errata notices

xmj: I remember that I did the upgrade for the jail before, anyway it's working now

Thanks, appreciate your help

first upgrade attempt and i got an error termbin.com/ly8m don't know what it means

i just did a git -C /usr/ports/ pull so i think i have the latest

do i install those packages then try again?

probably a dumb question, but you are all called bsd, but you do have different kernels, right? you just kept the name.

monkey_, I know you said it is solved but for the future I think there was a missing -r in freebsd-version -kru which also then lists the running kernel too. For those who have not rebooted since the kernel upgrade.

In a perfect world all three of those emit the same version strings. If not, then it's a clue.

And freebsd-version is -j jail aware too.

rwp: the three lines have the same version

However I did reboot the computer

Sometimes, a -p<n> only affects userland and not the kernel. Then, you can have different values. For instance, 12.3 is currently -p6/-p6/-p7.

V_PauAmma_V: so I don't need to do anything or worry right?

Look like amd64 is tier 1

lavaball: yes every *BSD is a different kernel. Many UNIX systems called their kernels unix too despite being often wildly different

jb1277976, based on your pastebin, I don't think you do.

vortexx, thanks for explaining.

lavaball: you're welcome

Hi guys, I need a little help... I'm using csh shell. The command history is saved at ~/.history. I want to prevent a user from deleting or modifying this file. I tried "chflags uappend", but FreeBSD cannot append history into this file, and ends up creating another .history_whatever file.

Just pre-ordered a VisionFive2. Hopefully freeBSD will run on it :)

While money can't buy happiness, it certainly

lets you choose your own form of misery.

fortune++

devnull, Do you really need this feature? It is easy to defeat. But if so then maybe (I have not tried it) you could set up .history as a pipe with a reader that records the information elsewhere. Seems like trouble to me though.

does FreeBSD have, by default, the kind of stack protection that caused OpenSSL to downgrade the severity of the current bugs? openssl.org/blog/blog/2022/11/01/email-address-overflows

Title: CVE-2022-3786 and CVE-2022-3602: X.509 Email Address Buffer Overflows - OpenSSL Blog

hey guys, maybe someone can assist me. I can't seem to get a jail to ping anything apart from it's host. vnet config

seems that I can ping the host from the jail, and ping the jail from the host, but cannot ping the router from the jail

raw sockets enabled, etc etc

Hmm i'm getting an error when running virtualbox Unable to allocate and lock memory. The virtual machine will be paused. Please close applications to free up memory or close the VM. I ran dpaste.org/V6D8b im allocating 4096GB what could be wrong? i ran a vm with 4 gigs on current and it worked fine. currently on 13.1-RELEASE

devnull: csh probably writes a new file and then renames it to .history to avoid partially written files and if it does only keep a limited number of history avoid accidentally removing it as it has to override the whole file when removing stuff at the start

codersmoke: you need to make the bridge interface of the host the network interface of the jail for it to work without nat via firewall stuff

nimaje, It makes sense. This explains why "uappend" didn't work and the system just created the . history_whatever as it failed to rename.

interesting

nimaje thats it. Using bash shell, it worked. uappend worked in .bash_history.

hi all

@vortexx, can you explain that like i'm a 5 year old?

morning

.oO( Hmm EN-22:2[16] found their way to my email account but have not seen yet EN-22:24.zfs for 13.1; all: freebsd.org/security/notices )

Title: FreeBSD Errata Notices | The FreeBSD Project

I have just upgraded my box to 13.1-RELEASE-p3, now I have no hardware acceleration on my Intel GPU!

hm, could EN-22:27 have been my problem with trying to update from 12.x -> 13.1? till the loader all was fine but after that it was just a blackscreen, I didn't have any idea how to debug that, so I ignored it after installing 13.1 was working fine (I only had 21 and 22 in my mail inbox, but in my feed reader I have all 7)

Opening xterm glides really slow over the monitor. It's horrible.

iio7: what was the prior version?

iio7: /var/run/dmesg.boot output as pastebin (bsd.to)

koobs, prior version was 13.1-RELEASE-p2

do you use drm-kmod for intel driver?

nimaje, That is what I am fearing if I would update from 13.0 (or, I would rather not reinstall from scratch)

Oh there is a p3 now?

Yes

Only checked last night

Thanks! Installing now

koobs, yes.

iio7: do you use ports or packages? if packages, quarterly or latest?

iio7: that /var/run/dmesg.boot and /var/log/message output (for the last boot) would be great (in the event the kernel mod is not loading correctly)

koobs, thanks for your time, one sec, I'll need to reboot.

And my desktop is running with p3 thanks for the headsup!

Ah and they fixed the dependency of KDE to samba 413 also

koobs, somehow kld_list="i915kms" was changed to kld_list="i914kms" in rc.conf, I have no idea how that happened.

iio7: changing it back fixes it?

koobs, yup.

win

On a completely different note, FreeBSD has so many tunables and settings available in both rc.conf, loader.conf and sysctl.conf, is that unique to FreeBSD? I don't see that many options on the various Linux distros config files.

There are GRUB options while loading kernel to start a system. System MIBs do exist on Linux too

MIBs?

Having not tried to set an option does not mean lack of it (or the complexity itsefl)

s/itsefl/itself/

MIB: Management Information Base

iio7: linux has plenty too

iio7: rc.conf is 'rc' (post init) time, think about it as userland startup

parv: boot environments for the win, I could just rollback when that happend, ignore the update for a while longer and then decided to do a reinstall (with copying all systemconfigs to the other zpool in the machine and copying them back after the reinstall, I just forgot that I had some symlinks setup for poudriere, so that costed me some time)

iio7: sysctl.conf is run during rc

loader.conf has sysctl tunables that have to be set much earlier

koobs, basically, most of that is configured using systemctl for a distro with systemd.

Eh? GNU/Linux still uses sysctl.conf.

iio7: *nods*, same here, vast majority of tunables here are userland (runtime) tunable

Ah, reading scrollback. iio7: loader.conf is roughly equivalent to GRUB and the initramfs on GNU/Linux. sysctl.conf is analogous to sysctl.conf. rc.conf is a centralized way to specify services to be run, which varies wildly by Linux distribution.

Thanks guys for clarifying :)

pleasure

I have just logged into one of my Linux boxes, oh yes, sysctl.conf has a ton of kernel parameters, I didn't know that.