does top show anything busy?

no not much

nothign about that drive

zpool status -v

what exact command did you do?

fuser /bd

before tthat, zpool export bd

do a control-T in the session that you did the export in

interesting

load: 2.18 cmd: zpool 97210 [tx->tx_sync_done_cv] 289.08r 0.00u 0.05s 0% 7456k

mi_switch+0xc2 _cv_wait+0x113 txg_wait_synced_impl+0xb9 txg_wait_synced+0xb zil_close+0xdf zfsvfs_teardown+0xb4 zfs_umount+0x119 dounmount+0x482 kern_unmount+0x332 amd64_syscall+0x10c fast_syscall_common+0xf8

is the disk still running?

yes

constantly

was there a lot of recent write activity on it?

yes

what type of disk?

well i deleted few GBs

hdd

it's apparently waiting to finish synchronizing all write operations to disk, but I have no idea why that should take so long

bad /slow disk maybe

but its new

do top -SH to see system threads - is the geli thread for the volume busy?

no

what about g_down or g_up threads?

none

zfskern* or kernel{z_* threads?

none

Zpool status is it scrubbing ?

gstat -p shows activity on the device?

by the way I found super fast loging of xorg logs saying [802036.674] (EE) Wacom One by Wacom M Pen eraser: Error reading wacom device : Invalid argument

wacom is unplugged right now

RhodiumToad heavy activity with da0 in gstate

100 busy

busy with read or write ops or both?

r/s , ops/s  ms/r is up

did you try zpool status on the pool?

normal

no scrub in progress?

no

last on 27th mar

Are you using that caching crap on kde ?

Kde uses a caching thing that requires MySQL and it's io intensive

Think it's supposed to mimic index service from windows

I do not have kde

there should be nothing in userland accessing the filesystem at this point as far as i can see

by the way, I have this issue and xorg is goign crazy forums.freebsd.org/threads/fighting…com-intuos-s-ctl-4100-working.77785

Title: Fighting to get Wacom Intuos S (CTL-4100) working | The FreeBSD Forums

Think there's some magic in ports that allows you see what zfs is doing

Oh that would do it as well

not sure how to stop xorg logging millions of lines to disk in /var/log/Xorg..

cpet but that drive is not OS drive.

Really doubt it's millions but ok

so writes won't be there

cpet willl millions in a day.

It still scans all the things

few dozen a second

how do I stop xorrg loging

I set pretty much everything to not log and turn it on when something bitches

Beladona: what's the full output from zpool status?

Actually what does ps auxwww let's see what's running and hogging

all normal

how do i stop xorg log

Probably get better results from one of the other tops such as htop

You read the man pages and restart

ya ok but how do I stop xorg log?

it's really not a good idea to get sidetracked into another problem while trying to solve something

damn HHW :/

paste the actual output from zpool status on termbin

I honestly think it's that xorg logging but I'm sure he can figure out how to stop logging by reading the man pages

RhodiumToadhttps://pastebin.mozilla.org/4DWaPC0K/raw

Setup makes no sense but ok

Paste in what ps auxwww outputs

also paste the second block of output from top -zSH -b -s10 -d2 40 (this will take 10 seconds to run)

cannot find man.freebsd.org/cgi/man.cgi?query=Xorg&sektion=1&format=html

Title: Xorg(1)

or the whole output if that's more convenient

found it. its gelli

found what?

i wonder who runs bsd.to? i saw a few people have noticed already the ssl certificate is expired

Oh nos you can still use it

Not sure why you need all your pools encrypted but ok

Topic should be changed to reflect though cloudfare shows a cached version.

might be koobs? not sure

can you still use it? it seems cloudflare won't let you

no

One reason to encrypt your pools is so you can simply throw a failed drive away without having to try to wipe it or physically destroy it.

I have about 60 drives I need to destroy... They are failed drive with data on them. Don't have to do that anymore.

yeah, definitely great for a failed drive scenario

since i seem to keep mine until they do fail

i think the most obvious thing is theft, though

For a laptop... But even for a server I suppose

RhodiumToad system crashed.

I am back and good. drive umounted anyway due to crash

xorg error was due to webcamd when I plugout wacom tablet. for now won't do that. will figure later

Hrm by PC is fine, oh it crashed but it's still fine

Erhard: if it's not enterprise you really don't need to do that I would sell them for 5 bucks a pops rather than destroy them just format then to CIA specs and they are good to go if you don't have data restore clean rooms really doubt you can find anything useful on drives formated that way

cpet: The point was about drives that actually fail... You can no longer wipe them with a computer. You have to destroy them, but, as you say, the likelihood of someone recovering them in that case is slim.

Normally I will wipe old drives and send to electronics recycling/reuse.

i sell them after I format them

a drive that has failed very few people can actually recover data from them without doing some magic, you cant really buy new HD mother boards as a end user so im not really worried about that but i know some people are, I would destroy them if they held trade secrets but even that can be fixed by formating as most drives have a expiration date so they are normally good drives that large datacenters just replace as they expire.

I the past I have run an open HDD and swapped the internal board with another good one, ad recovered the data (because I needed it). POint is that a clean room is not needed, but this was an old drive, and modern ones are likely far more sensitive, and harder to recover.

is this home end user or biz/corp end user ?

What? My use? Both. That data on the drive was personal stuff years ago. Nowadays I have auto-snapshotted systems and auto snapshotted to secondary and remote servers. So data loss is more difficult (though never impossible).

hi, could you give me a tip how to modify www/apache24 in ports so that it produces debug symbols for gdb? and, once it does, is there a place where debug symbols are split off to? I know there's /usr/lib/debug but that's for userland. Is there no equivalent place in /usr/local? Does it mean everything built with debug symbol has them embedded?

build the port with WITH_DEBUG=1

(ports are supposed to respect that and turn on debug symbols)

the port does not expose that option, but I did find a branch in the makefile that handles it. However, it also sets -O0 which is not good, so I just patched in -g -ggdb3. But I'm just noticing that release mode also does INSTALL_PROG_FLAGS="-s" ... which I now assume is 'strip'

it's not an option for individual ports, it's a framework option

make WITH_DEBUG=1 ...

or put WITH_DEBUG_PORTS=www/apache24 in /etc/make.conf

yea just saying that the makefile reacts to that by setting configure flags -O0 (no optimizations) plus -g. I guessed that I could get the same result by putting it into the 'else' branch for WITH_DEBUG. And I almost got it right, but indeed, '/usr/bin/install -s' strips debug symbols. Is there some other magic flag that splits them off instead? I knwo some stuff does it but I never paid attention how

exactly (probably a gdb commandline?)

*gcc

if the port is stripping binaries even when WITH_DEBUG was set, that's a bug in the port, and you need to report that

but I'm not convinced you're doing it correctly

it's my mistake, I didn't realize what that line in the release path meant, and I left it in.

you shouldn't be touching the makefile

just put everything back as it was, and build it with WITH_DEBUG

I don't wanna, this webserver runs live and I don't want an -O0 executable running long-term like that

all it needs is commenting out that install -s flag and it should be good to go

ok, then also set DEBUG_FLAGS="-g -O2"

you do not need to touch the makefile.

ah you are right, I could go that route as well. The debug build also does 'enable maintainer mode', and I didn't know what that did so I steered clear of it.

for now I'm going with 'release, but with debug symbols added in and not stripped out'

I use mod_php(threaded) + worker_mpm(threaded) and it's been crashing apache once a month since 2017, so I figured I'd try submitting a crash report again and see if I can make php pretend to be thread-safe a bit better

I'm not sure mod_php is still recommended, better use php_fpm with some mod_proxy

hi all

I did see that but that sounds way complicated for a single hobby pc. On a 'real' webserver I decided to avoid this issue and it runs singlethreaded mod_php and mpm_prefork. And I'm unhappy about that. And aggressive crawlers caused it to run out of system memory a few times so I had to block them. Gimme those threads

the native mod_php directives are appealing, but the lack of proper multi-threading support with mod_php extension is still a problem, yes

FWIW, setting up php-fpm with proxy_fcgi is very simple, a few directives are needed.

anyone know which package i get the c++ std lib for arm from?

specifically looking for type_traits.h. it's neither in arm-none-eabi-gcc nor in arm-none-eabi-newlib.

/usr/src/contrib/llvm-project/llvm/include/llvm/Support/type_traits.h

/usr/local/lib/gcc12/include/c++/ext/type_traits.h

they are usually included with the compiler you are installing

so arm-none-eabi-gcc is just… broken?

don't know

well, it definitely doesn't contain that file (checked with `pkg query %Fp arm-none-eabi-gcc | grep type_traits`)

grepped through the portstree, there seems to be an gcc-arm-embedded port which contains it…

yup, that one works. :)

grahamperrin: a solution to the 13.1 to 13.2 update? Yes, it was extra zfs labels left over when when one of the drives was used whole-drive ZFS vs partitioned zfs.

grahamperrin: gist.github.com/dlangille/af34e873727c62689ec937530b9ce398

Title: gist:af34e873727c62689ec937530b9ce398 · GitHub

dvl: it's annoying when you add a new disk and you realize you didn't remove the zfs labels previously

aquamo4k: In my case, the drives were not originally mine. :)

dvl: how did creep in out of nowhere, did you just recently add the problematic drive

aquamo4k: No, these drives have been in use since 2016 I think: dan.langille.org/2016/12/18/x8dtu

Title: x8dtu – Dan Langille's Other Diary

huh, i would have thunk that bad label would have caused problems prior

I'll read your blog to try to figure it out

aquamo4k: Nothing in the blog yet about this. I suspect it's just that the boot loader changed between 13.1 and 13.2

i recently put my alpine linux workstation on zfs and appreciate "wipefs -a $DISK" and "sgdisk --zap-all $DISK" commands.

as opposed to using the dd method

I suspect there are anagolous cmds in freeBSD (maybe gdisk) - need to maybe look into that

just updated my freeBSD server to 14-current with the just commited openSSL changes, so far no issues.

can't say ive ever had a need to go beyond stable on anything other than a testing vm for upcoming upgrades.

you do you ! congrats

The testing, especially for big changes like OpenSSL 3, is really helpful.

If _only_ developers ran -CURRENT, a lot less bugs would be found in time for -RELEASE.

i currently use this server as a build machine mainly and I figured since pfSense runs current, I will too, especially since we are getting close to 14's release (i hope) :-)

Likely "pfSense" have people to solve|workaround issues if they would pop up. Do you?

i can just wipe and re-install; everything on the build server is on a git server so it's no worries

Is there a server for the git server ?

Is there a way to add VLAN tag to traffic coming from a specific MAC address?

most of the code runs on another system running gitlab in a podman container so yes.

Is doing a live-install with only IPv6 possible? I get a Fetch error with "Address family not supported"

plus I just like building things from source when I can

your mileage may vary but I'm old and pre-date pkg managers

aquamo4k, I do similar. My contention was more about citing the example with supposedly more resources than you|I might have

parv: ack, understand.

ah, I got it. I was only once again to dumb for v6

jemius: cool, i was just looking through the bsdinstall and fetch source and it looks like they both know how to use ipv6 on a quick perusal.

However, it seems there's a little bug with the installer navigation? When I leave the address-window with TAB I'm stuck between OK and CANCEL

try arrow keys left<>right?

aquamo4k, nope, stuck. That's on a V-Server remote console, though. Maybe it's eating key codes

ack could be. i am running mine on xcp-ng and didn't have any form nav issues. good lulck

In the installer one of [Tab], [Up], [Down], [Left], [Right] eventually goes to the intended choice (who reads the doc for the installer anyway?)

.oO( perhaps I should )

parv, the doc is not really useful here, or is it? Every human instinctively knows that it's the arrow keys and TAB

:-]

exciting freebsd.org/releases/14.0R/schedule

Title: FreeBSD 14.0 Release Process | The FreeBSD Project

why do I get a "Protocol not supported" when creating a SCTP socket? i thought that's a well supported protocol

jemius: what's your code look like?

meena, listen_sock = socket(AF_INET6, SOCK_STREAM, IPPROTO_SCTP);

that works on my linux desktop just fine

I guess the kernel module is missing in FreeBSD or the protocol was not activated when compiling

man.freebsd.org/src.conf(5) i don't see an option to disable it

Title: src.conf(5)

the manpage lists so called `options`; SCTP and SCTP_SUPPORt

man.freebsd.org/sctp(4) but you still can, in your KERNCONF

Title: sctp(4)

jemius: did you include the right files?

meena, what files? that's just a C-Program and it compiles witht errors and warnings. The error comes from the kernel.

this one: #include <netinet/sctp.h>

of course. If I hadn't – it would not compile.

already... wrote

bah

but yeah, if the kernel was compiled without, that makes sense

is there an easy way to show the pkg versions of latest vs. quarterly? either web or using the pkg tool?

markmcb: freshports.org

meena: perfect, thanks!

this Aldi

this also shows if and why something is failing (fallout)

☢

if i want to "export" a jail for backup, is it as simple as "tar czf archive.tar.gz /my/jail" (along with the config)?

i see a lot of references to iocage, but i'm trying to avoid extra tools if possible

markmcb: how did you create the jail? what Filesystem is it on?

zfs, created it manually

on its own dataset

then bsdinstall jail /my/jail

meena, the solution was to load the module. `kldload sctp`. Apparently it is not auto-loaded

good to know