would someone be able to update docs.freebsd.org/en/books/handbook/…book/#building-on-non-freebsd-hosts replacing 'brew install llvm' with 'brew install llvm lld'?

Actually, it seems that when using the default version of llvm, compiling breaks.

It's lasting longer with llvm@19

i remember getting advice that after i run freebsd-update fetch \n freebsd-update install \n reboot, when i log back in i run freebsd-update install again. it never seems to do anything so is it actually needed/best practice?

update/upgrade steps change over time, so always referr to the documentation

mjp that doc seems to state that we should still run freebsd-update install after rebooting

"Once the system has come back online, restart freebsd-update using the following command. Since the state of the process has been saved, freebsd-update will not start from the beginning, but will instead move on to the next phase and remove all old shared libraries and object files. freebsd-update install"

yes

ok i'll keep doing it. fwiw i'd update the message that it outputs. it's deceptive imho as it implies that it was redundant to run it

'no updates need to be applied'. doesn't say anything about having deleted old shared libs and obj files

freebsd-update is going away in freebsd 15.0 anyway (moving to pkgbase), so will need to learn new steps soon enough

ivy ^

oh

will we use the pkg command for both apps and OS updates?

yeah presumably, i have not looked into it much yet - just the piece in the quarterly report: freebsd.org/status/report-2025-04-2…or_pkgbase_in_the_freebsd_installer

pkgngng :)

speaking of which... what happened to the July 2025 report?

hmm

encounting an issue when upgrading from 14-STABLE to 15-STABLE

it tries to install the first tz zoneino file

(but fails)

install: /usr/obj/usr/src/amd64.amd64/share/zoneinfo/builddir/Africa/Abidjan: No such file or directory

looks like it wasn't built for some reason

my servers pass lots of udp packets, and they all have non-stop /var/log/messages saying Limiting icmp unreach response from # to # packets/sec. what exactly is causing that and how do i fix it pls?

fixed it! i set sysctl net.inet.udp.blackhole=1 and now they stopped

kerneldove_: gj man

ty

yknow I have a similar issue, I have messages like Limiting closed port RST response from # to # packets/sec

I have no idea what this means :D

im getting regular OS crashes on 14.3. fatal error mutex xxx own xxx is on list xxx xxx at line 151 in file freebsd-src/lib/libthr/thraed/thr_mutex.c errno 35

it's p5. haven't seen the problem on p3 or earlier 14.3 versions

can one install 15.0 betas from an image or is it more of a game of installing 14.3 and upgrading?

zip, the former. See for instance freebsd.org/news/newsflash/#2025-11-07:1 .

hi. I'm trying to debug an issue with nginx. I have a jail with PHP-FPM, with Wordpress blog. the problem is that all requests are always pretty slow. whenever i click on a link, it takes at least 1 second until it loads (connecting locally). This also happened to me with Zoneminder which also runs in PHP. clicking anywhere would take around 1-2 seconds, and then it will load. both apps work fine on Linux,

everything loads instantly. I thought there may be some issue with reverse DNS lookups but nothing of such is configured. I have no idea where to start

I would indeed check if it hangs waiting for a dns request

Check either direction, it could be your client trying to do a lookup, or the server

yeah but isn't DNS timeout usually well over 1 second?

in this case it's pretty much always 1-2 secs

Check for that first, before you dig deeper

it also seems to be exclusive to PHP scripts. static files load quickly

are you bridging or using veth?

*vnet

bridging

hm. i think the jails do host or vnet right?

i haven't used fbsd for jails + (in my case) nextcloud which is pretty php heavy but i don't remember there being any hangups with it

No, there shouldn't be any hangups. Can you test with a small php script to see if it's the stack itself or something that Wordpress does?

I smell a DNS request too, but not timing out - just the round-trip time.

i tried <?phpinfo();?> and while i'm getting a blank screen because it's a production php.ini, it loads pretty quick

i wonder if it's a mysql problem

guyss guess what

it was a dns problem

# cat > 50-server.cnf

skip-name-resolve

^C

and now it loads in 100ms

:)

it was mysql though

i'm using php-fpm over a tcp socket in localhost

will i get any significant benefits from using a local unix socket?

No

does freebsd have a thing which allows me to see the actual data going in/out of a unix domain socket?

like tcpdump but for unix sockets

I bet we do

truss?

truss is the wrong end of the stick I think

I have a bunch of $THINGs that stuff data into a unix socket

and I need to watch only that socket, not all the other stuff that is happening

sockstat(1) lists domain sockets. . .

all of those options aren't great

Look at the SOCKETS heading under fstat(1).

CrtxReavr: I need to see the data tho

[tj]: yup socat is the best option I think

or I could do this with haproxy

heh

I get that. . . I'm following the breadcrumbs.

kinda sick

I'll use a network socket

then `ngrep -W byline` because its so pretty

and haproxy into the unix socket

you'd think this is something that can be done with dtrace

sure easily

I will leave this problem for tomorrow's dch

tomorrow tj is already pretty committed

me too but I'm going to let dch++ worry about dch-- poor life choices

same

maybe tj++ will get all the packaging off the office floor

tj-- didn't and I'm certainly not going to

is there a wiki or something that explains how to bump zfs 2.2.x to 2.3.x on 14.x?

or would it just be better to just wait on 15.0 and 2.4?

Macer: my advice would be not to fiddle with zfs separately, and just wait for 15.0

for the truly foolhardy there is a filesystems/openzfs-kmod and filesystems/openzfs in ports

nothing wrong with the ports, but IMO its a lot less stress to use bundled zfs

you can ofc install / update to 15.0-BETA5 already for testing if you have a spare box

FWIW I have been using the beta for a while with almost no issues.

me too, in fact I am using it right now :D

=]

dch: yeah i figured as much. i'm a little iffy about messing with a functional zfs 2.2 on fbsd 14

considering you'd have to install from pkg/ports .. that just seems like a problem begging to happen

i do have a spare fbsd server but now that i have sol working with the ipmi i'd just rather stick with my isilon since it's for a nas

i'm just running short on pool space with one so i wanted to use the raidz expansion to add some drives to it

Macer: upgrade to 15 and run command zpool-attach(8)

im getting regular OS crashes on 14.3. fatal error mutex xxx own xxx is on list xxx xxx at line 151 in file freebsd-src/lib/libthr/thraed/thr_mutex.c errno 35

it's p5. haven't seen the problem on p3 or earlier 14.3 versions

V_PauAmma_V: thanks :)

hello everyone, i have been working on getting NFS exports to work on a freebsd server to some linux clients and hitting a huge wall.. i can get the mount to work but can not write to it. I know the uid/guid ARE different between machines, would anyone know a good tutorial that talks about "id" mapping for nfs v4

the NFS exports are setup to use the /etc/exports file format as the sharenfs option on my zfs dataset seems to work "less" not sure why, but i am getting a permission denied issue unless i am root on the share

maybe make the uid/gids the same

yeah.. i saw that and was honestly like.. i could hose my machine.. fun fun

but that is a valid option and i saw this "mapping" utility available on v4.. figure one last ditch effort

changing a uid is not difficult nor complex

voy4g3r2: this is something I heard from another admin, but the usermapping between freebsd and linux is incompatible. there might be some trickery involved to get it working

what I can tell you, v3 works fine between the two assuming uid:gid is the same between the systems. getting v4 to work would be wonderful though. I have been meaning to experiment and get it working, if possible

voy4g3r2, I never use id mapping but instead always use a unified account database so that all ids are the same everywhere. That's the easy way. Otherwise the confusion leads to insanity.

thanks, i will go down this rabbit hole to just get the uid:gid to match the freebsd box, seems to be least painful.

Get it to work for /one/ id as a proof of concept. Then everything else will flow from there.

rwp: i was cheating and using samba and was like why are my shares take 80% of a i3 for just "moving" data.. switch to NFS and it dropped to .4% it was great

plus never ran into this uid/gid delta before and for the 2 machines i want to connect to it.. seems like a NIS approach is a sledgehammer to put in a finishing nail

I also always use the sharenfs property and let zfs handle the exports. Again because everything Just Works then. With /etc/exports don't forget that after any change to it that you must re-export the file to have it take effect. And also that it is a mount time check not a run time check.

yup.. doing a good ole service mountd restart.. umount on client, mount on client, check logs.. over and over again :)

voy4g3r2, Another very important thing is that the time be in sync for all participating systems. Make sure all of the systems are running some version of the NTP protocol daemons such as ntpd or chrony or whatever. Don't "jump" the time with ntpdate.

Let's say my desktop has an NFS mounted $HOME directory, because it does, and the time is off, and then I run a "make" on something. If the time is skewed between systems then one can get into an logical infinite loop where files are never up-to-date.

well darn nab it.. its fixed after a simple usermod -u <#> <user>

oh, good point, let me check that .. but got one machine working

and what did i learn with this: Stop overengineering a problem that does not exist...

rwp: i take it, if a "jump" happens when the connection is alive.. chaos ensues?

voy4g3r2: you got all worked up about not much there eh

rtprio: yes.. yes i did.. for hours.. i was happy i did the switch but NEVER thought about writing a file and then i worked on my embedding model project.. it kept saying, i can not create files with a permission denied

my infrastructure is reading files, converting json to literal and semantic phrases.. all great.. loading the results in postgresql.. logging.. the whole nine yards...

but my d**** forgot i do it all in RAM to remove hard drives from the process.. as it is millions of files

load chunks, process, move on

heads down and kicking butt

thank you for assistance.. it was just a simple usermod command and all good.. err

nfsuserd_enable="YES"

supposedly this is equivalent of linux idmapd

if you go back to testing nfsv4 one day

well, as mentioned, I was told that the usermapping doesnt work properly between freebsd and linux. I hope that isnt completely true.

maybe I should just try it LOL

voy4g3r2, (sorry, phone call) Jumps forward in time are not a problem but jumps back in time really mess with some software that expect monotonically increasing time. One time shots where we set the time are not really a problem. But I have seen people run ntpdate by cron hourly and then over the next month the clock jitter repeatedly tickles the problem of not having a monotonically increasing clock.

I used to do chip design in a VLSI design lab and we had 328 people working there so you can imagine the infrastructure for it was appropriately sized. It was NFS everywhere. With that much activity happening if there was some way for someone to tickle a problem in any way then you can be assured that they would tickle it repeatedly!

rwp: oh yeah that sounds like a "band aid" i got ntpd running on my machines.. so i honestly do not think about it, good nugget for consideration

oh yeah.. people doing people things

i just went through that with a portable environmental particle counter.. we design how it will be used, qualified it for a use.. then i got a user doing somethign totally different.

it is like dude, what are you doing.. why are you doing that.. oh well it sounded interesting.. so i did it

it is just crazy.. and only becomes apparent when you increase the use case from 10 people to over 30 people

they did not like.. there tests for the last 3 days (89 tests) had to be redone and he had to write a lot of documentation explaining his deviation

Right! Things always work in one consistently used set of circumstances. Then many people are turned loose on it and that's when you really find all of the corner cases.

yup

like this whole problem, you just helped me with, i am liking it has to be complex and something real hard.. nope.. simple

KISS for the win!

Pro Tip: If your manager visibly cringes every time you say the phrase "Test in production", try easing them into it by substituting some synonyms. Like "experiment" or "validate". Let's validate the release in prod and experiment on our users." There now! Wasn't so bad that way, right? :-)

hahahah

well i am the "manager" and my team knows better.. i hope

this was a rogue person at a site.. who was not trained.. who deviated from their own processes

but i have no clue how that relates to your ntp thing.. it just popped in my ehad with your rogue person and ntpdate you

this scenario happened on friday

One of my responsibilities was making things as safe as possible for people to make mistakes while doing work that they were not quite knowledgeable enough to do. They must do it. They will make mistakes. They need to be able to make mistakes and learn from them and get better for next time.

yes

That's what I thought of when you mentioned a rogue person doing things.

it was a mess.. he updated the firmware on the device (have no idea how) then went to a different menu (that only exists in that firmware version) and used it in production. it was a lot of patience to impale on the sword for him.. figuratively

it was a fun little project.. i got to dust off my serial communication skills baud rates/parity and capturing communication packages

mzar: if i upgrade would i even have to attach it?

oh i read that as import

it's not super important ... i was just going to add some 8TB drives to the pool to expand some space because it's at like 78% .. it can wait a couple more weeks for RELEASE

there's no point in waiting if you're going to do it

rtprio: right now the point is down time... i took it down yesterday to swap ram. it's a pain to bring it down since it feeds a lot of things

well... swap ram and work on getting the ipmi sol working

hey has anyone here managed to compile dxvk (Vulkan for WINE) in FreeBSD? If so, how do you do it? I tried 'winetricks dxvk' but that fails to compile.

uhm

pertho: `winetricks dxvk' isn't going to compile dxvk. It's going to install the pre-built binaries into your WINEPREFIX.

FreeBSD is the BEST OS for the thinkpad-X20

thinkpad-X220

no OS is good, shred it all

:)