I can run multiple `nc -l :: <n>` or `nc -l 0.0.0.0 <n>` or `nc -l 127.0.0.1 <n>` and I don't get port in use error on the second instance.

Is that expected with FreeBSD or is there something unique about my setup that would allow it.

Like have multiple IPv6 or IPV4 addresses

,/34

.

anyone here running FreeBSD ZFS on AWS ECS/EBS? I used the official AMI for 14.2 for root ZFS and ashift is set to '0' by default. Not sure if this is intended? Should ashift be set to '12' for AWS EC2 with EBS disks?

Do we have an example in base Kyua where we're testing a command line utility?

s/ECS/EC2/

pertho: does it say "One or more devices are configured to use a non-native block size." or nah? if it doesn't it's probably fine

rtprio: would that be in dmesg?

antranigv: are you asking how to run kyua or if there is documentation?

pertho: in `pool status`

antranigv: tests(7) is the place to look, it includes this example:

$ kyua test -k /usr/tests/Kyuafile bin/cp usr.bin/cut

rtprio: zpool status shows no errors

[tj] both. I have a jail manager that needs better testing, I want to do something like `make test` and I'd like it to use Kyua to run automated tests (create a jail, do the init commands, etc) and then I'd like to automate that process in our CI/CD.

I'm just wondering if I need ashift=12 for EBS disks or else there could be some performance penalty?

antranigv: I am not aware of an example of kyua being as easy to use as `make test`

since they're virtual i would imagine that you don't need it

[tj] doesn't have to be, I just need examples of testing a command line utility. (expected outcome, expected output, etc). are you saying that `cut` and `cp` have Kyua tests?

the man page is saying that

the kyua tests ship in /usr/tests

some kyua tests are along side the tool in the source tree and some are in /usr/src/tests/

rtprio: ah cool.. good to know

[tj] thank you! time to start researching!

I'm happy to help if you have more questions - this isn't straight forward to use

for individual tests you can use kyua debug

but not for sets of tests

i am trying to push a git repository via ssh (from windows) to a remote (freebsd server) and i'm getting a error: unable to write file ./objects/18/82886fd3af212fa4b54325fd6bd6ca6fb38584: Permission denied

I can ssh to the server and create a file in the ./objects folder. any ideas as to what could be going on?

that looks like git

what is the ownership and permissions on whatever/objects/18 and the file (if it exists)

owned by another user who is working in the repository as well.

drwxr-sr-x halemu git halemu is not my username

then unless you're in the same group and it's chmod 664 it's not going to work

we're both in the git group

but christ, it's 2024 you could spin up a gitea instance faster than it'd take you to figure this out

*2025 hah

idk what that is

do you know what umask is?

yeah, but knowing what files should have and whether i should chmod -R the whole directory or why i should have to at all is still a mystery for some of this stuff

so each user will need to set their umask correctly so that files they create will hav e the right permissions

and each file will need 664 and directory 775

jnewt: An error writing to ./ means the current directory. What are the permissions on the base directory (.)? The "git" group may have read access but not write access.

find yourgitdirectory -type f -exec chmod 664 {} \;find yourgitdirectory -type d -exec chmod 775 {} \; and make sure everyone has umask 002 in their shell rc file

If you do a "chmod -R 0664 ." in the base directory, that should fix your issue for anyone in the "git" group. However, I do see you have a set-bit setup for the group in your permissions as well. I'm not sure if that's needed or not.

ek: don't do that

Ah. There's directories.

I thought it was just files.

I suppose you could just keep it really simple and do: chmod g+w .

And just be done from the base directory.

yes, you could use that syntax i suppose

jnewt: gitea is a selfhosted git platform. can run sqlite. self contained.

Wouldn't remove the setgid bit (again, not sure if it's even needed.)

it is, he has two developers in the `git` group

Right, but are the executing the code together or something? It certainly isn't needed just to share write access.

what isn't needed?

The setgid bit on the base dir.

right, setgid shoudln't be needed if everyone sets their umask before shitting across the repo

I thought gitea was dead?

This is an impressive project: github.com/izabera/pseudo3d

Oh, no, that was gitosis or something.

With zero special terminal config, it ran in putty on a remote FreeBSD server, with quite reasonable performance, I'd say.

CrtxReavr: Interesting, for sure.

ek: no, it is not dead. some folks didn't like the vibe and make forjo

tsoome

oops

hey, another dummy question. Can I change a zpool (single vdev using 2 disks stripe - raid0 like) towards a zpool single vdev with 2 disks mirror?

or is it that the only way is to do it destroying and starting from scratch?

cybercrypto: since that's smaller than your original config, you would have to re-create

rtprio: That's what I suspect from docs. I am talking about the same 2 disks becoming a mirror. Guess I will need a backup and planned work.

rtprio: thanks.

here's a stupid question, does anyone know if I can convert /var/db/freebsd-update into a cache directory for reverse proxy? I don't want to wait an hour again.

antranigv: i don't think that would work, but you could probably pre-populate the /var/db/freebsd-update/files directory?

cybercrypto: yes, i have 20tb of files to shuffle around, to further complicate things, insufficient sata ports to run both pools at the same time

rtprio probably, yes. I do have a /var/db/freebsd-update/files directory already populated, but seems to be large, and my router is tiny, so lets see how it goes. maybe for the next run in a bit, after I upgrade from 13.2 to 14.0

I'll check again just to be sure.

cybercrypto, afaik this is possible if you have space

Just tested in VM

For example you have pool with name mypool with 2 disks stripe da0 da1.

zpool remove mypool da1

zpool attach mypool da0 da1

rtprio I just realized I can do NFS mount of that directory. never mind :))

VVD: really

then you can zpool create newpool da1; copy your stuff; and then add the mirror

antranigv: can't you mount it over NFS if it's your LAN?

Remilia one message above ^ :D that's what I realized as well

oh

I glazed over that, sorry

rtprio, u don't need create new pool

after "zpool remove mypool da1" u get pool mypool with 1 disk da0

after "zpool attach mypool da0 da1" u get pool mypool with 2 disks mirror - da0 and da1

"zpool remove mypool da1" can work very long (depends free space on pool)

I swear to god I configure NFS for living, how is showmount not showing anything?

firewall ports should be 111, 2049, on TCP and UDP. did I miss something?

Can I do `bhyve -s 3:0,virtio-blk,/path/to/rootfs` where /path/to/rootfs is a directory?

FragmentedCurve you probably need p9fs for that :)

FragmentedCurve: probably not; why are you asking ?

I'm trying to mount an install of src as a root partition in bhyve

But I can't get bhyve to treat it as ufs

Unless I put the directory into an image with makefs

ok, follow antranigv's guidance to use p9fs, but you can use NFS too

Ok, I'll look at that.

I believe there is a typo in the portsnap(8) man page. At the bottom, a directory is shown as "/var/db//portsnap", when I think it should be "/var/db/portsnap"...

also weird that when one installs portsnap, /var/db/portsnap isn't automatically created.

these are only mild flaws, is portsnap working for you dfdx ?

maybe you want to switch to git-lite or gitup ?

mzar: i would not consider documentation errors mild flaws. in any case, portsnap is working for me, thanks.

cool, but portsnap is deprecated, please take a look at git

I read that portsnap was deprecated, i just couldn't find out what replaced it. If it's git, great, I can use that going forward.

Thanks.

I am installing www/nextcloud from ports and running through the `make config-recursive` step now. I noticed a config for openssl-3.0.15_1,1, which already seems to be installed. Is that normal? I definitely have the /usr/bin/openssl executable.

it's also asking me to configure bash, which I also already installed via packages earlier today.

should I uninstall openssl and bash before compiling these via ports?

i'm being asked to config all sorts of bizarre deps, like groff...

i mean i love groff, don't get me wrong. i'm not even asking *why* it's a dependency. I'm just curious why i'm being asked to configure it when it's already installed.

hm... nextcloud is only webapp, so you need http server, php engine and database, you don't have to install it from ports

mzar: yes, agreed. that's why I find this so weird. there is a nextcloud package, too, but it uses mysql instead of postgresql, which I prefer. so I made the switch in ports with `cd /usr/ports/www/nextcloud && make config`. but now running `make config-recursive` is making me configure all sorts of stuff, like just now i configured curl (!!!)

if you install it from ports or as a pkg it will probably pull apache and mysql, but if you want to run it on nginx and postgress you have to intstall nextcloud by nahd

hadn

hand

yes, precisely. I have always used postgresql and nginx on my other systems for nextcloud, so I was hoping to replcate that here (besides, my nextcloud database backup is in postgresql).

this is my first time using FreeBSD ports, so i'm not sure if i'm doing anything wrong.

it will only deteriorate your user experience

seems like everywhere I look everyone is discouraging the use of ports... why even include it as part of the OS at this point.

dfdx: ports are basically build recipes for third-party software on FreeBSD. you are meant to build them using something like poudriere rather than use manually. using them manually is largely an accident due to their history rather than the recommended way to interact with them

it's not like in 1990s, for 3+ FreeBSD systems you usually want to configure own poudriere and build ports in one place

most people that build their own port tree will use poudriere to build a package repository and then use that package repository via pkg

dstolfa: I'm coming from Gentoo so i'm very familiar and comfortable with something akin to ports.

IIRC Gentoo was FreeBSD inspired

yes, gentoo's portage is excellent, in my opinion.

dfdx: you can imagine poudriere + pkg as freebsd's "emerge" split into two parts: build the repository and install on a box (not an exact analogy, but it'll serve for the purpose of this point). cloning ports and going `make install` would be the equivalent of bypassing emerge and manually running the build steps

dfdx: so we have own builders running poudriere or poudriere-devel for such deployments

dstolfa: ahh... which of course, even on gentoo, no one really does (unless you're a developer, say).

so the lesson here is: read up and learn about poudriere, and then figure out how to install nextcloud with its dependencies via pouodriere.

do I have that right?

for single FreeBSD running nextcould I would install webserwer php modules and database using pkg and official pkg FreeBSD repository

poudriere is used for the build step, pkg is for the install step. official freebsd packages are distributed in that exact way, but building your own via poudriere allows you to target different architectures which may not be supported or change options to your liking sort of like gentoo does, although USE flags are more generic than what you'd see in freebsd ports

mzar: nextcloud is but the first of many things I want to install with my own custom modifications (e.g., mastodon being another one, which is "just" a huge ruby on rails program).

I keep getting "mountd[49337]: Warning: exporting /var/db/freebsd-update/files exports entire / file system". thoughts?

dstolfa: that was not what i thought you were going to say.

who cares right now about targeting different architectures? I'm on this machine and this machine only. how should I install nextcloud?

dfdx pkg install nextcloud? probably prefixed with phpXX-

pkg install nextcloud-phpXY if you don't want to change anything

antranigv: that will pull in mysql as a dependency; i prefer postgresql. what should I do in this case?

i have been running nextcloud for a number of years and my nextcloud database that i've always used (and have backed up, and plan to use to populate my freebsd nextcloud installation) is in postgresql. so i can't just switch to mysql right now.

oh

I was writing a long message on how to do it properly with Jails

for that you'd probably want to use poudriere to build nextcloud yourself and then pkg add it (no need to host a repo using nginx)

but in that case... you might need to build it yourself

but

before we do that

let's check the Makefile and see if you can install from pkg AND just install the deps for PSQL version, without building

meanwhile, can anyone help me with NFS? I just need a rubber ducky

I poked around the Makefile and saw both a MYSQL_USE and PGSQL_USE line items.

I also want to enable redis -- which the normal package doesn't pull in.

dfdx I think if you install the pdo's manualy, you might be able to get away with it :))

but I might be wrong

when I run `cd /usr/ports/www/nextcloud && make config` I see options both for postgresql and redis, which I selected.

antranigv: when you say install the pdo's manually you mean via packages?

dfdx yup!

sure. I don't mind doing that. do I basically just do everything manually except for installing nextcloud then?

maybe not

actually, I think I would do `pkg install nextcloud` and then do `pkg install <whatever deps needed for psql>` and finally do `pkg install postgresqlXX-server`

It may not necessarily be nextcloud that's dependent on mysql, maybe something else in the dependency tree is dependent on it.

TommyC would this help? cgit.freebsd.org/ports/tree/www/nextcloud/Makefile#n57

TommyC: yes, I thought about that. for example, when I ran `make config-rescurive` i was getting config options for sqlite. I can imagine a scenario where one of the deps itself requires sqlite.

i feel like i'm in dependency hell all over again.

antranigv: lines don't render well in lynx :/

but it is looking like nextcloud itself is what depends on sql and not one of its dependencies (which makes sense to me)

i'm not sure if i should pay attention to `make all-depends-list`, which is 945 lines long...

if this is like a fresh brand new install then it's gonna be pretty long, yeah

it is a fresh band new install. i have installed bash, ssl, and... I think that's about it. i'm not sure where to get a list of hte history of packages i've installed.

but it's less than 10.

oh, i installed nginx. so that's at least 3.

dfdx: FWIW i'm not personally aware of any easy way to say: "build and install this port with these configurations" since i've never needed to really do that. the way i build my ports is by using poudriere to basically build the entire port collection and then use different machines to pkg install from that single repository. you could avoid hosting the repository and then simply use pkg add

/path/to/pkg.txz instead of pkg install .. but it's still not as simple as emerge if that's what you're looking for

dstolfa: so you build every package in the ports tree? and then pkg add is just like a glorified install(1)?

i guess I still don't see how that helps if you want to change dependencies, e.g., from mysql to postgresql, as an example.

dfdx: i build everything because the boxes i need to do it on break KBI which means i can't just use the official freebsd ones, i don't personally have usecases where i need to just modify & install a few things

well, poudriere allows you to specify options

so you can say to use psql and redis instead of mysql

dfdx: They did say "... i've never really needed to do that"

Fair enough. But isn't that kind of what `make config` does in /usr/ports/www/nextcloud?

dfdx: pretty much

maybe I can skip running `make config-recusrive` entirely?

Just curious though, why can't you convert the database (from pgsql) to mariadb?

yes, but it does so in a "dirty" environment as opposed to a clean one in a jail using a tool that is widely tested for building your own ports

TommyC: I don't know how to do that, and I

i've worked on postgresql for years and don't really want to switch to mariadb just becuase FreeBSD can't handle dependencies :||

it would be nice to have a tool that does the equivalent of a gentoo emerge, but i'm not aware of one atm :/

for a while there was synth that i believe was somewhat close to it, but i don't think it's supported anymore

Gentoo's emerge is phenomenal. the problem is that you get all the other linux problems that come with Gentoo.

that aren't necessarily gentoo-specific.

What are "... all the other linux problems ..."?

like having to configure your own initramfs if you want to combine LUKS and ZFS.

I have had nextcloud on MariaDB, but migrated all the instances to PostgreSQL one by one

dfdx: Does dracut not have a plugin to support that set up?

dfdx: oh actually.. last commit is a week ago: github.com/jrmarino/synth

maybe this helps?

not sure if this is the most relevant critique of gentoo

given dracut supports that out of the box

TommyC: the point is why do i even have to think about it? I could care less about the intricacies of my bootloader.

i don't think synth does quite as much as poudriere to ensure clean environments for builds, but it does have a synth install that could solve your problem

in fact, the default config file has both of those modules in it

you don't even need to edit it.

I was using portmaster in the past, but it's still time consuming process when compared to centralised poudriere server

duncan: if you can find me a tutorial that lets you install ZFS on LUKS encrypted hard disks and walks you through using dracut to roll your own initramfs I will read it.

configuring dracut is definitely in the handbook

it is possible that ZFS config isn't in the handbook, I don't know

duncan: ZFS is not in the handbook.

LUKS is also not in the handbook.

ZFS has its own wiki, ZFS/RootFS. But there are problems with it, and it doesn't incorporate LUKS.

in FreeBSD this is all very easy with GELI.

I'm not really sure what your criticism is, is it that it's gentoo? or is it that it's linux?

It's that its linux.

OK, typical IRC blether, then

I was trying to praise gentoo, not criticize it, so not sure why you jumped on that.

more to the topic, dfdx, if i was trying to do what you're trying to do now, i would probably look at testing synth (pkg install synth) in a local vm or something to make sure nothing goes horribly wrong, and if it works, just use that probably?

okay, thanks for your help. I will experiment.

dfdx: pastebin.com/98U4nqa7 <- pkg install $i , one by one, and then you can install nextcloud by hand

mzar: amazing, thank you. I tried coming up with a similar list, but it wasn't as complete as this.

could anyone invite github bot to #freebsd-pulse again ?

VVD: your suggestion quite good.

u r welcome :-]

VVD: I believe I will need to recreate. I am talking about zroot pool (not sure if the same applies) I am not that knowledge on it.

just try

imagine a freebsd standard installation with zfs on root using 2 disks striped. thats what I have now.

I want to convert that freebsd install from striped disks towards mirrored (so I get redundancy of one disk)

I can always reinstall... but I was wondering if any other way, therefore, my questioning.

cybercrypto: you can do that

see zpool-remove(8)

zfs is flexible

if I remove da1 from zroot pool - I believe I will loose data (since the zroot pool is stripe and not mirror yet

cybercrypto: read zpool-remove(8) first

remove: Removal of vdev 1 copied 15.1G in 0h0m, completed on Fri May 15 20:04:34 2020

your zpool will remember this

mzar: you saying that even if it is currently striped 2 disks = removing one disk will make zfs to recopy data into the remaining single disk?

I can't comfirm, neither deny, please read zpool-remove(8)

mzar: I can successfully use p9fs to mount root as you guys suggested above. This is only in CURRENT though, correct?

GOD DAMN IT JERRY

all this time

it was the NFSd tunable

and it was stuck in V4 mode

and I kept wondering "why isn't V3 working????"

I think it's a bug.

I will to reproduce today

OK, good luck with debugging antranigv

thanks for reporting FragmentedCurve

mzar: this is fabulous = remove: Evacuation of /dev/vtbd1p3 in progress since Fri Jan 3 20:24:48 2025

mzar: by removing one disk out of the vdev strip pool - the zfs is populating the remaining disk to guarantee the data.

cybercrypto, zpool remove copy all data from removed disk to remaining

VVD: Yes!

I tested this behavior several days ago

it works since we switched to openzfs

VVD: now, I have a 'spare' disk to attach again (this time, setting as mirror vdev)

worked like a charm

cybercrypto, I already wrote commands remove and attach

backup is stanfby anyway... but I can confirm I will not use bkp for this operation :D maybe another time

VVD: I know, I acknowledge your suggestions (history chat above) appreciate that.

running zpool labelclear on the device before attaching would be nice