there any modern ci/cd that's opensource and also actually really good?

where in the handbook do i can learn zfs boot environments?

polyex: nowhere bugs.freebsd.org/bugzilla/show_bug.cgi?id=263331 (but you could write this!)

Title: 263331 – Explain use of bectl (if using ZFS) in the updating FreeBSD section of the FreeBSD Handbook

unixwitch so we got manpage for bectl just not handbook docs?

yeah

Title: Managing Boot Environments | Klara Inc

looks like boot envs have something to do with jails optionally too

polyex: yes no: you can jail into a BE, like you can chroot / jail into anything

meena so i have host freebsd, i take some be snapshots, then instead of rebooting into a be to test it out, i can launch it in a jail?

or, you can launch a jail do an update, reboot into it, see if works, and activate it for real if it does

so it's like, be's are shared. that's kinda weird. so if a jail is booted with a be and makes a change, then the host is rebooted into it and activated, whatever the jail did to the be now reflects for the host running it?

it's not that the BE is shared, just that any changes you make to a BE (including a non-active one) will be visible if you reboot into it... booting a BE in a jail is just one way to make changes to it

ya but a BE belongs to 1 system, so i'm just surprised another system (a jail) can run it

wouldn't they have different configurations?

depends what it's running, jails can share an IP address with the host and the hostname is set in jail.conf... you might have issues with network daemons not starting properly

also mounting /var on both the jail and the host seems a bit iffy

when we create a BE, what paths or files does it snapshot? i'm wondering what base considers "the operating system" to be

although the default BE config seems to put /var inside the BE, so perhaps that's okay

polyex: ROOT/<bename>, unless you specify -r, in which case ROOT/<bename> plus all child filesystems

anything outside ROOT/ is not part of a BE

so only stuff in /root?

ROOT (in caps), /root is the user root's home directory

ohh, ROOT is the zfs dataset right?

yes

zroot/ROOT/14.0-RELEASE_2023-12-22_232055 8K 783G 1.39G /

zroot/ROOT/default 13.0M 783G 6.16G /

zroot/ROOT/main 11.6G 783G 6.60G /

how can i find out what dirs and files are in zroot/ROOT?

-r should probably be default

find / -xdev -print, except this won't include any child filesystems of the root BE... but those aren't used by default, there's only one fs per be

(for a BE other than the active one, find /mnt -xdev -print, assuming the BE is mounted on /mnt)

meena: i saw something about this the other day, it seems there was a desire to only enable -r if bectl detects a 'deep' BE, but i'm not sure why that was... it seems like you'd always want it enabled

deep?

deep meaning there are filesystems beneath ROOT/<bename>, like ROOT/<bename>/usr or something

ty

sway with multiple monitors seems kind of weird, i thought the workspaces would be global, but they're per-monitor, so you have to look at the bar to see which workspace is on which monitor... i guess you get used to it

i'm not looking forward to migrating to wayland. i don't know it for fact but it seems like it's not clearly better than x, only arguably

it's not generally better *but* i do think the design is better in the specific case of KMS/DRM, which is nearly all x86 graphics cards now. since everything X used to do on a driver level is now done by the kernel, and modern applications have abandoned things like Xt entirely

but isn't it cool that x lets you run bin on 1 comp but show the gui on another over network?

i'm still a bit suspicious of it because some people (like GNOME) are pushing weird ideas like client-side decorations which i find pointless

ya gnome sucks

polyex: the X model was nice for things like dumb terminals, but those are hilariously insecure nowadays (unless you use IPsec or something, maybe...), Wayland doesn't prohibit something more like RDP for that although i don't know how much that exists today

(i know there's some support for VNC, but that seems to be more about connecting to Wayland remotely, rather than running remote apps locally)

isn't it cool that… yes, yes it is, but how many people use that?

^

although to be fair i do run X apps via ssh X forwarding occasionally

ofc you can still run X apps in Wayland but it would be nice to have native support for that, i think it's still a somewhat common use case

let's try some other exercise: isn't it cool that Unix emulates a 1940 teletype? yes, sure, but wouldn't it be cooler if it pretended to be a computer?

I'm always running X apps remotely. Mostly xterms....

fwiw, this use case is common enough in some environments (enterprise desktop) that companies like Microsoft and Citrix have entire product lines built around it...

true

modern Unix Hardware en.m.wikipedia.org/wiki/Teleprinter

Title: Teleprinter - Wikipedia

i wonder how hard it would be to get the apple magic trackpad working in freebsd. i think there's an open source Windows driver somewhere...

or does atp(4) already support it?

(the manpage says it's for internal trackpads only...)

unixwitch: Linux has it working so, no?

ah, it's supported by bcm5974(4) apparently

any tip for how to fsck 8 disks at once without opening several terminals?

cant be done at boot becouse they need to be attached by geli first

tmux / screen or: fdisk CMD PATH & fdisk CMD PATH & ....

suffinxing & basically backgrounds a task

suffixing*

you also could use 'daemon' its in base

chatgpt made a script for me using tmux, ill try that

thanks

np

beware that both tmux and screen are not in base (if that matterS)

yeah I use tmux from time to time :)

sopparus how accurate are the scripts chatgpt makes u?

:-)

polyex: for stuff like this its almost 100 %

nice

I managed to get bing AI to write perl a week or so ago

more advanced stuff, not so much..

was quite impressed

after all these years, we finally have a use for generative "A" "I": writing Perl.

:)

I like just chatting to it and asking it increasingly silly things its quite entertaining

when I asked it to write a websocket adapter in POE it actually nicked all the code from a module I wrote years ago which was intersting

i'm trying to sketch out support for automatic IPv6 CLAT configuration (for 464XLAT/NAT64)... the macOS approach is that the CLAT prefix in the RA is configured as an interface attribute (visible in ifconfig), which seems like a reasonable way to make it available to something else that does the actual config

probably better on the forums or the mailing list ;)

yeah, i haven't thought about this enough for that yet :-)

There used to be (back on leenode) #ipv6 that was very good for well chat about ipv6 it may have moved across here, worth a shot at least

#networking is still here, they might be able to help too

i think on the networking side everything is already present (ipfw can do the actual CLAT), it's just a case of working out how to fit the parts together

#perl ;)

are we no longer getting 14.0-STABLE images? freebsd.org/where

Title: Get FreeBSD | The FreeBSD Project

if all else fails, just throw some perl at it?

or throw perl in general :P

jbo, download.freebsd.org/snapshots/ISO-IMAGES/14.0

Title: Index of /snapshots/ISO-IMAGES/14.0/

unixwitch: hm, I looked up "flang" and it is a tool ("miner's pick with two points"), well but as slang it seems to be used as whatever you want according to www.urbandictionary so I'm sure you can use it as swear word

you can use whatever you like as a swear word if you try hard enough

and regarding perl mcmillen.dev/sigbovik

Title: 93% of Paint Splatters are Valid Perl Programs | Colin McMillen

if you use a swear word enough, it can even change from a swear word to a "normal" word

thank you daemon

np

i don't suppose there's an existing script which can be used to bootstrap a bhyve vm from a source tree? (i can do it manually, just trying to save the effort)

unixwitch: we got build targets for various vm thingies

ah... release/Makefile seems to have something for that. thanks

i forgot release includes vmimages now

unixwitch: regarding multimonitor sway, it just does what i3 does there

that's probably true, but i've never used i3, so that did not inform my expectations :-)

unixwitch: speaking of vm images.. i got a response on teh CURRENT mailing list on the installer stuff.. it looks like that there is a "propogation delay" with images and checksums usually take time to "catch up"

not sure if that is a valid answer.. but it is STILL doing it, so going with that unless something changes :)

voy4g3r2: yeah i saw that, but i don't really see how that would break installing from disc1... the mirror should replace the files atomically?

and that shouldn't be the cause of the issue on 14.0 since that was released ages ago and couldn't possibly still be syncing to mirrors

(the problem Alfonso mentioned is when, say bootonly has synced to the mirror but the sets haven't, or something like that, right? but disc1 should include both the sets and the manifest...)

how do you guys backup your machines? zfs send|recv ?

jbo: yes, i use syncoid (part of sysutils/sanoid) to do it, which so far seems to work well

it has some nice features like using ZFS user properties to disable backups for particular filesystems

I'm trying to figure out how to set up a proper ZFS send|recv based backup system. basically my idea is to have a backup storage destination server in my rack. all other servers are zfs sending to that machine. And then from that backup storage machine I would selectively backup to a remote location (which is $$$)

i.e. I currently have a 1TB account at rsync.net where I can zfs recv (this already works)

but locally I want to backup about 10TB

unixwitch, so you don't dick around with custom shell scripts for your backup purposes?

jbo: i used to, but not anymore, syncoid does everything i need. i *think* it should work for your use case, since you could set the do-not-backup property on the filesystems on the local backup server ... but i haven't tested that.

unixwitch, yeah so currently I too have custom scripts for the backups via zfs send|recv. I perceive my setup as "fragile".

unixwitch, one of the things I feel I never did properly is incremental backups. Basically I have to scripts: One which creates the initial snapshot on the source host, sends it to the destination host and then keeps the snapshot alive. and then another script which does the incremental backup. but that seems so fuckly.

yeah ... this is one reason i didn't want to write my own scripts. doing incrementals right and handling all the edge cases is not trivial

exactly. another problem I had was dealing with mountpoint properties.

I would like them to be replicated, but then also I managed to make the destination host fucked up because it mounted the incorrect datasets :D

In today's "I'm not working" accomplishments, I've fixed two Nginx @LibreNMS graphing issues (it was permissions for nginx-status on the host) and resurrected TimeMachine backups on my air01.

nice dvl! :)

jbo: thanks. It's been a good week so far.

All these things which linger because you're doing other things...

I got about 5 years worth of those :p

this reminds me that i'm actually not backup my desktop right now, i should probably fix that since it only has a single crappy ssd in it

unixwitch, sanoid seems somewhat reasonable on first glance. Although I would have preferred a solution that doesn't advertise it as "for linux, but might maybe sometimes partially work on FreeBSD" :D

unixwitch: but ... with zfs on it?

I like sanoid and syncoid (same package) for snapshots. Very flexible for my stuff.

dvl: yes, but i'm paranoid about this because i've had (consumer) SSDs lose all data on power failure before. perhaps newer ones include power loss protection but i don't really trust cheap hardware

I use syncoid to backup my bhyve instances

unixwitch, so does the zfs recv host also need to run sanoid/syncoid?

unixwitch: I've taken to using a Documents directory - all data goes in there. Then I use syncthing to replicate that elsewhere. At that elsewhere, it is snapshot'd and backed up.

jbo: no, from the remote host's point of view it's just connecting via ssh and running some zfs commands

excellent

unixwitch: The bonus, I have more than one laptop. The Documents directory is synced across ALL of them.

dvl: i've been thinking about trying syncthing, but i have things like (for example) 4 copies of the ports tree that i wouldn't want to put in there :-)

so when shit hits the fan and everybody is dying from some weird fungal infection I can always walk up to my backup host, connect a terminal and I see all the stuff on my ZFS filesystem there without the need for any special tooling? sandoid/syncoid is only "client" ware for the backup source host?

unixwitch: You can easily tell Syncthing which things to sync.

jbo: The result is just ZFS filesytems.

jbo: yeah. on the host you just see all the files on the local filesystem, so you can access them, or zfs send them back to the client, or whatever. (note that by default this means it sends encrypted datasets unencrypted... but you can enable encryption on the backup host separately, or there's some way to make it use zfs send -x but i haven't tested that)

so I'm basically replacing my custom crap shell scripts with sanoid/syncoid? that sounds like the solution I was looking for. I really don't want something that does something weird/non-base on the backup destination host. it also sounds like unixwitch went that path already.

this sounds exactly like I was looking for :)

s/on the host/on the backup host/

thank you guys - much appreciated!

jbo: no worries, enjoy.

jbo: dan.langille.org/2019/11/11/zfstool…-sanoid-snapshots-on-the-local-host and dan.langille.org/2022/12/27/using-s…backup-zfs-snapshots-home-assistant

Title: zfstools & sanoid – snapshots on the local host – Dan Langille's Other Diary

jbo: one thing to be aware of is, if you use sanoid on the client to do auto-snapshots, you might also want to run sanoid on the backup host to have to automatically prune the old snapshots, because syncoid will send those over. but that's optional

dvl, that looks like it's your blog/guide?

(just so I know how trustworthy it is)

jbo: Yes, I wrote that.

excellent - thank you!

unixwitch, I seem to remember from my previous custom crap scripts that there are some flags for zfs recv that would automatically destroy older snapshots or something like that, right?

i haven't looked into that (because i actually want this behaviour) but possibly, yes... you can give custom zfs send/recv to syncoid i believe

custom arguments, i mean

aye

one thing I hate about my current solution is that my source host is ssh'ing as root to the destination host. that is really not pretty. but I assume that has nothing to do with syncoid/sanoid. I remember something about zfs allow-user or similar.

oh, dvl's post actually has some zfs allow magic in there - I better read that now :)

personally, I'd like to have one account per source host.

so i cleaned the grill over the fan intake on my desktop, the idle CPU temperature dropped by 5C and the peak temperature during poudriere dropped by about 10C... i probably should do that more regularly

my solution to that is to simply live in a cleanroom

jbo: yes, zfs allow, so no ssh as root. Also, you can restrict the ssh usage to certain commands only. via ~/.ssh/authorized_keys

unixwitch: got graphs? Post them on the intrawebstuff

no graphs, although i'll probably put it in grafana at some point so i can post some next time i clean it (in 5 years' time) :-)

dvl, any chance this works with doas too? AFAIK the major restriction of doas vs sudo is that doas cannot do wildcard command whitelists

unixwitch: I noticed my IPMI snmpd interface provides temperatures.

jbo: I'm sure it does

for monitoring stuff (including IPMI) I can recommend zabbix.

i just have dev.cpu.0.temperature displayed in my desktop taskbar at the moment. should check if that appears in node_exporter

it doesn't really matter right now but it would nice to have in summer for our server since it gets quite hot in that room

One thing I've separated: Nagios for monitoring (is it running?) and LibreNMS for metrics (how much foo has bar done today?)

dvl, or just zabbix for both :>

jbo: I've never had a reason to move from Nagios.

at an old job we used to use Ganglia for metrics... that was pretty nice for monitoring large numbers of servers, but nowadays it probably doesn't do anything grafana+whatever can't do

It reminds me of : use a web browser for web browsing, use a mail client for reading mail, and a news reader for reading news. Dates back a few decades.

unixwitch: agree..i was not 100% sure what that even meant.. but i am leaning a lot towards virtualbox

but finally got another 4tb hard drive on sale.. time to see if i can 'swap' out a 3tb to make an offsite backup

can I disable all DOCS entries from ports by default?

also is it possible to get all of these option windows out of the way? I think my compiling would have been done long ago without them

fowl: i guess "OPTIONS_UNSET+==DOCS" in /etc/make.conf should do the first thing, although i've never tried it

fowl, you can have OPTIONS_UNSED+=DOCS

use 'make BATCH=yes install' to skip options dialogue entirely, if you're installing manually

unixwitch, it's +=, not +== (I do assume it was a typo)

er, i meant to type +=, yeah

I usually have OPTIONS_UNSET+= for EXAMPLES, WAYLAND and PULSEAUDIO :>

thanks friends

[00:09:11] [01] [00:00:19] Finished sysutils/sanoid | sanoid-2.2.0: Success

here we go

fowl: fwiw, if you're building ports frequently, i recommend looking at poudriere or synth (which build packages on a local filesystem you can install with pkg) - or at least portmaster or portinstall, which automate a lot of tedious stuff

er, portupgrade, nor portinstall

a not-so-complete "complete guide on poudriere" I wrote when I first started using it: blog.insane.engineer/post/freebsd_poudriere_guide

Title: Poudriere: A complete guide

fowl, I can certainly recommend using poudriere even if you're just building locally.

the nice thing about poudriere/synth is not having to deal with these "python/perl/whatever upgrade" flag days... it just does the right thing by default

i remember having all kinds of hassle with those with postupgrade... back before we had poudriere

no poudriere no fun

even when building everything locally

s/postupgrade/portupgrade/ ... i do not know why i cannot type anything today

then you can at least easily do resource restrictions on jails etc.

unixwitch, too much whisky?

vodka, but yes, probably related

hah :D

time to submit drunk patches to fix net-im/tokodon

don't do that

jbo: it's trivial enough that i think i'm safe bugs.freebsd.org/bugzilla/show_bug.cgi?id=276025

Title: 276025 – net-im/tokodon: add purpose to USE_KDE

(and i did test it)

hm... are we meant to bump PORTREVISION for these? i think this is the first time i submitted a patch to an existing port

yes, you should.

although I am only a committer since september so probably ask somebody more experienced to be sure.

but from what I know you should bump PORTREVISION for this.

the rationale is that anything that changes the resulting/built package binary should do a bump

yeah, i guess this is required for pkg to pick up the change

yep

dvl, regarding your blog post on syncoid: why are you adding 'mount' to zfs allow on the receiving host?

jbo: I don't recall, but I think it was required to get it to work.

hm, now i got it to build, i don't actually like tokodon very much, it only has a single column layout... but it seems to be the only mastodon client that isn't web-based or an electron app

unixwitch, time for an upstream patch then ;)

i think adding multi-column would be a pretty huge project. maybe i'll have a look if i can't find anything better

dvl: You generally don't need to mount a dataset as part of send/receive. In fact, oftentimes you explicitly don't want it mounted.

mason: All I can say is I didn't add it without reason.

dvl: I have an example of explicitly suppressing mounts in my tool: github.com/ChibaPet/backup-zfs

Title: GitHub - ChibaPet/backup-zfs: Shell scripting to automate back-up via ZFS snapshots.

mason: If you go through the article, there's a lot of trial and error document. However, I definitely do not recall the mount issue.

SCSI sense: ILLEGAL REQUEST asc:21,0 (Logical block address out of range) <--never good when you had a drive to a box

was reading about ucl and someone said it fails the norway bug but what is that?

.no

reminds me of: ruudvanasseldonk.com/2023/01/11/the-yaml-document-from-hell

Title: The yaml document from hell

what's the tld got to do with ucl?

ahh

read the article I linked :)

> {"geoblock_regions": ["dk", "fi", "is", false, "se"]}

:D

so the norway problem is basically the pain of dynamic typing, applies to config langs

polyex, yes: hitchdev.com/strictyaml/why/implicit-typing-removed

Title: The Norway Problem - why StrictYAML refuses to do implicit typing and so should you - HitchDev

so what's a better config lang then?

personally I like XML

but you'll find plenty of people hating on that as usual with everything

dvl: Can you link the article? I'm just going off of channel traffic but I'd enjoy reading it.

not very user friendly but i also like xml

polyex, for simple stuff I like TOML

toml doesn't have norway problem?

mason: [Dec 30 12:24] <dvl> jbo: dan.langille.org/2019/11/11/zfstool…-sanoid-snapshots-on-the-local-host and dan.langille.org/2022/12/27/using-s…backup-zfs-snapshots-home-assistant

Title: zfstools & sanoid – snapshots on the local host – Dan Langille's Other Diary

ty

polyex, TOML has problems too. XML is the only thing I know that works reasonably well "everywhere"

hence all my shit is XML

i usually write my own config parser with a named.conf-like format. i know this is not trendy nowadays...

would you make a config file format for a program be in xml format tho jbo?

polyex: TOML requires quotation for strings

unixwitch, trendyness is part of the problem. XML isn't trendy. but if it works it work. I see no reason to re-invent the wheel all the time just to end up with yet another solution that just has other problems :)

polyex, yes. I would. I use either XML or TOML.

what about cue or nickel?

shall we discuss all 74820 obscure formats now? :D

some prefer Lua for their configuration files

jbo: i have noticed that about half the time, when an application uses an XML config file, it applies XML in a really weird/unidiomatic/bad way

to which I say, I salute you

Java is usually okay for this but other stuff not so much

unixwitch, I have also noticed that some people use screwdrivers in weird/unidiomatic/bad ways

abusing a tool does not make the tool bad IMHO

XML is kind of hard to read/edit for an unitiated person

see JSON. nothing wrong with JSON but for some reason it gets used in places where it really shouldn't be. doesn't make JSON bad.

jbo: i didn't mean to say that XML is bad, just mentioning my experience with it :-)

unixwitch, you're right - sorry if my response sounded triggery.

imo JSON should only be placed between the browser and the server

Remilia +1

one of the main things I like about XML as a format is that it properly supports attributes

note browser specifically; I think actual non-browser client-eserver APIs should not use JSON

<timeout type="ms">400</timeout>

no question what 400 is.

-e*

err, that should have been unit="ms"

XML is also great because you can process it without programming. XSLT is great.

"oh, you want to have an HTML rendering of this config file? there you go"

"you want to have a flattened hierarchy of this nested node? there you go"

and then of course xpath. you can actually search XML files properly.

i just twitched seeing this xml stuff and think of ICH e2b(M) transmissions and how ugly it was: admin.ich.org/sites/default/files/i…les/ICH_ICSR_Specification_V2-3.pdf

I think I extolled the benefits of XML a week ago on tokio discord hahaha

i can probably still be tracked in the european union working on that stuff

Remilia <3

i had to give all my info and my first born.. to be able to work on that

but for client-server communication people should really consider using protocols other than HTTP(S) and running strictly typed stuff

or especially server-server

+1

I wonder if JSON has the Ontario problem because I think that one is abbreviated ON

er

YAML, not JSON

and on is true in YAML

i remember the horros trying to use ANT configuration which was all xml

and the horrors of java

Remilia, last time I checked on and off were supposed to be treated as strings in YAML

for the love of god, why can't this damn machine find the new drive i just added

voy4g3r2, camcontrol rescan all?

camcontrol devlist shows the drive!

(don't do that if you have a dozen storage racks with 4000 drives) :D

haha, i got 4

voy4g3r2, if camcontrol shows the drive then I don't understand why you say that it isn't finding it?

i just want to add this 4tb one and take out a 3tb one.. so i can have some kind of offsite backup

voy4g3r2: how are you determining that the system can't see it? note if it's a new disk it won't appear in gpart until you actually create a partition table on it... someone else here ran into that recently

jbo: because i try to add to the zpool for reslivering and it says no bueno

errr!!! i am following the zfs docs on freebsd, it does not mention gpart :)

just add the disk and do a zpool replace

jbo: you posted a link earlier about the YAML docoment from hell, quoting, 'The literals off, no, and n, in various capitalizations (but not any capitalization!), are all false in yaml 1.1, while on, yes, and y are true.'

voy4g3r2, you have to partition it before resilvering, no?

i am also cranky because penn state is losing to ole miss right now

voy4g3r2: can you show the output of camcontrol devlist, and the zpool command you're running and the error?

Remilia, you're right. part of the big mess with YAML is the various specs. it's possible that I am quoting some 200x spec

voy4g3r2: yeah gotta watch Ligue 1 or La Liga

Remilia: i am trying to have some kamasai washington the epic going.. i have the game on silent

regarding resilvering: one thing I learned the painful way is to always create a zfs partition on each disk rather than using the full disk. one has to replicate that gpart partition table before resilvering, no?

voy4g3r2: those do not really tell me anything haha

Remilia: en.wikipedia.org/wiki/The_Epic_%28album%29

Title: The Epic (album) - Wikipedia

this is jazz heaven!

just a little under 3 hours

but anyway.. need to get outputs onto bsd.to

Title: dpaste/YQN4 (Plain Text)

i am scrubbing as the dang zpool found 5 errors

> scrub in progress since

can you even modify the pool during scrubbing?

> missing pool name argument

storage is the pool name

because i am "smart"

voy4g3r2: that seems like some sort of write-protection on the disk...? or something... does 'gpart create -s gpt da3' work?

but you didn't specify that, voy4g3r2

voy4g3r2, I also usually name my pools 'storage' - I see nothing wrong with that :)

jbo: yes, first time i made a mistake, i ran it again.. with the name

voy4g3r2, right - my bad.

let me try that gpart thingie

voy4g3r2, can you actually read from the disk?

unixwitch, wouldn't it be more wise to copy the partition table from one of the other disks?

i did not do what unixwitch suggested yet

jbo: i just suggested that to see if a partition table can be written at all (since that's what zpool wants to do)

can always replace it with another one after

my goal is to REMOVE that diskid/blah blah with this new drive

unixwitch, aye. reasonable. I was previously asking whether zpool automatically creates the partition table.

then MOVE the diskid blah blah.. to my pole barn to have an "offsite" backup

zpool does not know anything about partition tables

that's... true.

and you do not need a partition table if you are not booting off those drives

hmm, in Solaris, zpool would write a GPT label to the disk when you added a whole disk to the pool. has that changed?

this does make me feel _a little_ better

i am like.. just put the drive in, it will just "work"

voy4g3r2, one of the things I learned is to create pools with GPT labels: blog.insane.engineer/post/freebsd_zfs_gpt_labels

Title: FreeBSD: ZFS pool using GPT labels

my old colocated server had zpools over raw disks

Remilia, the problem with raw disks is that if you replace a disk and it happens to have a few bytes less than the other ones you're screwed. Hence I always create a dedicated partition on each disk leaving a gig or so free.

i am hoping after this is all done.. i am changing that damn raid setup

# gpart create -s GPT /dev/da3

gpart: geom 'da3': File exists

well thanks..

voy4g3r2, gpart show da3

# gpart show da3

=> 40 281474976710576 da3 GPT (128P) 40 281474976710576 - free - (128P)

if you're sure that is the "new" disk I'd just nuke it via gpart

128P?! that does not seem right.

jbo so when you said no earlier, were you saying ucl DOESN'T suffer from the norway problem?

jbo: bought it 2 hours ago.. it better be.. it was even in a bag

assuming there's no data on this disk, maybe 'gpart destroy da3' and then try zpool again

polyex, no, I said ".no" when you "asked" what the norway problem is :p

voy4g3r2, oh, is this an external USB drive? those usually come with partition tables from the fab

voy4g3r2, gpart destroy da3

ohh lol

so does UCL suffer from norway problem?

jbo: an internal drive in a usb enclosure.. yes

hahahahahah

128 PB is insane

# gpart destroy /dev/da3

gpart: Device not configured

it is 4TB

voy4g3r2, gpart destroy -F da3

voy4g3r2, alternatively, just dd if=/dev/zero of=/dev/da3 bs=1M count=1024 or something like that to nuke the partition table.

voy4g3r2: `sysctl kern.geom.debugflags=16; dd if=/dev/zero of=/dev/da3 bs=512 count=2`

or maybe more in count

well the destroy says nof confifugred again

Remilia, yeah that is more accurate. I tend to overdo it on the bs/count for "safety" :D

voy4g3r2, just dd the first few bytes.

jbo: you need debugflags too

Remilia, what? why? never did that before.

well, maybe this changed, but before you could not dd over a partition table without it

protection

Title: dpaste/NRuB (Plain Text)

that is just crazy

voy4g3r2: your drive is DOA

Remilia, interesting - I didn't know that and certainly performed this "act" without it successfully several times.

ya ucl does suffer norway problem looks like. DAMN

check dmesg

voy4g3r2: do you see anything in dmesg?

Title: dpaste/oY4F (Plain Text)

hmm... can you read smart data from a USB drive?

I mean, it should not show 128 PB

ahhhH!!!!

voy4g3r2, does /dev/da3 even exist? does dmesg show that if you plug it in?

either the drive is DOA or cable failure

Remilia, or USB host port :>

it certainly exists judging by dmesg

that's a weird error though, it must be really broken

jbo: i would not think the USB host thing as i am listening to music off the zpool

is that a multi-drive enclosure

yes

.__.

I dunno, either check if the drive is sitting in it properly or RMA it

unless that error is from geom trying to read the backup GPT label at 128PB...

voy4g3r2, so the disk you're replacing is a standard "internal" HDD that you mount in the enclosure?

(which would certainly be out of range)

jbo: correct

voy4g3r2, and you're sure that the disk is fucked and not the controller in the enclosure? :D

jbo: re debugflags, correction: it only applies to MBR stuff, sorry

is that one of those enclosures with a built-in USB hub and then one USB-to-SATA per drive?

i am defaulting to drive, at this point, as there are 3 other drives in the enclosure.. and i am listenign to music off that device

Remilia, good to know - thanks for the info! Why is that tho? MBR surely is in a similar place as GPT at the beginning of the disk, no?

ask someone else, I always assumed that kernel simply tries to protect the MBR

jbo: amazon.com/dp/B0734G79FW?psc=1&ref=ppx_yo2ov_dt_b_product_details

voy4g3r2, yeah but if the enclosure has a built-in USB hub and then one controller per bay you might have a broken controller too. do you have a way of hooking up the new disk you bought to your machine directly (internal SATA) port or a standalone USB-to-SATA adapter?

Title: Amazon.com: ORICO 5 Bay 3.5inch Hard Drive Enclosure USB 3.0 to SATA Magnetic Tool-Free External HDD Docking Station Case with 12V/6.5A Power Adapter for Family Storage Expansion Up to 90TB (5x18) - DS500U3 : Electronics

jbo: yes, i have an enclsoure i can put it in

voy4g3r2, I'd try that now.

lol the thermal/cooling design of that enclosure is stupid af

so weird camcontrol sees it.. but after that it is foobar

jbo: they are always like that :\

there are very few good external enclosures

jbo: haha.. it is in a basement, it is plenty cold.. in another area of my house

Remilia, icy has some where you mount them the other way around (rotated 90 degrees)

icybox are nice

yeah

I had a trayless rack for years

i am going to move the drive to another bay an dsee what happens

if the one bay is shot, i am returning this damn thing

voy4g3r2: also it is not about ambient temperature because drives get seriously hot without proper spacing or airflow

yep

voy4g3r2, for fun and giggles: is the broken disk by any chance the one closest to the front panel of the enclosure? :D

haha its the bay

Seagate's enterprise drives went up to 55 degrees without active cooling in a server room with 14 degrees ambient

see, stupid cooling solution. obviously that disk died :p (could be other reasons too, ofc)

simply by being next to each other in a standard tower case

a single disk in the same case never went over 30

Remilia, have you seen the icy dildo? :D

IB-1823MF-C31

I'm trying to slowly upgrade an old system from 11.0 to 11.4 - is that supported ?

well something didn't like something

the whole zpool went down

when I try freebsd-update -r 11.4-RELEASE upgrade I get Fetching 2 metadata files... failed

last1: it might be too old to use freebsd-update

last1, I assume/guess that the official repos no longer provide binary updates for 11

jbo: a single nvme..

jbo: even after new(er) updates are provided, the old ones stay

# zpool clear storage

cannot clear errors for storage: I/O error

oh joy

freebsd-update-build is wicked and you don't want to have anyone get rid of the latest patchlevels for an EOL version

last1: what's the system doing? it might be a great time to rebuild and migrate to 14

too many possible points of failure

I need to check the software running on it first

it's custom

so is there a way to get to 11.4 first ?

Remilia: at this point, i am going to find an "old" computer with multiple bays

i figure this would be "easier"

that is what i get for doing that

brb

xmj, is that a guess or a fact?

EOL is EOL

but first must move to another machine, so i can do this..

luckily voyager2 wasn't built with USB enclosures on board :>

jbo: do you know of an unofficial repo where can I get 11.4 from for upgrade ?

jbo: haha.. my wife goes.. well just buy another raspberry pi..

i kept eyeing these lenovo mff machines

voy4g3r2, RPi is the USB enclosures of the embedded world so not sure if that will be a better solution :p

jbo: yes that was her dig at them

as Remilia, I just don't trust them. but if I need one, then it's icybox

ever since i tried to put pihole on the network.. she hates them

sounds like time for a new wife? :p

i got sick of all this "spying" my work computer was doing on my network.. so i blocked it

(obviously kidding)

and i am like who actually clicks links in google searches

ah, must be something else, I tried upgrading to 12.4 , same error

yeah.. that was the evil eye look

last1, traceroute update.freebsd.org

(or just ping, I guess, if your network "allows" that)

lets try moving this drive to an enclosure

figure if i can isolate the drive issue.. ti is just that raspberry pi of enclosures..

yeah that's fine, I suspect it's some stale files in /var/db/freebsd-update

if xmj is right then you should definitely be able to update from an ancient version

(via the regular channels)

dang it..

what was that gpart command again? sorry.. did not write it down

just type "gpart" and hit enter - it's fairly self explanatory

18:59, <jbo> voy4g3r2, gpart destroy -F da3

?

freebsd-update fetch / install first did it

then upgrade works

go figure

nice! seems like xmj was correct about ancient updates still being available - nice! :)

how can it not be supported

# gpart create -s GPT da0

gpart: provider: Operation not supported by device

is that just a permissions thing?

i am as root under su

or as Remilia said early on: DOA

time to figure out hte RMA process..

or i can just go to the dang store i bought it.. no restock fee

is there a wireless-network-management system that works on freebsd? as in, configure some small embedded systems as access points, and have them connect back to a central server for configuration and authentication, like Ubiquiti or MikroTik or whatever

unixwitch: iirc i got the ubiquiti one working on freebsd. it's just a java jar

rtprio: does that require ubiqtuiti APs? i was looking for something non-proprietary (otherwise i'll just stay with mikrotik)

as in, freebsd controller + freebsd APs

rel ubiquity controller on freebsd: blog.insane.engineer/post/freebsd_unifi7

i don't think you'll find one of those. yes, their management requires ubiquiti APs

Title: UniFi 7 controller on FreeBSD

yes, you need their APs

hm, shame. surprised this doesn't exist, but i guess most people who want this just pay the $$$

I am actually looking for a different solutions. I didn't have good times with ubiquity gear the last two years.

i mean, with puppet (or hand spun gui) and a few freebsd systems with wifi cards you could make this i guess

but a university or corporation aren't going to deal with it

i also am not sure most wifi drivers on freebsd give you the ability to run multiple AP's on the same card?

rtprio: i'm not sure a solution like that would support .11k (or whatever the current version is)?

then again mikrotik still doesn't do .11k anyway... although it's supposed to be coming soon

rtprio, part of the issue would be to handle the handover from one station to the other. not sure how you'd address that with puppet

containing woes of kerberos: postgresql 16 doesn't build with base Heimdal, it's missing a gss function it wants :(

s/containing/continuing

jbo: i forgot about that. freebsd is by and large not up to the task

jbo: that's what i mean by .11k

unixwitch, I'm not familiar with the wifi lingo, apologies :)

is it .11v now? too many wifi standards to keep up with...

seriously

i noticed .11ax is out now and we haven't even upgraded to .11ac

I know very little about 802.11 but I feel like that's one of those standards that got out of hand :D

oh no that's wrong... we're on .11ac, .11be is the new thing. so i guess we're only one version behind since .11be isn't out yet

I dunno, I'm still working with the 2.5MB/s I get on my laptop

we do have net/wifibox for the inclined ones

bz@ is still working on iwlwifi (I hope)

jbo, Do you use wifibox yourself? (I do)

i might just buy my users some USB-C Ethernet dongles for their phones

job, I took "we do have net/wifibox for the inclined ones" to mean that you do|would not?

Sorry, I meant to address jbo ^

parv, I don't. I am one of those "the leaner the better" type people.

so I'd rather deal with 25 Mbps than having a freaking linux vm running to proxy my card

jbo, Understood; thanks

parv, from what I have heard, it works well tho. Personally, I would just set it up myself rather than running some obscure 3rd-party provided VM that gets untethered access to my network card

FTFMP: "wifibox deploys a Linux guest operating system with the help of bhyve(8) ..." Heh, I'd just roll out a home built tiny Linux device if I want to roll my own AP. No need for the overhead of FreeBSD + bhyve. ;) TBH, I'd truly prefer to roll out a home built tiny FreeBSD device, but we're not *there* yet.

unixman_home: i believe this is for using unsupported wifi cards on desktop/laptops, rather than for APs

unixman_home, But using an AP & having to be able to connect to said AP from FreeBSD (due to lacking support for speed or at all) are two different things when

unixwitch, parv, ah, that makes more sense. I was looking at the mention of AP above and was thinking that was the recommendation for a FreeBSD AP.

s/when//

unixman_home, go grab some ThinkPad would be my recommendation.

jbo, yep, that has been a recommendation for a long time now in the FreeBSD camp.

I'm currently using a Carbon X1 Gen9 if that is helpful

but that's mostly because I work on desktop machines and only need a laptop for on-the road stuff so smaller/lighter is preferred. Also had a P2000 mobile workstation with a xeon, ECC memory and quadro running FreeBSD. that was fun too.

I am not going to be getting anything soon. I need to pay off the credit card bill for Christmas first. ;)

yeah, I am currently also living off of water and toast sandwiches :D

:D

I guess vodka is cheaper than whisky

voy4g3r2: btw what make/model is that HDD?

Remilia about to update the HDD bingo spreadsheet

haha

as the year ends, I'd just like to take a moment and publicly claim that FreeBSD rocks

5 commits says it's a WD device

I mean, I had quite a load of HDD reliability data back when I worked at a data-centre

it's so much better than Linux distros, in every imaginable way

last1 +1 better tell the cool kids over at reddit tho

jbo: WD is a lottery, 'was this made at their Seagate factory? no? you might have won as long as it is not the problematic WD factory'

never had any issues with enterprise drives

still feels incredibly insane to me that the mergers were just allowed

gold enterprise / hgst are just rock solid

Remilia, funny that you say that. I hear so many people fame about WD but I had so many lottery experiences with them. I didn't know that they produce partly at seagate but fun fact: once I finally was fed up with WD I only bought Seagate and I am happy since.

Remilia, are you saying that the WD drives fabbed at Seagate are the worse or the better WDs?

i found myself happier buying more expensive drives with the 5 year warranties

in the long run

pretty much all my systems run IronWolf Pro these days

jbo: a very small % of WD drives is outsourced to seagate and they are Bad

Remilia, interesting - what are your experiences/opinions on straight seagate disks?

while WD's ex-HGST drives are still made at IBM's former factories

jbo: I always had really really bad luck with them

Remilia, that's interesting. Seagate is the only brand I never had problems with so far :/

WDs used to die left and right in my scenarios.

at the data-centre WD and Seagate comprised about 80% of the total and had failure rates above 15% over year-after-purchase

WD had ≈12-13%, Seagate was at 18% iirc

hmmm

Hitachi was at 0.8% and Samsung at 1.5%

do note this is from more than a decade ago

Title: Backblaze Drive Stats for Q1 2023

I am subscribed to their reports yeah

it seems like they got better

does look like Seagate is not doing great comparably on first glance

note Backblaze mostly use consumer, not enterprise, drives

oh

yeah hence I said "first glance"

I don't even bother putting consumer drives into a desktop

I ran consumer Hitachi/HGST drives since 2010 in my desktop PCs and servers with zero issues

until I dropped one at the airport haha

it developed like 40 bad blocks

aye

alright, time to leave the office o/

those are consumer grade drives

ok, speaking of drives

i have 93,728 hours on my drive. how much longer can it go?

rtprio: 10 and a half years. With drives old like this I usually had problems to get them running once they stopped spinning (due to power outage)

that is weird

all my HGSTs from 2010 spin up just fine haha

well that was easier than RMA.. they took the hard drive back and are like.. this is the 4th one this week..

errr!! if that was the case, you should just remove from the shelf

Well, have to admit the last drives getting that old where 2 pound scsi ones :)

seere: funny story, i had to replace a raid card in this box and when i turned it on again the bios battery was dea

so i finally got Tokodon (KDE Mastodon client) working, after installing a bunch of stuff related to gnome-keyring, and... honestly, it's not really that good. are there any other desktop fedi clients that aren't based on electron?

unixwitch: how do you feel about `toot' ?

i thought that was for iOS, or is there another project with the same name?

Title: Introduction - toot

it's tui

hmm, that doesn't look terrible, thanks

elk is another one i liked, but it's not desktop: elk.zone

ouch, ports version is toot looks pretty outdated

s/is/of

i think i just used pip

outdated means it might work with non-mastodon instances! ahaha

oh this issue wasn't because it's outdated, it's bugs.freebsd.org/bugzilla/show_bug.cgi?id=274411

Title: 274411 – devel/py-urwid Update to 2.2.x breaks several python packages

enough of this crap.. a trip to a microcenter.com is in order

hodge podge of stuff is just pointless..

25k square feet of computer stuff.. there HAS to be a machine somewhere.

Python and Ruby seem to have such tightly coupled packages. Updates seem to be a struggle, every time.

i'm not sure if there's something stalling the fix for this, it's been in maintainer-approval? since november

The maintainer is python@

maybe they're waiting for reviews.freebsd.org/D39288 to land, although that's been in review since march

Title: ⚙ D39288 devel/py-setuptools_scm: update to 8.0.4

i'm starting to think just installing python stuff via pip might be easier :-)

unixwitch: given the python package version requirements, virtualenv and pip may be a good solution in many cases

yeah, one virtualenv per app, or whatever

although that sounds like hell to keep up to date... perhaps there's an app for that

i think this is just a rough period for python right now with the PEP-517 updates

or python ports, i should say

The very reason i trudge though packages not to need to use single versions of packages. You quickly loose track.

taking inspiration from meena who told me to annoy people to get things done, i mailed -ports about this...

Hi! is it possible to change the boot device ? i copied a disk (with dd) (ada0) to a (nvme0) and now freebsd fails to boot with: Starting filesystem checks: Cant open /dev/ada0p1 UNEXPECTED INCONSISTENCY.. THE FOLLOWING FILESYSTEM HAD AN UNEXPECTED INCONSISTENCY: msdosfs: /dev/ada0p1 (/boot/efi)

oh its the fstab entry ?

hernan: yes, edit fstab to replace ada0 with whatever your new disk is called

probably either nda0 or nvd0

ok thanks

heh: 220-mx1.freebsd.org ESMTP Postfix (Postfix FTW!)

is there no way to view portscout.freebsd.org by port rather than by maintainer?

Title: portscout - new distfile scanner

unixwitch: have you heard of pipx?

Also, very happy to be an inspiration

i have not heard of that

"It is not recommended to install pipx via pipx" Inception fail

ah so this is like ports for python

oh well, nothing is perfect, certainly nothing that's used for packaging

you install python binaries in their own venv, you can also chose which python to use

but... i feel like for the benefit of the User Community(tm) it's better to fix broken python ports rather than give up and use this

what happen to gnome cant install it doest see it in the repos

richard_: what package are you installing?

x11/gnome?

well before i did pkg install gnome but now it doest work

what error do you get?

pkg: No packages available to install matching 'gnome' have been found in the repositories

what freebsd version?

14

that's odd because pkg.freebsd.org/FreeBSD:14:amd64/quarterly/packagesite.txz does have gnome in it

what did they change it to

richard_: can you paste the contents of /etc/pkg/FreeBSD.conf, and do you have and files in /usr/local/etc/pkg?

oh, and what arch are you on? amd64 or something else?

s/and files/any files

i gave mutt another try and with this new sidebar thing it's a lot better for managing multiple accounts

still takes way too much effort to configure (they really need some sort of 'account' concept in the config) but i think it'll do

Title: dpaste/izGu (Plain Text)

richard_: any files in /usr/local/etc/pkg? and what is the output of 'uname -m'?

repos

richard_: any files inside repos?

Remilia, still up?

so the freebsd 2024 survey asks if i "primarily" use freebsd on servers, desktops, or laptops. i have 1 server, 1 desktop, and 1 laptop running freebsd. so... i don't know how to answer this?

FreeBSD.conf

all the below

richard_: paste the contents of FreeBSD.conf?

unixwitch, I had similar issues. I have 8 servers, 3 desktops and 1 laptop .__.

AumShivaya: it is a single-answer question!

and a hand full of networking gear running bare bone FreeBSD ofc

jbo: i actually have 2 servers running freebsd if you count my secondary DNS VPS... so... i guess i'll say servers!

FreeBSD: { url: "pkg+http://pkg.FreeBSD.org/${ABI}/latest" }

unixwitch, I decided to choose "server" because I don't want them to think that they should spend money/effort on trying to make a fucked up desktop OS :D

I just don't really see a point in making a desktop OS if you can have a well engineered, robust and performant server OS on which you can just install desktop components.

richard_: it seems like the latest pkg build does not include gnome. i do not know why that is, but you could ask on freebsd-ports@ mailing list, or on the forum, if no one here knows. quarterly does seem to include gnome.

amd64

I use FreeBSD as a desktop, I feel it is fine, but not for Steam gaming. Oh well.

AumShivaya: have you tried to steam gaming ?

its on a vm

Still prefer FreeBSD over Linux. I do not know why

yeah, I have steam installed under linux emulation

"How long have you been using FreeBSD?" <check '25 years', feels old>

tried installing/playing 3 games, nope.

debian sid is on my hardware

AumShivaya: it works ? i never tested

well, it doesn't load any games properly so not worth

but the Steam UI looks ok

ok

Probably I could try more/different games but nah

but it works what you gave me though x11/gnome

AumShivaya, github.com/shkhln/linuxulator-steam-utils/wiki/Compatibility

Title: Compatibility · shkhln/linuxulator-steam-utils Wiki · GitHub

"Performance is a major security consideration for server workloads" what does this mean

that's 131 playable games

okay, but I do not think it would have any games that I play

plus I do not think that controllers work on FreeBSD Steam

especially not via bluetooth

I learned to separate gaming hardware from productive hardware a long time ago

so I just boot in to Linux on the rare chance I feel like playing a game

makes sense

the only game I feel like playing these days is factorio :x

i've been meaning to Factorio and X4 on freebsd. those are the only games i actually like nowadays

Hi! i copied an hdd with dd, and now it stops the boot process after i select option [1] in beastie (the normal boot).. it displays: Loading kernel... /boot/kernel/.. and then stops with: Start @ 0xffffffff8038b000 ... EFI framebuffer information... dimensions 1920 x 1080 -- what could be wrong ?

and they both have Linux versions, which is promising

unixwitch, what's x4?

oh, scifi simulator thingy - nice

jbo: do you know Elite? imagine Elite, but single player, but with the ability to build space stations and actually influence the economy and take over the universe (if you want)

unixwitch, factorio is really just SDL2 with everything else written in custom C++. I asked them do provide a freebsd build but they were "nah, we're doing nintendo switch instead" :<

(Nintendo switch is basically freebsd)

to be fair, PlayStation is also basically freebsd but we can't run PS games :-D

yeah, I'm not complaining :D

unixwitch, that sounds lovely

unixwitch, also, pm

although it would be nice if they'd do like a Quake-style thing where they release the source to the engine but you still have to buy the data files

the "problem" I have with factorio is that I know how low effort it would be to provide FreeBSD builds compared to other games. But I fully understand why they are not doing that.

jbo: i saw your PM but i am just like "... yeah" so not sure how to reply :D

fair enough

unixwitch, I'd argue that factorio is a lot more engine than assets compared to an arena shooter.

yeah, maybe

i guess if they did release the engine source, it would be like OpenTTD where someone comes along and replaces all the sprites

i can't answer "Which sources do you use most for help with FreeBSD?" becuase you can only give five answers, wtf

only select freebsd journal and move on to the next question :p

(obviously kidding)

i deslected 'wiki' because it's usually the last place i go

it gave me what i wanted ty

"It’s unclear if or when my code will be reviewed" well i'm glad they asked about this one at least

although "Do you consider yourself... A committer?" is a weird question. that's not something you 'consider' yourself, you either are or you aren't

Remilia: it was a wd blue.. don't buy a wd blue..

voy4g3r2, if it helps, I made good experiences with ST10000NE0004-1ZF101 so far

jbo: i got a 3 toshiba drives

and a motherboard, some ram, a case.. power supply

microcenter.com is evil!

unixwitch: some projects welcome people who are committed much sooner as committers than FreeBSD. And some people have enough imposter syndrome that they could have commit bit too src docs and ports and go, yeah, but this is just a hobby, you know

i suppose