it shoudl install to /boot/modules

firmware fiels are compiled kernel modules

not actual bin's

chris@beastie:~ % cd /boot/modules/

amdgpu.ko amdgpu_green_sardine_ce_bin.ko amdgpu_renoir_me_bin.ko

yes, in fact, it does install to /boot/modules, I am not an idiot

says the perosn thikning spamming is asking for help

but ok

i once put in a filter in the wrong spot when I was in .mil

did I feel like an idiot no

i just laughed took it apart and put it in the other location

this is normally when I walk away and go watch some TV

almost never have got along with DOW types, be they contractors, be they gunpowder...

also what the hell is a topic whore, from your gecos

MelMalik: people who will either mention to stay on topic or to goto -social when the channel has been silent for hours

MelMalik: google has a couple articles about it

buildworld completed

6.5 hours

actually not too terrible to buildworld

now use tripwire to check all the checksums from that build and cross reference them to the one in base.tgz

I expected 10 hours but hey ho

just to be safe you know

10 hrs takes me 25 mins heh

pax.

i would have made world on all my systems by the time you finished

what is that a celeron ?

i did build openbsd kernel on a rasp pi once took 3 days :)

MelMalik: fax

cpet: intel.com/content/www/us/en/product…-up-to-3-30-ghz/specifications.html

polarian: i take it that building your system instead of using pre built packages fall under your logic of everything is compromised ?

I notice that freebsd cant turbo either

or maybe thats because I have libreboot

it can powerd is your freind

cpet: powerd doesnt properly do much actually

normally the CPU itself deal with that on newer CPU's

for me

I assume its coreboot nuances

try using something like -j16

cpet: pointless using so many processes, its an old cpu

6.5 hours doesnt seem unreasonable

ccache, nve drive, some magic with zfs

old cpu, HDD, and limited memory :)

also cpet I am compiling from source in protest of pkgbase

i would blame the HD before the CPU

why protest you can still install non pkgbase

because freebsd-update is deprecated now

pkgbase is recommended

good

and freebsd-update will be removed in 16.0

you know how many times I had to explain to people that you need to run that like 4 times

I already had a long moan about #bsdcafe about this

and people would run it 2 and have a broken system

its not 4 times lol

its the same process of building from src

kernel first, reboot, then userspace

and a 4th one just to be sure

cause why not

lol

what you percieve as a flaw, I see as a benefit

albeit, freebsd-update is slow as a shell script, but freebsd-rustdate is much faster from what I heard

if you say so

but I will oppose turning freebsd into Linux

which is what pkgbase is

i have been updating my systems from src since I knew what freebsd was

disagree

the dependency hell of Linux

you have dependency hell anyway

yeah

metapackages for all the regions of base

do you want to manage it in big monoliths or in fine gradations that reduce network traffic

i like the idea of pkgbase making smaller jails

MelMalik: sure, but I rather have a complete base system which is one blob, than lots of small packages with its own dependencies and dependants which make up base

instead of hacking src.conf compiling and then still have a bunch fo crap a jail doesnt need

the one blob is an illusion

cpet: sorry to break it to you, but the base metapackage will still pull in all the same stuff :)

uh well I think freebsd did a good thing making pkgbase a thing

you will still be stripping things out manually

cpet: its horrific...

not if you use -jail

pkg will update the kernel and the utilities at the same time

this is one of the many reasons I dont like Linux

the thing which made freebsd (and other bsds) so great to me, is now being removed :(

i dont like linux cause of systemd pkg hell has been a thing for a long time

uh well youll get over it man

people come people go

at the end of the day no one cares what you use

the difference between ports and base is that ports arent critical for your system

if a port breaks, you still have a functioning compiler, userspace and kernel to repair it

correction: not normally critical to be able to boot

they are thinking about making pkgbase and ports seperate

MelMalik: alright sure, theres some which are

but they are rare

cpet: "thinking"

afaik it will be a opt-in configuration value

I have considered that it may be meritorious if I attempt to make pkg zfs-aware; and if an update comes down the wire on the base repo, a new bootenv can be proposed to be created and the update be done there instead of in the current bootenv

no points for guessing how i was inspired to think that up.

in theory it should be common to create a boot env before you do any updates

this is why I have like 60 of them

but since I have so much space I dont worry about them

I see why OpenBSD folks get so pissed off at freebsd sometimes

each OS has its pros and cons

i still have a linux install with virtual box to run windows case bhyve doesnt like TPM

"most gnu+linux-like" goes for the good and the bad with this

i think someones told me how dumb it is to run riscOS

but i like it

you are always going to have these issues when projects do large changes

not realizing all the work that ivy did to make it happen

now saying that building from src will probably never not be a thing

i saw this same issue when they deprecated mergemaster in favor of etcupdate

i have a dir in /etc/ named perconfs and each servcie I use has either -f path/to/conf or the equi so updating etc is never an issue for me

It is a delicate issue. Continuity is one of the reasons I choose BSD. At the same time, innovation requires change. So I don't know.

bhyve <3

you can still use the "traditional" tarball sets instead of pkgbase

o0x1eef: i do wish thay they would go back to the old way of including drm and then using fwget to install firmware

vishwin: up until 16.0

vishwin: also they have been tlaking about this since 2016 its nothing new

yes i've been more than aware

im sure if we ever get a GUI installer people will complain about that as well

i'll be quite surprised if the jimmying I just did actually works out...

and it did not.

you need to pass -c to pkg(8) when you have multiple boot environments mounted at once

why do I get the feeling if I logged into your system I could fix it in the time it takes to make a coffee

but sadly this isint 2003 and people dont do that any more

cpet Eh. That misses the point entirely. Linux is a good example. There is no continuity. It reinvents itself every few months. It is hard to keep up. Change is constant. That's the other extreme. So people are right to be concerned.

cpet because you have no idea about my setup

o0x1eef: most enterprise linux distro dont change much

MelMalik: and you make it hard to help you help me by posting things

and I think I'm going to leave filters on, then.

Again, brushes over the point entirely

o0x1eef: i remeber when I weanted to change my server from OpenBSD to Debian, the fatc of how systemd does the logging made me not continue

MelMalik: i personally dont care if you ingore me or not really me being here is a past time for being bored

o0x1eef: ok

mer's response heh

think theres one for openbsd as well but syspatch and sysupgrade are good

hasnt been updated since 2021 but doesnt stop you from creating something similar

Hello again what did I miss

I might try getting the driver crash logs tonight or tmrw from the laptop

you missed a party

JetpackJackson: MelMalik: is upset can he cant get his gfx to work, polarian is upset cause of pkgbase, o0x1eef8 is upset that i mentioned pkgbase is a good thing

JetpackJackson: pretty much sums it up

Ah

Alright thanks

JetpackJackson: any time

i am not a he.

ok

MelMalik: makes sense now really

cpet That's not accurate :) I'm not upset for any reason, and especially not because you mentioned pkgbase as a good thing.

cant help it if youre in denial now

JetpackJackson: asked and i delivered

You delivered, at best, half truths

possibly

At risk of restarting the argument, pkgbase is cool! I'm using the minimal set tho and I can't figure out which set has sysrc lol

JetpackJackson: optional

Ah thanks

Although there isn't any issue with me using a text editor to edit rc.conf, right?

nope

thats how we did before sysrc and service

That's what I usually do, and sysrc for scripts.

we would edit rc.conf and then to a sh /etc/rc.d/bleh start

If you use an editor that does not always write newlines at the end then just make sure there is an ending newline in the file.

o0x1eef: did you get upset when they added in sysrc and service ?

knowing change is bad

use systemctl enable/disable

you can write your own script, your own service

:(

I think you mean service $foo enable/disable.

i sure hope so

scott go back to red hat.

many linux distros use systemctl not just redhat

cpet When did I say change is bad?

o0x1eef: about 10 minutes ago

Re-read what I said, then think about it a bit, and then maybe reply. Good advice for life.

rwp: well that works too but imo they don't use that in a lot of the distros these days.

o0x1eef: re-read what ?

i dont take what is said in here seriously

The message where I supposedly said change is bad, but I never said that, and I think your ability to comprehend other people is handicapped by your knee jerk reaction to reply without deliberating a bit first.

o0x1eef: I am a 90% disabled veteran

so thank you for calling me handicapped

:D

ooo weechat looks different...

something is clearly wrong here

dunno why this is happening

way too many colors for me

way too many channels as well

thats just libera

I am in many more when I have all my networks open

anyways the channels are the problem

it shouldnt be like this, yet it is

I update to 15.0-RELEASE, and updated all the ports

ah fuck it let me recompile weechat

normally this is a term issue not really a client issue

but then again im handicaped so i dont know wtf im talking about

hmmm

maybe I should recompile st then

could try to install more fonts

i see st uses Liberation Mono which is patched on freebsd

to use Inconsolata

Hello! I have a router appliance that I have run FreeBSD on for quite some time. I have in-place upgraded this before. It is currently running 13.1-RELEASE-p7, but I am unable to do the typical things due to the information it is trying to retrieve from update*.freebsd.org servers no longer being available. What are my options to support this device?

Upgrade.

compile manually ?

despite st saying a simple term that sucks less maybe it sucks more if it cant properly display weechat ?

i can try manually compiling. however, are there any mirrors that would have this metadata available? Fetching metadata signature for 13.1-RELEASE from update2.freebsd.org... failed.

13.1 is EOL

so no

Not officially, no. If someone else happens to have a mirror of the old 13.1-RELEASE stuff, and if you trust them/it, you could use that.

youre only option is a freebsd-update upgrade to 13.5

To upgrade to any version it needs the 13.1 metadata

seeing how its a router i dont see the need to upgrade to 14 or 15

alright recompiling st fixed the problem, thanks for the heads up cpet :)

I kinda knew I needed to but ngl, couldnt be bothered to find my st config :P

polarian: guess in not always a handicapped veteran

Yknow im glad I reinstalled cause that showed me that the WiFi issue wasn't my fault lol

cpet: well me not wanting to use an EoL operating system is not valid?

twsta: 13.1 is EOL upgrading should require the metadata for the release being installed

gotta recompile graphics/png with the patch added manually to fix cve

then 15.0-RELEASE update is finished, after like 9 hours now :P

i am sure without the compile it wouldnt have taken 9 hours

I had iwn panic the first time I booted into 15.0-RELEASE xD

cpet: I am running `freebsd-update upgrade -r 14.3-RELEASE` and it is trying to get the 13.1-RELEASE metadata first

cpet: you are right :)

it would have taken maybe 1 hour

twsta: well src compile it is then

twsta: and learn to update sooner rathger than 5 years after the OS is EOL

ok 3

13.1 (May 16, 2022)

Lmao you just told me that you don't see a reason for me to update and then you say that >.>

i said for you to upgrade to 13.5

But yeah i am reading the documentation on how to src compile. Seems simple enough

13.5 is supported

geez

Yes, but the same problem presents itself regardless of what I set the version to

2022...

Release 13.5 (March 11, 2025)

I would hate to think how many vulns that has

also 2022 --> 2025 is 3 years, not 5 >:)

polarian: honestly I can say that network is probably compromised

polarian: as you mentioned earlier :)

lol

so you dont think im paranoid? :P

They really don't serve the metadata after EoL? it can't be that much information to host

oh i still think youre paranoid

polarian: but in all reality it doesn't really matter what I think

19:57 < cpet> twsta: and learn to update sooner rathger than 5 years after the OS is EOL

19:57 < cpet> ok 3

Damn, how do I not even have git on this thing >.<

it basically just does pf and dhcpd so i guess i never needed it

git clone github.com/freebsd/freebsd-src /usr/src && cd /usr/src && make buildworld && make kernel && make installworld && etcupdate

twsta: I have to ask, why is a clean install not an option here?

TommyC: Well, it is routing the packets that are allowing me to talk to you :P

twsta: Sure, but once it's reinstalled you can talk to us again then.

I also think my dumbass set a firmware password a long time ago and i forgot it

i tried shorting the pins but i couldn't get a USB to boot

I probably would have already installed NixOS otherwise

I do love FreeBSD though, it's been rock solid for so long

I got a bit lazy with my homelab for awhile. Sue me!

I sold my rack and all my extra shit. Now I have a Framework Desktop, a Framework 16, and my DAS. Headscale glues it all together and it is very nice.

the only issue I have with those framwork PC's is the CPU is soldered on with no way to upgrade the CPU itself

but then again its no different than my lenovo mini PC

but atleast thats a Core 7 Ultra

yeah I think that's my sign to sign out of the BSd stuff and do VoidLinux package jenga

bye

I wonder how people who run freebsd on servers are doing the upgrade to 15 and/or pkgbase

freebsd-update

I mean like are they waiting to update? Are they just zfs snapshotting? But that wouldn't help uptime

If something goes wrong

I think everyone applies a different mindset to that problem. Some will wait, some will install right away, some will do the upgrade through rotation of servers, some might not bother upgrade, etc.

I'm being lazy and not digging through the source yet, but does smapi(4) present itself through sysctls somehow? I've not seen anything documented yet.

Or is it not built by default? I didn't consider that.

Speaking strictly for myself, I don't plan to update my single machine (and daily driver) until 13.5 reaches EOL, and even then it will be probably to whichever 14.x is the latest then.

mason, "/usr/src/share/man/man4/man4.i386/smapi.4" means (IIRC) that it exists only on the i386 (32-bit) architecture.

!

Hrm.

There's an out of tree SMAPI driver that sort of works, but it has issues: github.com/sghctoma/thinkpad-smapi

wow, `pkg clean` is a thing and can recover soooo much space. I wonder if that should be a periodic thing

There's also (in case you didn't know) "pkg autoremove", which lists and offers to remove all no-longer-required packages auto-installed as dependencies.

Interesting. There are no periodic(8) scripts that call pkg.

Not in base but there's 2 in /usr/local/etc/periodic, installed as part of pkg itself.

Aha, interesting

Thanks

o0x1eef, V_PauAmma_V: ah OK, thanks

I'm currently on 14.3, and am looking to switch over to 15.0-RELEASE. I know there is pkgbase available now, but I am guessing that the pathway to upgrade is the uusual freebsd-update method and once I'm on 15.0-RELEASE I should use pkgbase. Or can I use pkgbase to upgrade from 14.3-RELEASE to 15.0-RELEASE?

mns: The Foundation ships a conversion util.

mns: github.com/FreeBSDFoundation/pkgbasify

flua :)

Without saying anything bad about pkgbase I think I will miss freebsd-update though. It's not gone yet and I am already nostalgic for it.

root@server:/etc/periodic # cd /usr/local/etc/periodic/daily/

/usr/local/etc/periodic/daily/411.pkg-backup /usr/local/etc/periodic/daily/490.status-pkg-changes

anyone with 15R and pkgbase see a lot of base orphans with "pkg orphans"?

seeing how those pkg's aren't really part of ports wouldnt be surprised

all the pkgbase stuff lives in /usr/src/release/

but im sure markmcb will mess with it then come back in here with "my system doesn't boot up any more"

cpet: it's new though. i hadn't seen this behavior in any of the pre-release versions

what does pkg autoremove say without actually doing a pkg autoremove

again without actually doing a pkg autoremove

just says nothing to do

all the orphans look like: FreeBSD-zoneinfo-15.0 ? orphaned: base/FreeBSD-zoneinfo

pkgbase was named from pkgbase to pkg

could be why you see those

rwp: I thought about using pkgbase on the fresh 15 install on a ThinkPad here, but seeing it still marked as "experimental" made me shy away. There'll be time to convert at a later date. (Thank you all for bleeding on it so I don't have to.)

I guess I'll keep an eye on it a few days and if it doesn't resolve I'll report a bug

markmcb: you can use sqlite to remove it from the db

its basically saying i have a package thats installed but I cant find it any where

mason: most .0 releases will tend to have a lot of issues which is why most people avoid them

Yep.

this my desktop is not production and my server is just a personal server well breakage is fine

plus the chances of me being able to fix those breakages is high so

Fair.

I have multiple FreeBSD machines and will upgrade some of them to the 15.0 but I have always held off upgrading production systems until the .1 point release appears.

i did remeber when I ran current there was a change to the twe driver that you had to do some conversions I frogot to do it and well my raid went to hell

but thats what you get when you run bleeding edge

which is why im not a fan of those linux rolling distros cachyos tends to break more often than most other distros

but since I only installed it to run some VM's reinstalling isnt an issue

I find it interesting how we always use the term "bleeding edge" now and the term "cutting edge" I have not heard anyone say for literally years.

I ran -CURRENT until recently and reported a couple of bugs, but otherwise it was very stable, just annoying to constantly update, that's the main reason I jumped onto 15-STABLE.

people are becoming vampires

:(

scan: resilver in progress since Fri Nov 28 12:11:07 2025

5.18T resilvered, 95.01% done, 03:48:57 to go

data

Macer, You are performing a en.wikipedia.org/wiki/Long-term_experiment

somebody needs to do it

although. there is something to be said about how awesome that pitch experiment is :)

I wonder if -STABLE will become a lot more popular with pkgbase

-STABLE has always been popular as people dont really understand it's purpose

ah the good ale days

I ran -CURRENT on an SMP box from 5-CURRENT until around 6.3-REL

no other OS has ever been as unstable before or since, excluding various microsoft products

yeap I remember the 5.x days

it was fun. back then I still played Quake sometimes, which amusingly ran significantly faster on freebsd linuxulator than native linux

doom 3 did as well

mason: thanks for the pkgbasify link

mns: note that first pkgbasifying and then updating didn't work for me... use freebsd-update the last time, and then run pkgbasify

else you need to fiddle in the freebsd 15 keys yourself

i've upgraded some jails and an ec2 using pkgbasify and then updating.

At this point is it basically wise to start with pkgbase on a new install?

that works fine, i tried it recently

isley: how did you get the keys then? they are in FreeBSD-pkg-bootstrap, but i think the pkg binary of that doesn't run on 14.3

there's a mailing list post about fetching the keys that changed and putting them in place.

and also fixing the pkg FreeBSD.conf

OMG ... the resilver finally finished 5 days later lol

it's trying to continue the raidz(2) expansion now

that sounds exciting, like edge of your seat stuff.

13.1T / 69.2T copied at 29.8M/s, 18.90% done, 22 days 20:12:19 to go

ask me again in 22 days :/ lol... i'm sure that estimate is a 'just started' one though

Macer: One ETERNITY Later

ouch, my lil 15.0-RC2 VM does not want to upgrade to RELEASE

zip: you using "freebsd-update", pkgbase, or source build?

pkgbase

I'm not against obliterating it and doing a reinstall, it's doing nothing except being FreeBSD 15

I'm just wondering if this was a known issue or if upgrades will be painful post RELEASE

not sure. both my vms are source build for 16/CURRENT right now..

likely to keep them that way too. I blew up several installs using pkgbase prior lol

Is there a way to query what remote package provides a command or file? I know you can do it for a local package but I can't figure out how for remote. Like "pacman -F /usr/bin/foo" when foo doesn't exist on your system but you want to find what package installs it

zip: that one is documented, see freebsd.org/releases/15.0R/relnotes/#upgrade-rc

JetpackJackson: see pkg-provides.osorio.me

Thanks!

Ahh

My Google fu was lacking

Thank you

it should be called locate not provides :)

so did pkgbase stick to using the pkg command and having to specify package names (ie: FreeBSD-*), or did they make it pkgbase for base, and pkg for third party?

Demosthenex: My understanding is that it will use pkg with package names. Based on what I reading here: github.com/FreeBSDFoundation/pkgbasify

meh

is there a way to get the device name of gpt labeled partitions in zfs?

never mind...

gpt/F01-03_Seagate_BarraCuda_ZCT.. REMOVED 0 0 0

another smr drive bites the dust

there are a few days. gpart show -l will show you all your partitions and their labels

ah ok. i didn't notice that it puts the dev name on the top

wow so the drive just completely fell off the map

like it was ejected off the controller

this one was a read error ejection though. usually they fail like they mean it

ah well.. i should have 2 8TB drives delivered today so i guess i'll wait on those and resilver again. at least this time it won't be writing to a smr drive.

i can't wait to see what the difference is because 5 days was torture.

Hi all! I am trying to run `make buildworld` but getting this error: dpaste.com/4PBCPRAEE

I am on this git commit: e19230d49ffc02944d93645800da934b7aeee078

This is a new install. Am I doing something wrong?

if its a new install why build the world?

rtprio: Doing some kernel stuff, I like to build everything at the same time. I guess it isn't technically required...

leah2: provides makes me think of the occasional usefulness of pkg info -r <pkg> and pkg info -d <pkg> for required-by and dependencies

i just recently upgrade to freebsd 15.0, now nginx wont start

Performing sanity check on nginx configuration:

ld-elf.so.1: Shared object "libssl.so.30" not found, required by "nginx"

Starting nginx.

i did rebuild all packages

argh! forget it, i build nginx manually becouse of a plugin, forgot that. sorry!

sopparus: might be able to have the module added to the port

its there, its webdav extensions, its just not enabled by default

ahh

so it seems that smr drives still fall off the face of the earth with zfs

Are these the ones where the tracks overlap partially?

yes

in my case 8TB seagate barracudas

while trying to raidz expand i've had 2 get wrecked.

and it's only 25% complete

so i'm sure i'll lose more in the process and there is no way tot stop it now

one i managed to swap out with another smr drive and the resilver finished but then another one died during expansion

What's smr mean? Never heard that before

Shingled Magnetic Recording.

Oh neat

Does it not play nice with zfs

what does SMR play nice with?

It doesn't play nice with writing non-sequentially to the disk. So most filesystems not specifically designed for it don't work well, if at all.

Ohh

That reminds me I should adjust my zfs setup on my PC, I have a 1TB SSD and a 1TB hdd and I set it so that they're one big drive but I think that might be problematic

Cause speed differences

Is my conjecture accurate or am I just making stuff up

that is a depends situation

do you "need" 2 terabytes of space

spinning anything will be slower and if you can swing it.. 1tb ssd has "main" drive and could use the 1tb SMR as "backup"

It's for PC games, the SSD is that emmc stuff so it's like attached to the PC

Is it possible to remove it from the pool?

Like

Idk what the best course of action is

hehe.. that is a multi-variable equation

I can hear it spinning up when I boot and do stuff

there is an ability to add and remove from pool but it is dependent on what you are working with

Like the good clicking noises and such

that is through the zpool utility

Uh it's cachyos

oh yeah.. i have spinning drives myself, in my server, and i use ssd as my main jails/bhyve/stuff area.. i have them on 2 seperate pools

Ah ok

when yhou do zpool status.. what do you see?

The only thing I'm using this thing for is steam games

that would be a good indicator of where you can go from there

Uh last time I looked it was healthy

thats great, the key thing we want to see.. is what tyhpe of zfs you setup..

I can check when I get home, is it OK to ping you later?

i may be around, if i am not.. i am sure someone can help.. if available

Alright

that link i sent you, is a great start.. in particular you want to do the command in 22.2.3 of the link i sent

zpools tatus

Sounds like a plan

Oh ok

Sure I'll do that

the handbook is awesome and trial and error.. you won't know until you try and the fact you are using this as a gaming device... the big thing is to be able to backup data

not too challenging.. it becomes more difficult the more space you have.. like i have 3tb on my setup.. so finding a way to offload becomes more difficult

Yeah I'm using my borg backup script to backup stuff like local games and mods

Most of it is steam though so that's in the cloud

cool cool.. well when you try later.. i may be around.. just depends on time and maybe can help you then

good luck

hodapp: yeah it most certainly does not play well with anything. lol

I don’t even know if there is a way for zfs to even address it. I just had 2 drives get removed back to back. I resilvered once just to have another fall off.

I have a mod_perl problem on 14.3 and also 15.0, I get something like this on both: ld-elf.so.1: /usr/local/lib/perl5/site_perl/mach/5.42/auto/APR/Request/Apache2/Apache2.so: Undefined symbol "apreq_handle_apache2"

I normally disable the mouse in the console when X is not started up with this in rc.conf, moused_ums0_enable="NO", but after upgrading to 15.0, the mouse pointer is back. Anyone who knows why and how to disable it again?

Hmm, looks like the driver has changed.

Macer, When you finally get a non-SMR 8TB disk in what I would do is I would do a bit clone of the SMR disk reading only over to the CMR disk writing only. That will make a full bit clone of it. Then put the CMR clone into the array and it will all be fast and then run a scrub.

Reading from the SMR is mostly normal speed. Writing to the CMR disk is normal speed. And then everything will be fast for the array. It's just the non-sequential writes to the SMR that kills use of it for RAID. It's not just ZFS either. SMR has been a huge problem for Linux MDADM RAID too.

voy4g3r2: termbin.com/9tn7

rwp: ah

i'm not even sure if the disk that is dead will even work now. last time i pulled one and moved it to a different bay it would just spit out cam issues

so maybe take it out and dd the whole thing in somethin gelse?

i mean a resilver to a cmr drive doesn't take that long at all imo. less than a day for sure. the smr took 5 days lol

and the dd relies on the disk actually .... working.. i'm not sure if it pushed it over the edge where it actually broke something mechanically because of the random writes... they're just garbage. there really needs to be a huge call to boycott them

i think in 2.5 admins they stated that a 30TB of CMR density translates to like 34TB of SMR space... i mean is that seriously worth it considering how horrible they are? :)

I figured out my pkg orphan problem. The default /etc/pkg/FreeBSD.conf has FreeBSD-base disabled by default, even if you do a pkgbase install. Seems an odd choice.

Unless I somehow triggered an edge case. Anyone else see this when fresh installing 15.0R and pkg-base?

yeap pkgbase is disabled by default

Macer, For disk cloning I recommend GNU ddrescue which will try and retry a failing disk cloning it to working storage. Always use the logfile feature. Always. Using the disk may use up the disk causing it to fail further just the same as any use of it. But it's been most successful at getting data off of a failing disk and onto a clone.

And with ZFS I had enough of a failing disk cloned that it was enough to rescue the ZFS pool. After a scrub it showed full recovery and no lost data.

I pulled the disk and did the ddrescue copy on a different system and then installed the cloned drive back into the zfs array. Ran a scrub. That minimized the amount of wear on the failing drive so that all of the use of it was copying the data off of it onto a different drive.

cpet, that makes sense for existing users as things changed a bit, but doesn't that seem like a bug? i.e., if you install and choose pkgbase, wouldn't you expect that repo to be enabled by default?

I think I see the issue: cgit.freebsd.org/src/commit/?h=rele…13fb0788f65002a40f47e02c08e23047525

The override was automatically created

Macer, My reasoning for your current case is that while resilvering onto a replacement CMR the other half of the mirror is still SMR and it's still slow for all of the normal activity that is happening there. In my case it would be because I have root on zfs. If this is purely a data array then it's less of a problem. Just two resilvers.

When replacing a drive it's good to keep the old drive into the system so that it is available for reference while silvering the new drive in case there is an additional new failure.

like muscle memory, I copied /etc/pkg/FreeBSD.conf to /usr/local, which overrides this change

I guess this is just pkgbase limbo for now

markmcb: I believe this is due to allow users to decide if they want a standard "pkg update && pkg upgrade" to include base or not. Until there is a way to discern when a user requires just port packages or base packages installed, this seems like the safest route.

Of course, there is also the "-r" arg to determine what repos to check and use for updates/upgrades. I wrote a little update script that I keep tweaking as options and ideas change. But, so far, so good.

ek, it makes sense now. effectively there are 2 /etc/pkg/FreeBSD.conf files that ought to exist, but only one can so this is the workaround

ek: not exactly: the pkgbase repo is disabled by default so it doesn't break non-pkgbase users, the installer creates this override so pkg upgrade works as expected for pkgbase users

it's less about user choice (you don't really have a choice here) but rather about doing the right thing for both cases

markmcb: Well, /usr/local/etc/pkg/repos/*.conf will override /etc/pkg/*.conf. So, you can certainly copy /etc/pkg/*.conf to /usr/local/etc/pkg/repos/*.conf and make changes there. This will avoid conflicts moving forward pkgbase updates in the future.

ek, understood. when I install a system, i typically do exactly that and change quarterly to latest and i'm done. what I didnt realize was by copying that file I was disabling base. the result was all base packages showing as orphaned.

ivy: Ah, yes. Much better explained. That's basically what I meant. I don't use ports packages on my productions servers so this seemed to make sense to me.

markmcb: Yep. Fair enough. Just remember to change "enabled: no" to "yes" and you should be good. I also use latest instead of quarterly, I so I completely understand.

I do also make sure to check "pkg repos" after any repo edits just to be sure. That helps.

also i recommend *not* copying /etc/pkg to /usr/local/etc/pkg/repos, that will mean you don't get updates to the base files, like URLs or signing keys. just create a new file in /usr/local with { enabled: yes } or whatever you need

ivy: ah right, really good point. thanks

ivy: So, the latest /etc/pkg/*.conf files won't contain a reliable config for future repos?

ek: they will, but if you copied the entire file to /usr/local, it will use the URL you put in /usr/local, not the updated URL in /etc/pkg

I didn't have /usr/local/etc/pkg/repos/ when I last tested/upgrade to 15.0-R, but /etc/pkg/ did. So, I figured /etc/pkg/ would contain the reliable source moving forward.

yes, it does, that's why you shouldn't just copy it to /usr/local - because then if it changes in future, the changes won't be picked up, because you're overriding them in /usr/local

ivy: Ah, gotcha'. That goes without saying. Thanks for clarifying.

Wait, wait... Is /usr/local/etc/pkg/ going to be updated moving forward with pkgbase? Is that what you're saying?

no, i'm saying the URL in /etc/pkg/FreeBSD.conf might change in future, therefore you should not override the URL in /usr/local

which is what will happen if you copy the entire file from /etc/pkg to /usr/local

Ah! Okay.

Yes. This is in respect to quarterly vs latest (or whatever you're wanting). I understand.

I thought I was thinking in reverse.

you can override the url if you want to switch between quarterly/latest, but in that case you still don't want to override the signing keys or other configuration

is there any way to enable "fastest mirror"? the mirror I get lately maxes out around 300k. i see a port, but wasn't sure if there was a more standard way.

Absolutely not. Just the choice of quarterly vs latest or base_release_* vs base_latest and whatnot.

As long as /usr/local is still the way to override without changing base, that's perfect.

markmcb: There is a "ports-mgmt/fastest_pkg" you can try to use.

I tried it while back and it did seem to work if you'd like to change your mirror URLs directly.

what should I do if a kernel module load goes so wrong the system doesn't even appear to panic (which would at least give enough time for a netdump), it just falls over?

MelMalik: Boot into single-user mode and disable the kernel module.

ek, The kernel module is already disabled, I was loading it as a trial run.

How did you originally load the kernel module?

kldload

I guess I'm asking more of a dev question - what would I do to figure out what's actually on the computer's mind at the time that it dies?

rwp: yeah. i'm leaving it in

i'm just waiting on a couple drives to show up that will hopefully be here soon so i can start the resilver

but now i'm going to have to work on replacing all the SMR drives OR just buy larger drives and replace the pool altogether.

i wonder where the sweet spot is for drives nowadays. maybe 16TB?

Macer: I quit buying drives when I couldn't get used spinners for $10/TB anymore