mvanbaak: just hope you don't get shock and awed :p

Speaking of wireguard running too slow, I had a fun one the other day. Turns out my home ISP was running a v4 over v6 tunnel or something that was adding extra overhead for every packet. The standard wg MTU adjustment wasn't enough, so I was getting packet fragmentation like crazy. Dropped the MTU and tunnel was back to normal speeds again.

o\

is vm-bhyve still the go-to for running bhyve guests?

yes

thanks

i have optiplex 3050 micro running 15.1 and can't get any audio from the front panel minijack

sndstat shows: Installed devices:

pcm0: <Realtek ALC255 (Front Analog Line-out)> (play) default

pcm1: <Intel Kaby Lake (HDMI/DP 8ch)> (play)

No devices installed from userspace.

it's unmuted in mixer

has anyone tried to use: github.com/FreeBSDFoundation/pkgbasify to transition their freebsd 15 from "traditional" upgrades to the pkgbase? I was thinking of doing in a jail.. and seeing if anyone had an experience good or bad using this?

voy4g3r2: I didn't do it on a jail specifically. I ran it on a regular host though. No issues (beyond one of my own making and failing to read the 15.1 upgrade notes in full and skipping the ports kmods update before rebooting haha).

For my jails, they are treated as disposable (managed using AppJail; easy to recreate). After the upgrade, I switched all of them over to pkgbase too (created a new pkgbase jail base, which AppJail can automate now, and rebuilt the jails off that).

ohhh interesting take on the appjail and base idea.. thank you ianthetechie

Sure thing voy4g3r2! If you wont to take a look, my whole setup is open source. Nothing about the AppJail setup itself (the docs are good and the maintainer is GREAT), but maybe some inspiration for how to set up the services: codeberg.org/ianthetechie/appjails

Currently it seems my rsync, on 14.4-RELEASE-p6, requires libcrypto.so.111, which I don't have... This seems to usually be a ports-related thing from reading, but I didn't compile rsync from ports

am fully upgraded and freebsd-version -kru shows 14.4 for everything, as far as I can tell

nvm, had missed a freebsd-update fetch

mewt: ldd `which rsync`

yes, did

And the result was...?

dch, ping

dch, it's not smooth sailing after all

dch, did the thing, took out the other old disk too (only booting from one now). the system is booting but then I get:

ZFS WARNING: Unable to open $dev for writing (error=1)

it lists that message both for the new drive and the now disconnected other old drive

if I have both the new drive and one of the old ones plugged in, the machine boots happily and zpool status looks good.

what does `zpool status -v` and `zpool list -v` show? output of `gpart show -l` is probably also useful

dch, paste.jvnv.net/view/36cX2

looks good to me. not sure why gpart backup|restore created the -free- (3.0K) section on the new disk tho.

difference on the new disk (ada1) between `gpart show` and `gpart show -l` tho: paste.jvnv.net/view/8eM7D

ah, that's just labels, never mind.

jbo: from a zfs side other than having the raw device name ada0p4 instead of gpt/disk0_zfs or whatever zfs seems happy

partitions look 4k aligned so this seems all good

dch, both old disks in the mirror didn't use GPT labels yet. I added the new one with a label.

that shouldn't break things, right?

no

even if the device names get renumbered for some reason zfs will figure it out

ada0p4 is the one I want to replace next but first I need to be able to boot from the new one and that's where things go wrong

it just looks nicer with gpt ones

I assume this is both rebootable at will and close at hand

yep

jbo: if you only have the new ada1 present, does it boot? and if not how far does it get?

ada1p1 should be an EFI partition, with msdosfs type

dch, that's where things are going wrong. if I only have new ada1 plugged in, it starts booting, but then eventually fails with the "ZFS WARNING: Unable to open $dev for writing (error=1)

so you get as far as the beastie boot loader?

yeah

not an EFI problem

and then it manages to load the kernel? but then fails?

I just repeated the experiment for a 4th time but this time I first detached ada0 from the zpool and now it seems happy...

which is concerning because if the drive would physically fail the machine wouldn't be able to boot from the other, good drive?!

:D

are you sure it ever booted from the other drive?

yeah

either you only have EFI partition set up on 1 drive, or your firmware isn't set up to boot from the 2nd device

I guarantee it booted fine from either of the old disks 1 hour ago before I started dicking around

ok thats good

the ZFS WARNING which prevents booting is concerning

so this looks like a hybrid boot system, bios+gpt layout

I think the error is coming from geom tasting

dch, after manually detaching the old drive from the pool everything seems to be fine...

jbo: so make sure all the partitions are labelled, and initialised, the bios boot sectors are written, and the efi partition is mountable as msdos, and has the correct bits present

now both are new disks and it boots happily from either without the ZFS write issue

ddr5 unbuffered ecc ram

sorry wrong channel

DDR5 is only quasi-ECC isn't it?

e.g. corsair.com/us/en/explorer/diy-builder/memory/is-ddr5-ecc-memory

You can buy real ECC with DDR5.

The decaf default ECC is worthless.

haha

the built-in ECC is cuz they run at such a high frequency, they'd fail without it

it is basically dishonest when manufacturers use it to sell the sticks

i ignore on-die ecc as a feature

it's on-die ECC

I never spent as long as with DDR5 to source proper ECC given that everything is marketed as ECC now

"Everything" is marketed as ECC?