Hey guys... got a fbsd VM in a boot loop ending with: bsd.to/XnGp

Title: dpaste/XnGp (Plain Text)

Any ideas on how to restore to a working state ?

Is that all you get? With verbose booting?

I get a lot, it's just that's the last part

with verbose & safe mode enabled

any ideas on how to force-check / repair the disk?

What had happened before the boot: sudden power loss, dying disk, other?

most likely - yes, I wasn't at home at the time

(most likely power loss)

You could boot from a installation image & choose "live disk" (or whatever "live X" is called). Then poke the disk from there

hm

What is the file system on the disk?

I wonder if boot partition was damaged ...

from what I remember its ufs

tried running fsck -y /dev/vtbd0p2 ... spammed a lot but after reboot - same thing

as for the boot partition... it boots, reaches some stage and reboots

no help.. ran fsck a few times.. marks the fs as clean but still in a reboot loop

The bothersome thing is that it says device not found. That looks more like a hardware problem to me. Maybe check the cables?

the first (virtual) sata drive is not occupied as is the 3rd and 4th

the 2nd is (scsi2)

the cables work fine, it's a virtual disk on a hdd - the hdd works fine :) there are other vms running off it atm

and theres a ton of stuff in lost+found

is that virtualbox

no, pve

Someone knows how to remove pulseaudio without removing its dependencies using pkg? I would like to use OSS on FreeBSD

freebsd-update now creates boot environments by default? (`CreateBootEnv yes` is commented out in its .conf on my install but it created boot envs for me anyway)

can I just `zfs destroy` a BE to get rid of it or is there more of a procedure to that?

beadm list;beadm destroy why not freebsd.org/cgi/man.cgi?beadm

Title: beadm

Title: HOWTO: FreeBSD ZFS Madness | The FreeBSD Forums

rennj: do you perhaps mean bectl

Remilia: it's much the same

grahamperrin: sure, but beadm is not available by default

grahamperrin: thanks for the review, pushed an update to my report

grahamperrin: the only difference appears to be the language they're written in

Difference between what and what?

Sorry, I should've made it clear what I was talking about; beadm is Bourne shell, bectl is C (well, it's a C library with a tiny front-end binary, so that things other than the binary can use it)

There are slight differences, the first one that comes to mind is option -T

Huh, that's neat.

One other advantage is that bectl has a more complete manual page. :3

The default creation of boot environments made fixing a mistake I made last week super easy. FreeBSD ROCKS! :-)

As I learned things boot environments was originally a recreation of the Solaris feature created for FreeBSD as beadm which is still maintained in ports.

But being such an excellent feature it was rewritten in C and moved into core as bectl.

Fortunately you can read about beadm in old docs and map to using bectl exactly.

So much original documentation writes about beadm. But now bectl is in core and upgrades with it.

And also when I want to know how some particular part of it is implemented then the shell script beadm is available for reading very easily. Super cool.

meena (and all): sorry for not reviewing sooner. Thanks for the notes

rwp: I need to do BEs with UFS…

meena, Do BEs work with UFS? I thought it needed ZFS features in order to support it.

I would imagine that with UFS one would need to boot something live like the freebsd-installer and then use it for recovery.

Title: UFS Boot Environments | The FreeBSD Forums

meena, rwp: <vermaden.wordpress.com/2021/04/02/ufs-boot-environments>

Title: UFS Boot Environments | 𝚟𝚎𝚛𝚖𝚊𝚍𝚎𝚗

From the original developer vermaden! I am definitely bookmarking that reference! Looks like a user space management of DIY snapshots on UFS.

"We've got it all on UFS"

I am going to need to work through some examples of that technique in order to grok it. Looks like a useful technique.

However it needs to be a strategy intentionally deployed explicitly. It's not set up by default, as it is with ZFS.

rwp: yeah, it looks like lots of work

main reason i want to attempt it is so i can upgrade a few VMs with PkgBase

I really wish the VM images we offer would come with a choice of ZFS vs UFS, because i would certainly always pick ZFS

I assumed that the problem was cloud hosting on affordable sized systems.

meena: soon, hopefully, maybe

kevans: can I help?

meena: you can keep an eye on reviews.freebsd.org/D34426, but I'm not sure what else

Title: Login

.oO(login?)

kevans: poudriere image has that, but i suppose we don't use poudriere for creating images

nah, release/ scripts

kevans: but even so, the default images are pretty crappy starting off point

kevans: why is sendmail enabled, but no SSH??????

(i have heard the argument, and I still think it's crap)

I am not the one you need to convince :-p

kevans: who is?

*waves hands* the community *waves hands*

Propose change -> get community acceptance -> convince re@ to accept it -> score

the missing step between #2 and #3 is, naturally, 'make the change'

or that could be part of #1

sendmail takes about half an hour to start on a vm, because of DNS/hostname crap

CURRENT on my VM boots in 13 seconds, and would probably boot a little quicker if i bootet NODEBUG…

well, I'd have to install nodebug for that…

Is there no option to fire off "sendmail" in the background & let do its thing without blocking rest of the booting|"rc"-ing?

As for enabling ssh in VM, where do I sign the petition?

parv: i'll probably start a petition next month

it'll probably be in phabricator, if phabricator hasn't been replaced with something more sensible

meena, Aye. Will ask you then🤞 unless I see that elsewhere

parv: I have a lot on my plate with the whole cloud-init thing, but it's exactly that work that is pointing to lots of deficiencies wrt cloud environments

meena: Maybe seed /etc/hosts with what sendmail wants...? Or change the boot order so it starts after BIND or whatever's gating it?

mason: some of that implies that i can modify the image before first boot, and if I could easily do that, I wouldn't be complaining about it

First, I am biased, but... Postfix by design will run without DNS and therefore will start immediately and run locally until networking becomes available.

rwp: i a allergic to mail

so that's a different issue there

morning

what's the difference between a relent/13.0.0 branch and a stable/13 branch?

former gets only fixes for security advisories/erratas, latter will become next 13.x release so it gets backports from main

yuripv, awesome so if I git checkout stable/13 and do a git log will it only show me the commits along the stable/13 branch?

that's how git works, yes :)

you don't really need to checkout it, just do a `git log stable/13` if that is what you need

yuripv, awesome, thank you now if I git bisect does it only consider commits along that branch?

yes, but as someone's (yours?! if i'm not mistaken) experience shown, it will sometimes land in some vendor commit? :)

yuripv, yup it was mine, but I think kevans pointed that out and kindly showed me the switch to prevent that

has anyone used usb disk enclosures with zfs?

yes, and usb flash drives as well. they're enough like internal spinning rust to be managed the same

geli-encrypted zfs, too

crb: twitter.com/FreeBSDHelp/status/1457531204228374534

Title: FreeBSD Help on Twitter: "@probonopd @NealsJoe @FreeBSD_RE FreeBSD Development Lifecycle Version 2. Would appreciate a review, and if you're happy with it, I'll start on a Wiki page. Will be creating ports and docs versions too t.co/gGRjWRwSYV" / Twitter

koobs: oh, cool, thank you

pleasure

arielmt: will plugging them into a different port screw things up?

reason i ask is because it's hard to find a sas disk shelf that is cheap and supports sata

sas2 shelves should be really cheap now (used ones)

not by my experience, but the order in which you plug in usb drives definitely will if you rely on /dev/da* device names. also, my experience is limited to single-disk pools.

zfs usually doesn't care if the drive device name changes. I thought it reads the info from the disks themselves to determine the structure of the array

What do you need in a "shelf?" You looking for easy hot removal isertion?

Or just a way to hook up a bunch of disks? Because you can just get an IT mode HBA and wire them up like normal drives.

you still need a case to put those drive in, and shelf (or chassis with proper drive bays) is easier

Of course, but their are plenty of cases that hold 8+ drives.

You only need the fancy hotswap bays if you want hotswap

Erhard: i just need it as a disk array without hot plugging

yuripv: that accept sata drives?

i had a 4 disk enclosure that was esata; worked alright for zfs until it lost power one day

rtprio, Was the enclosure rendered useless due to power loss?

parv: the zpool was-- i had to restore from backups.

rtprio, Ah, ok.

oh damn

is that a common thing?

CCFL_Man: if you lose more than two drives on raidz, yes

and if you just do jbod?

Obviously, need to use RAID-Z3😏

it only had 4 drives!

CCFL_Man: yeah, jbod might have handled it better.

it only happened once in 3 or so years, then i finally got a chassis with 8 drives in it