why does zfs hang and trash the disc so much?

the processes are stuck in tx->tx state

so i have a crash report that shows that the drm-kmod is giving me kernel panics. i was told to remove the pkg drm-kmod and install the port. is it possible to install the i915km port without installing the whole drm-kmod or do i have to install the whole port?

Anyone seen this imgur.com/a/ieYJ094

Title: Crash2 - Album on Imgur

Anyone alive I'm in single user cause I was told to install the drm-know port instead of the pig rebooted and I'm kernel panicking every time. I dropped into zroot shell since that was the defaults in the install im mounted rw but networking dosent work I only see lo plugged into a usbc ether port and did ifconfig and I don't see an interface. Any

ideas?

jb1277976: Double check loader.conf and rc.conf for typos maybe, from rescue media.

dkfkd: what kind of disks are you using? I used ZFS on an SMR drive once, an 8TB barracuda, and it made the disk thrash ridiculously.

hello all

hello

how do i fix incredibly small screen resolution in the console?

gnathat: the question is ambigious; do you mean that the resolution isn't your panels native resolution, or do you mean that the text is small? Is the system using BIOS/UEFI-CSM or UEFI? What console driver are you using?

gnathat vidcontrol

that doesn't really help, no.

@debdrup I was messing with the /boot/loader.conf

#kern.vty.fb.default_mode="800x600"

but that made the text bigger and the screen smaller

cmake, without gui flag seems to pull in a load of x11 and qt stuff

As description says, it's meta port. If you need just cmake CLI, you probably want cmake-core

i had a message about it conflicting with cmake-core

I've just uninstalled cmake seems to get rid of all the deps

and kept cmake core

thanks

is cmake without the gui flag not the same thing though technically

Kinda sad but it was a learning experience. Had something urgent come up with mybfreebsd laptop and nobody was here and since I'm new the forums don't approve my message right away so I had to reinstall. Didn't lose anything but there should be a fix or somewhere to go if this channel is all sleep

If I were to write custom init.d scripts, should I put them in /etc, /usr/local/etc, or somewhere else entirely?

/usr/local/etc, I believe.

Sorry, rc.d not init.d; and thanks.

Where is the state of rc system stored? i.e. which services are started already?

ccx: service -e

those are enabled

ccx: The state itself is in rc.conf - the man page for service(8) can give more detail.

Thanks, though I'm speifically asking about state of the services, not which are enabled. Looking at /etc/rc.subr seems to default to $rc_pid but surely not all scripts leave running processes behind.

Thinking of it, do rc.d scripts bring up their dependencies on manual start? Or is that just for ordering?

ccx: see rcorder(8)

ccx: Look at force_depend in some RC scripts and in /etc/rc.subr

Thanks again!

Hey folks. I'd like to get a clear head to look into my analysis of this issue.

Which issue?

I have a jail that runs process A which spawns process B which spawns process C. That runs perfectly for a few minutes. Then system logs show that processes C and then B receive a SIGBUS.

And presumably you then get a core file that that you can backtrace with lldb, since you built it with debugging symbols, to find out what caused the SIGBUS?

thanks debdrup. I will reconfigure to null-mount rw to get the core file

before I go into that: dpaste.org/pAK85

'supervise' runs python3.9 as a child. Any reason why the parent (supervise) should crash if a child does?

the error on stdout is indeed "Segmentation fault", which I'll investigate as soon as I get the core file. But why does the parent crash too? Might that suggests an overall memory exhaustion in the jail? Is there any such memory separation in the first place?

I found the cause of the crash of "python", and I'll fix it. I'm still confused as to why that should make the parent process crash. There is no core file for the parent process.

Without a core file and a program built with debugging symbols (as well as debugging symbols for the userland, libraries and kernel), it's very hard to say anything.

the core file showed that a library was missing (I use "micro jails" with only the required files). The missing file crashed the logging process. Still unclear to my why that caused a domino effect to the other processes. They are completely independent. I now had the library placed, it's been running for ~20 minutes without crash so far.

Hi, does anyone know how to access the hallway track of the vendor summit? I am logged in via Zoom, but I think I am beeing blind. :(

gbe: it's in the topic in #devsummit on EFnet

I can PM it if you'll permit me to :)

debdrup: a PM would be nice, since I am currently only on Libera. :)

When should you compile ports/package and when should you install a pkg ? is it just time or what?

jb1277976: I'm not sure I understand the question.

When should you compile a port or when should you install a package ? what are the reason why you would do one or the other?

Packages are built using the default configuration for ports, so in case you want to use some option that isn't default, or if you want to build it for a specific architecture and/or micro-architecture.

If you need security updates quicker than the project build them is also a good reason for using ports.

Thanks debdrup

so if I use ports it uses the default configuration like BATCH=1; make clean install clean ?

pkg*

I believe so, yes.

If you examine the ports tree make files, you'll see that internally, it's using the 'package' target as part of the 'install' target.

question: why doesn't freebsd have a good progress bar when downloaing or installing something?

gnathat: what part are you talking about, specifically?

dont know if it is specific to macports in macos, but that progress bar is great when im downloading.

macports is an entirely different system, though

Quite a sizable amount of ports have distfiles which don't report the size of the file correctly, too.

i know it is different i am just wondering why we dont have a progress bar

Because nobody implemented it? :D

If you can work out a way to do progress reports when the remote size is wrong, it'd be cool if you could provide a patch :)

another question has anyone had slow download rates in the pkg? took me 1 minute to download 20mb pkg

That's occationally an issue if the servers are being heavily used, which can happen - the package building process is being adapted to be able to use regular CDN services like Akamai, but it requires a not-inconsiderable change to make it work properly.

is there a way for me to see the status of the servers through the terminaL?

i've also had the geolocation get confused and go the wrong way across the ocean

I have a friend who is yesterday compiling all of the ports that they use.

Compiling on a 4GB RAM system. 36 hours later I am told it is 90% of the way done.

seems fairly normal

Yes. Normal. A huge advantage of the pkg binary packages are that they are already compiled and ready go to.

depends on how long your port list is of course

And if Firefox, Chromium, LibreOffice, are in the list or not. Those are some huge monsters.

I was just reacting to an earlier question about when to use pkg packaged binary ports versus source compiling ports.

firefox isn't so bad. chromium is a horror to build

last time I tried it was 9 hours using 4 cores

and that was likely more than a year ago

The project builds close to 40k packages, so if you've only got a couple hundred, you can probably finish that quite a bit faster than the beefiest servers in the FreeBSD build cluster.

4GB or RAM would have been massive in the early 90's

gnathat: w3m and pkg-status.freebsd.org

But something like chromium in the early 1990's would have been unheard of too.

my first fbsd box might have had 8 or 16 megs, I forget

maybe even less

16MB? Surely that'll be enough for anyone. :)

back in the days when you could boot off two floppies

s/Surely that'll/Should/ There. FTFY!

Were they floppy, though?

3.5", so floppy inside the hard shell :-)

boot.flp and kernel.flp

I can't remember the size of those two, but they surely couldn't fit on an actual floppy.

As in, a floppy floppy.

back in the day, they really did

ive been moving documents to plain-text and the size differences of the storage im saving is massive

or do you mean a 5.25" floppy? some of the later 5.25" formats could hold 1.2MB or so

never realized how much memory an storage eye candy modernity makes

oh yeah, it's crazy

RhodiumToad: I mean the 5.25" that are actually floppy, as opposed to just called it.

you know things have gotten well out of hand when emacs not only isn't the largest process, but is dwarfed by a bunch of things

HiFD could almost fit a modern bootonly image.

I don't think the 5.25" ever came in 1.2MB capacity, but I don't really remember.

man floppy install was the worst; esp when you only had 3-4 floppies to use

heh

I probably did the from-scratch floppy install maybe 2 or 3 times, after that it was updating from source for me

(cvsup! remember that?)

having used microsoft word for practically all of my life there is a difference of about 500kb in most documents. that accrues significantly. why is word taking up all the space

...probably not a question for #FreeBSD.

i know im just rambling now . ill see myself out

Anyone who knows why OpenJDK 17 requires fdescfs and procfs mounted?

what's the best way to go about expanding a volume for a bhyve VM that isn't a zvol? `truncate` a larger disk, dd over, repair headers, and expand as normal for linux?

(the VM is linux)

gewt: you can truncate in place to expand it

oh, cool. i don't do much with truncate so haven't learned all its power yet

just make sure the value you're truncating is larger than the disk or otherwise you will, uh truncate it smaller

yeah, makes sense

i would probably just copy the image befor i start horsing around with it

but then inside linux, i ususually use volume groups so it's not hard to fdisk to resize, expand the volume group and whatever

yeah of course

yeah, it's already a volume group

i planned ahead, apparently

What are some nice apps I should use with FreeBSD to make life simpler? All I know is psearch

Ports

jb1277976: what are you trying to do?

FreeBSD hangs at boot on my arm64 VM with "cryptoarm AES-CBC" or something. I noticed this happened after I set 1080p resolution in /boot/loader.conf. Coincidence? Modern monitors use crypto on their interfaces, right? my ViewSonic 4k monitor perhaps isn't supported yet by FreeBSD 13.1?

Or it's the host OS doing something (not FreeBSD)

winlundn, Is that on ZFS? If so you should be able to use the Boot Environments to boot a previous file system clone from before when it was working.

rwp: ok

After booting a working previous Boot Environment you can mount the current one, fix it, and then boot back to it. "bectl mount zroot/ROOT/default /mnt", fix the loader.conf there, then "bectl umount zroot/ROOT/default" for example.

<dkfkd> "why does zfs hang and trash..." <- dkfkd: maybe a problem with the disk, or related hardware.

Why is zfs kinda forced on us during the install? Just wondering

How is ZFS forced on you? You can choose UFS if you'd rather.

V_PauAmma_V: I worded it wrong. I chose ufs and my system wouldn't boot. I chose zfs defaults and system ran fine at boot