the wabt port appears to be missing a crucial library, but "pkg provides" insists that it should be there. How can I debug this mismatch? bsd.to/q0fs

Title: dpaste/q0fs (Plain Text)

Why isn't /usr/local/var/{log,run,db} more common?

I almost always see /var/... being used, which is fine, but I'm updating my skeleton (template) port and am contemplating using /usr/local/var/... instead.

With zfs boot environments I see that usually only parts of the /var tree are included, which implies that upgrading base might not carry over some of those files.

Hmm, "A handful of ports create /usr/local/var and should probably be fixed."

hey, I have two machines - one uses 10.123.109.57, the other uses 10.123.109.58 and my intent is to have a VIP of 10.123.109.60 using ucarp. I have ucarp working, all it behave as expected.

Now, my point for this is to use HAProxy. so in the config, I'm setting the bind to bind 10.123.109.60:1234 for every whatever port. and that's for the listen. that works. Now, for the outbound, I am seeing attempts at connection on the native address.

how do I get the machine to use the virtual ip for outbound connections rather than the native IP? this seems like something I should be able to do and I'm positive this is some kind of routing or similar in FreeBSD ... so that's why I'm asking here not on #haproxy

virtual IP is created as an Alias on the same nic as the native IP. so the question could be boiled down to "How do I default outbound traffic to the Alias IP instead of the native IP?"

laidback_01: thru routing, usually

docs.freebsd.org/en/books/handbook/advanced-networking/#carp nothing here mentions needing to do anything special with routing

Title: Chapter 33. Advanced Networking | FreeBSD Documentation Portal

usually the software generating the packets needs to know to bind to the other IP, its not something done at the OS layer traditionally

can jails access serial ports or raw USB devices?

i'm guessing yes by just setting up the right permissions for the /dev file no?

i'm thinking of buying a USB zigbee dongle for Homeassistant which I have running inside a jail, idk if this dongle is a device in /dev/usb or if it acts like /dev/ttyS

hjf_: you can set up /etc/devfs.rules to unhide devices inside the jail and configure jail to use that ruleset

For example: github.com/cbsd/reggae/blob/master/templates/devfs.rules

Title: reggae/devfs.rules at master · cbsd/reggae · GitHub

meka: great thanks.. that's what i was guessing. awesome

hjf_: depending how you run your jails, maybe you'd like to use runtime for devfs: github.com/freebsd/poudriere/blob/m…src/share/poudriere/common.sh#L2006

Title: poudriere/common.sh at master · freebsd/poudriere · GitHub

Portlint fun: bsd.to/sXWm

Title: dpaste/sXWm (Bash)

good morning

im trying to switch my default version of java from openjdk 8 to openjdk 19 does anyone remember the tool to do that ?

im having a brain fart this morning

bsdbandit, java/javavmwrapper maybe?

Judicious symlink use?

meena, nacelle thanks for the information. I'll see what I can do with haproxy to convince it to use the alias address instead.

laidback_01: you just tell it, explicitly, which addresses to listen on.

listen is great. works fine. it's the outbound connections that are the problem.

weird, why wouldn't it use the listen addresses to respond to a request coming in on them?

Hi! When I run "iocage list" I get this message: "Please set a mountpoint on zroot/iocage/templates"

zfs get mountpoint zroot/iocage/templates tells me its mountpoint is /zroot/iocage/templates, though

templates/ is empty though, I might have accidentally removed its contents

hrm, just remounting it fixed it

*shrug*

that moment when you fix some problem by restarting the device -_- (not FreeBSD dw)

shitty device

hmmm should've used #freebsd-social

sjk: iocage has fallen out of support and hasn't received any Bugfixes in at least a year

meena: really? what's all the rage nowadays?

as a linux user, where would be a good place to get started with FreeBSD?

I'm interested in learning my way around it

lilufo: i think the handbook on freebsd.org is pretty good

thanks, I'll give a read through

sjk: pot, or cbsd, or, if you don't have many jails, you could just manage them manually with jails.confs, or, if you're adventurous, you can go for podman

lilufo, FreeBSD's guide, I guess.

FreeBSD is pretty straight-forward, compared to OpenBSD, I think?

sjk before I dive into a physical install on my PC, is a virtual machine a good way to practice and check things out?

Well.. if you're familiar with OpenZFS, that is

You can install Arch on OpenZFS

try with FDE and you'll probably cry from pain

FreeBSD makes it ultra-easy

oh nice

it's a good starting point among the BSD's?

I guess so.. it's friendly enough to me

But compared to you - I might be too elite

idk

shit that is easy to me might not be to some

for me*

ah

I prefer OpenBSD for desktop if you can, but if you can't FreeBSD is good, too, and FreeBSD is, in my opinion, better for storage server stuff

if for nothing else - at least for dynamic partition sizes and stuff trough OpenZFS

PredatorONormies: if you call yourself elite, I'm imagining you're at most 14.

I'm 9, actually

It would be extremely dope if I were only 14, and knew about FreeBSD, OpenBSD, some of their pros and cons, etc.

Imagine that

It would work in my favor

somwhat

Is that sarcasm?

I always say - age doesn't matter - one could learn a whole life-time of one's knowledge in just 1 year if they try hard enough.. it SERIOUSLY depends on how fast you learned your knowledge - and how much interested you were in it

I remember almost nothing from school

But shit that I learn - I often-times like it, and thus - it's much more easier to remember

people who hate learning (probably a big part because of FORCED schooling) - won't know shit even if you gave them 200 years of time

There's probably many channels for that conversation, but #FreeBSD ain't it.

Agreed

neutral

you win

Sigh. All of a sudden ezjail went extinct and I had to switch to iocage. Now it's time to switch again. bah!

Looking into pot now

(The software)

And with jail.conf.d, orchestrating thousands of jails with base utilities is doable.

hmm

I have seven jails.

That's easily doable.

I do enjoy pretending I am administering a huuuuge bunch of servers though

But if you can do a whole bunch of jails with just the regular tools

maybe I could do that too :)

youtube.com/watch?v=S3u8OtjfGFE is, until Someone™ get around to redoing the handbook, the best source of information on how to do jails with tools in base.

Title: 20 Years of FreeBSD Jails (2019) - YouTube

debdrup: yeah, he wrote a book about jails too

Yea, but I can't remember if the book c overs _everything_

is VNET recommended nowadays?

ah, Lucas' got a section on vnet :)

There's no reason not to use vnet.

Highly recommend the book, especially paired with the two zfs books.

spork_css: yeah, I read the zfs ones and just started reading the jail book

I think mwl stated that in order to write the jails book, he needed to write all the other freebsd mastery books.

what's everyone's fave jail manager? doing it by hand? iocage? bastille? ezjail?

I quite like the bastille thin jails

yeah, I use bastille. used ezjail, iocage, and bastille, and I swear something else, but I can't recall. Only done it the by-hand way a couple times. Been wanting to try out CBSD, but so far, the opportunity hasn't presented itself. Bastille is the most useful to me at this point.

yeah I used bastille for a juniper pulse connect vpn that takes over the default route.. got it set up for a VNET jail.. works a treat

and doesn't muck with the routing table of the host system

Just occurred to me today, if I can't get Azerothcore or Trinitycore to compile on FreeBSD, maybe I can build out a Debian jail per a somewhat recent BSD Now.

i use jail.conf + PkgBase: alpha.pkgbase.live/howto/jails.html

Title: Howto: Setting up Jails

That sounds cool as well.

mason: WoW :D

pertho: Yar. It ought to compile on FreeBSD but honestly the errors the code generates are a bit opaque to me. I'm not a C++ fan any more and I get easily confused. :P

Title: Dennis Ritchie - Write in C - YouTube

mason: I tried to fix the devel/adb port in OpenBSD but it's a C++ mess and I got lost trying to look at what to fix.. I think FreeBSD's port is newer but I couldn't translate it over.. so gave up

I think adb changed all its namespaces.. noticed in FreeBSD it's devel/android-tools-adb

i need to move some bhyve things to jails

Surprised it's not Java.

Yeah, heavyweight -> lightweight is a win.

Java... *shiver*

I'm trying to ditch all my VMs.

Can't find that port in head ports.

Oh it's inside androidd-tools.

freebsd bhyve VMs... under.. freebsd?

mason: so is azerothcore a way to host your own WoW server?

why use big resources when little resources do trick.

pertho: Yes. Specializes in WotLK.

So, client 3.3.5a

at the time i had more linux hosts

i'm trying to set up a new external flash drive for fbsd so i start with 'dd if=/dev/zero of=/dev/da0 bs=1M' and it starts but then soon after says "dd: /dev/da0: Input/output error". what's that mean?

Maybe it's busted. But why do you zero it with dd?

just to wipe anything that's currently on it

it's a brand new drive but i guess it could be a DoA?

anything else i could try?

Why not just create a filesystem on the thing?

well i gotta run gpart create -s GPT da0 first no? it's currently MBR/ntfs

You didn't specify a limit, so it keeps trying until you get to the end of the source. /dev/zero doesn't *have* an end.

s/trying/copying/

but it shouldn't input output error

polyex: zpool create elements /dev/da0; zpool initialize elements;

Why not? It tried writing past the end of the target.

V_PauAmma_V: ok, perhaps. if he was very patient

wait so what did i do wrong?

i used that dd line in the past and it worked

Without count= ?

ya i'm pretty sure

polyex: you didn't do anything wrong; it did the trick and blasted the front part of the disk. now make a filesystem

so the input/output error can be ignored?

Hold on. Let me try on one I know works and can be recovered.

tyvm

just tried to run gpart create -s GPT da0 and it said geom da0 file exists

i think i need the dd to work first before i can run that

or you could use commands to remove the old paritions

well i just tried gpart destroy -F /dev/da0 and it said input/output error but i guess it worked? gpart show shows a raw disk now

Now *that* is worrying.

(At 7 MB/S, 32GB is gonna take a while.)

just did newfs /dev/da0p1 it said newfs: wtfs: 512 bytes at sector 454354353: input/output error

Sounds like you got a dud. Or one that's lying about its size, as I'm told happens.

hmmm i'll try another. i ordered 2

... yeah; a fried of mine got a 1tb usb stick for $30; it was more like 32gb with some sector magic

might have had an mp3 player embedded in there too ;-)

this is a samsung t7 shouldn't that be good? or does conterfeit junk get dumped onto amazon?

SSDs shouldn't get zeroed like that.

they shouldn't?

why not?

No. Short answer: They have a limited number of times they can be written to.

There are other reasons that can depend on the firmware, but that's the most important on.

but it's just once to clear it, so that's only 1 write no?

What was on the disk before, and what are you going to be using it for?

new drive

gonna use it for freebsd-ufs fs

just to put files on and stuff

Oh, so you're doing a burn-in?

well whenever i get a new flash mem stick or drive i like to zero it then put gpt and fbsd-ufs on it

this workflow worked on flash mem sticks before, this samsun t7 is my first external

Do you have any errors from CAM in the system log?

would that be in dmesg or?

It'll be in /var/log/messages

grep for CAM or?

ya lots...

cam status: ccb request completed with an error

tail -n 500 /var/log/messages | nc termbin.com 9999

tons of those

Yeah, that's probably a DOA drive.

File an RMA with the seller or the manufacturer, depending on juristriction.

is my command list termbin.com/xgmo7 right for future reference? it's to take a new flash stick or ssd, wipe it, and put generic gpt/freebsd-ufs on it

Termbin is just a pastebin that's accessible by sending the data to it via netcat.

no i know that. go to that url

I don't see anything obviously wrong with that.

ty!

I'd recommend using 2M for the blocksize because it's a factor of 4 but isn't evenly divisible with 4.

2MB up to 1TB capacity, then 4MB for >1TB capacity?

Nope, 2MB for basically anything that doesn't _require_ a certain number of bits.

number of bits = block size in this context?

Some 520B sector disks will complain if you try to write anything but 520 bytes to them, but those are rare (and can usually be formatted to use 512B with sg_format).

No, blocksize means the kernel will send requests of that size to the disk, and it's up to the disk cache+firmware to decide how it gets written. 2MB happens to be a medium value between the native sector size of the drive and whatever the size of the cache is, as they usually top out at ~256MB.

where does "number of bits" come from then?

Too small and it'll take forever to write, too big and can cause issues with the disks firmware and cache filling up too quickly.

What do you mean number of bits?

"Nope, 2MB for basically anything that doesn't _require_ a certain number of bits."

basically i wanna know how i can know when something requires a certain number of bits, to know when to use a different block size than 2MB

Oh, I explained that already, but you didn't catch it I suppose - see the thing about 520B.

ah

Do you use enterprise SAS drives meant for enclosures made by the same manufacturer? (Ie. EMC drives for EMC enclosures in the past, nowadays Dell drives for Dell enclosures)?

If not, then you don't need to worry about it.

ok so 2MB blocks on all external drives no matter the size or type (except the weird 520 byte stuff) got it ty

ahh

2MB blocksize for all drives.

k i'll try that

..unless they're special enterprise drives.

btw the 2 new samsung t7s don't show up for gpart show now

even when i change the usb port or whatever

Yeah, they're dead. :D

wtf

That's why doing a burn-in test is good (though dd isn't the best tool, you ideally want something that can read and verify what it's written).

By "not evenly divisible by 4", do you mean "not a power of 4"?

V_PauAmma_V: that's entirely possible. :3

</math major>

For a good burn-in test, creating a temporary pool directly on the drive then doing zpool-initialize is pretty much the best way I've found.

My cstream dd-like tool has reproducible random generated output. So you can first write it to disk and then `cstream -i- | diff /dev/da0 -`

Verifying that all bytes are the same.

Shouldn't that be procedurally generated instead of reproducibly random?

Err, yes. It just uses random with a constant init.

And it's ascii.

I'm good at math that makes sense to me. When I was in 3rd-grade-equivalent, I had trouble with long division because I couldn't understand what that apparently pointless procedure had to do with figuring out how many of my 37 slices of pizza each of 17 people would get, with how many left over. It lasted until someone explained not how, but *why* it worked.

</tangent>

k i just plugged both drives into a mac and they both showed up just fine. reformatted them both just fine

What was the geom commandline to display the mapping between /dev/label entries and the original /dev/ entry?

gonna try rebooting or smth i guess

cracauer: geom -t

thank you.

No worries.

I have to say that I find the Linux way of using symlinks to be very convenient for the labels.

I guess it costs some indirections through the filesystem on access, though.

the drive shows up in gpart show after a reboot. running dd now with 2M bs. it's running however there were some CAM errors dumped into /var/log/messages

polyex: if you get _any_ CAM errors on a brand new drive, it's almost never a good idea to try and use it. Just RMA it.

oh really?

Especially if you're using a filesystem without checksums.

i wonder if samsung is putting their defective stuff on amazon

UFS on FreeBSD is at least better off than other Unix-likes, because FreeBSDs lack of block devices mean that the filesystem can correlate I/O errors to filesystem operations, but even with checksums like in ZFS you're still not gonna enjoy using a defective drive.

polyex: even if they were, it's pointless to speculate on it. All production lines, no matter how good, have a failure rate - otherwise, production lines can't be profitable.

after dd finishes i'll run it again on same drive but from a motherboard usb port instead of pc case usb port. unlikely but maybe the pc case usb port is defective?

oh

Even if Samsung is one of the more reputable manufacturers with certain of their product lines, they still have errors sometimes.

well i got 3 of these drives so i'll try all of them. if they all generate CAM errors i'll be really suspicious

might even become angry

I'm not familiar with that particular line of their product, but own several Samsunug SSDs in other products.

As for external disks, what I do is use one of these: akasa.co.uk/update.php?tpl=product/…oduct.detail.tpl&model=AK-ENU3M2-05 combined with an M.2 drive.

Title: Portable M.2 SATA / NVMe SSD to USB 3.2 Gen 2 Aluminium Enclosure | Akasa Thermal Solution

we had logic in traffic server for figuring out the native block size of drives, when using it in raw mode

I wonder if that's still in

Nice enclosure.

They're fantastic.

can u use regular nvme or only sata?

That's the advantage of it, you can do both.

(: sweet

And switching between them is entirely toolless (though a little fiddly)

What kind of throughput are you getting with those?

I don't have any USB 3.N Gen2 ports, so haven't been able to test, but Anandtech tested a previous version: anandtech.com/show/16954/akasa-m2-s…ng-spare-drives-a-new-lease-of-life - though I will say that to reach the highest speeds promised, you'll probably need to cool both the SSD and the controller chip on the drive.

Title: Akasa M.2 SSD Enclosures Reviewed: Giving Spare Drives a New Lease of Life

I'd say the main advantage of them, though, is to be able to get a reputable M.2 drive from a manufacturer you're comfortable with, instead of some NoName branded flash that's almost assuredly bought because it was discarded by someone else.

I do recall seeing over 1GBps though, so that's ~fast enough for me~

Wait, over 500MBps.

Well, I was wrong about dd reporting end of of as an IO error. "dd: /dev/da0: end of device"

k the dd finished, but it says the primary gpt table is corrupt or invalid

debdrup btw if i use 2M for dd bs, should i also use 2M for gpart add -l DISK -t freebsd-ufs -a 1M da0?

or are they 2 different unrelated things i think?

Might be critical.

k rerunning the dd command on same drive but connected to motherboard usb port. tailing /var/log/messages for CAM errors

in the meantime i'm checking out that akasa enclosure. looking niiiice debdrup

Costs a pretty penny, though.

Cheapest I can find in the US is $81 on Amazon.

dd still running but NO CAM errors yet

anyone ever see CAM errors when drive is connected to PC case usb port but not motherboard usb port?

still no CAM errors

gotta be almost done

ok dd finished and not 1 CAM error when drive is plugged into mobo

case usb port must have something wonky

USB cabling has always been a weak point.

hmm but drive not showing up in gpart show wtf

ah there it is

after i gpart create -s GPT da0 it's there

This stupid Dell laptop refuses to turn on after I put in a better wifi card :-/

I assume there's some device file I can read to get raw PCM audio from mic or other audio channels. Anyone happen to know which one i'm looking for?