lw: or hw, that is?

the door blew off

i bet lot of sw isn't built with "nails as fuses" approach

and are checked

how do you even imagine it being certified further

what about drivers license for computers?

lot of users seem to need that

but how do you enforce that?

also why isn't it engineering

yet it runs all?

i have no idea how to fix proposed issues

in that field

grand programmers guild?

wouldn't it be nice if i walk info software house and it won't collapse

but

the xz problem was only noticed after a stray lag appeared to a random guy, while in that severely regulated world, citicorp was asked by random student that looks like your building will fall over if wind blows from the corner

who knows really

on the engineering thing, it is strange how much of a separation there is between software and non-software.

even in situations where software is directly controlling expensive hardware

E.g. in the case of the Ariane 5 crash which cost almost $400 million dollars in damages... who was held "accountable" for that? Who lost their job or license? nobody.

woop woop bsd.network is coming back up again… like email I guess it will take a while to backfill those spicey hot takes

dch: o rly?

oooo yeah, there it is

status.bsd.network new status page and a funny update hopefully the recovery wasn’t too difficult

Title: Terrible Status Page

johnjaye: yeah, i read that "The system is designed to have a backup, standby system, which unfortunately, runs the exact same code." major fail indeed, double (precision :p) fail... but accountability is other topic, sometimes people get fired, sometimes not, depending, maybe you can't even build teams if you need to fire people all the time... also i didn't fully look into a5, but if somebody said that

it's cheaper to use same code for ...

... both, it's no longer sw issue

I'm using a neovim plugin that takes pretty snapshots github.com/mistricky/codesnap.nvim

Title: GitHub - mistricky/codesnap.nvim: 📸 Snapshot plugin with rich features that can make pretty code snapshots for Neovim

it has configurable font names, but I can't figure out how to format the name of the custom font I choose

is there some sort of tool or standard for how fonts are used?

fc-list is the best I've found so far for listing but I'm missing some subtlety in how fonts should be named

I think it's choosing the font but always ends up as italic

dch: fc-match ?

i got a web server that runs nginx and my node app. can i put both of those in a jail or can a jail only have 1 running daemon?

alepzi yes you can run them both in the same jail

i got a web server that runs nginx and my node app. can i put both of those in a jail or can a jail only have 1 running daemon?

just set up the normal rc.conf stuff and a jail can run any number of daemons just like a normal install?

pretty much

just set up the normal rc.conf stuff and a jail can run any number of daemons just like a normal install?

just set up the normal rc.conf stuff and a jail can run any number of daemons just like a normal install?

sorry dunno what's up with my internet

The question came through fine, all three times :P

sorry bout that

alepzi--: yes, you can setup a nginx and node app on the same jail

alepzi--: navigate to the jail and perform the actions you would perform outside of jail. thinkgs like pkg -j <jailname> install <node and other stff>

and the jail's /etc/rc.conf to enable services?

correct

ok tyvm

you are welcome

there any limitations on a jail vs a full install?

alepzi--: a jail can not be a highver version than the baseline operating system

anything else?

alepzi--: docs.freebsd.org/en/books/handbook/jails

Title: Chapter 17. Jails and Containers | FreeBSD Documentation Portal

yep, read it

there are a few areas.. thick jail vs thin jail vs vnet

so i am not sure what you mean by limitations

it comes to personal preference at this point, i am NOT a fan of thin jails

ya seems most ppl advocate thick jails

but reason is subjective and not based on the handbook.. i just read what the differences are and picked thick jails as it is more isoaltion and hard drive space is cheap

then some will say they prefer vnet setup so they can build out a network.. which is fine too

ya i hear that too, to use vnet with jails

so that's my plan, vnet thick jails

then maybe thin jails within thick jails if i ever want to nest

uhh, i would not recommend that

have distinct jails

don't recommend what exactly, nested jails?

yes

why not? jc

just make stand alone jails

that makes it more difficult to manage.. how many levels deep are you going?

host -> jail -> jail -> jail

host -> jail

have multiple top of hte level jail and then figure out how to get jail #1 to talk to jail #2

it is not to say you could not nest jails inside of jails, i just think it would add complexity to the situation that is not needed

ok ty

would anyone have tips/tricks on why a bhyve client could not communicate with a host (through a simple ping) but can download pkgs and browse web? the host can ping the bhyve... the bhyve can ping everything EXCEPT the host

voy4g3r2: host is blocking ping ?

voy4g3r2: try to listen on a random port in the host ie. nc -l 9999 and then try to connect from vm into host ie. telnet $host 9999

lower your firewall

the host can ping the client

the client can ping the host and ANY other machine

i can even ssh into the bhyve machine no problem

are zfs features useful to thick jails or only to thin jails?

oh i guess zfs clone can't be used with thick jails

alepzi, What's your definition of thick jail that it can't work with zfs clone?

i guess thick jail means it has its own base

A zfs clone is a live file system like any other. But it is sharing disk space underneath with another branch of the file system. The usual purpose of a clone is to share that space. If one removes and replaces all files then the disk space is no longer shared. But that is not the same as not working with it. Right?

zfs clones are good for when someone has a template that is all set up. Maybe change a couple of things in /etc for it. Then clone out a hundred or thousand of them and they all share the disk space and can be forked off almost instantaneously. All of the disk space is shared.

But it works best when those images are temporary, created on demand for a short run time, and then discarded afterward freeing up the space.

so clone is best for ephemeral jails, not long running ones?

Right!

If one is going to create a long running image where everything over time is going to change then a clone does not save any space. Yet zfs still needs to keep track of all of the reference pointers. Better there just to start off with a fresh empty dataset for it then.

All of those things will "work" though. You won't notice any failures or it not working. It's just a matter of disk space usage. The organization chosen will have an effect on disk space use.

i better learn more on zfs fundamentals like clone vs copy vs template...

ty

clones suck

how do i upgrade it?

why

ketas is illustrating exactly the problem with clones and *long running jails that need upgrades*.

if i disconnect it

it's fucked

no "dedup"

i use nullfs mounts

i actually have own etc's under /root

bonus: no space usage

i have jail dist over which i tar | tar some changes

it's a zfs dataset

so i can just destroy it on upgrade

after jails stop

but sometimes you need writable etcs

oh, /usr/local comes from on of the jails

one

others have it ro

is there any way to get freebsd faster in read and write on disks ? i have ssd disks in my box and its extremely slow ... i have an identical box but with debian and that is 10 times faster. 1048576000 bytes transferred in 79.344914 secs (13215415 bytes/sec) <-- -this dd test is from freebsd box ... 1048576000 bytes (1.0 GB, 1000 MiB) copied, 14.2255 s, 73.7 MB/s <-- this one is my debian box. 2 identical

servers

What bs= value do you use in dd? (default or explicit)

dd if=/dev/zero of=/tmp/file.txt count=1024 bs=1024000

did i write

"1024000" is a weird blocksize, in that it's not a power of 2. For dd, FreeBSD performs best with "bs=128K", IIRC.

use status=progress too

but the box is extremely slow no matter what i do

n30: ssd can become fragmented over time. maybe thats the case ?

also, what FS is it?

and which ssd

samsung evo 850

and its ufs filesystem

13m/a, 73m/s only?

err 870

i hope the reads are better

not on the freebsd server, its really crappy here

i have had that same question years ago. UFS write speed sloooow vs linux, but on sata hdds. I dont think i found the answer. but no problem with ZFS

i had slow zfs years ago

=p

it was fucking slow and i attributed it to zfs

Verify that the SATA interface is running at 6 Gbit/s, I use "smartctl -i /dev/ada0 | grep current:" to report it.

it was fast as crazy before but not anymore

zfs is still kinky

but how can ssd to be that slow...

rwp: i dont have any current there .

rwp: pastebin.com/Gquu6QiY

Title: [root@mail ~]# smartctl -i /dev/da0smartctl 7.4 2023-08-01 r5530 [FreeBSD 13.2 - Pastebin.com

n30, termbin.com/napb

rwp: i have raid on my box . 2 disks in raid 1

but it have been way much faster

lacks smart

how?

I always raid everything too. That example is just one disk in the pair. But I am using zfs pretty much everywhere.

it's connected to usb or what?

pkg install smartmontools

The most recent versions of packages are already installed

oh right

Just for background information "geom disk list" will list the devices, and the sector and stripe sizes. Is it possible that it is a 4K AF disk but used as a 512 byte sector one?

are we dd'ing hdd or ssd here now

There is also "diskinfo -cit /dev/adaX" there too. Here is an example from here: termbin.com/7jn0

rwp: ill test

Title: [root@mail ~]# diskinfo -cit /dev/da0/dev/da0 512 # secto - Pastebin.com

da0 ? is that usb ?

USB storage is always terrible performance. And often glitchy.

i wish usb is better

iirc usb3+ actually had some better standards?

hernan: nah not usb

its hp raid array

but where's the ssd you tested

da0 is the raid 1 with 2 pieces of ssd

oh

i have an identical box with debian on that works ten times faster

this freebsd box was fast as crazy before with both reads and writes

hence the wrong data

but why is it ffs 13mbytes/s

i have for some weeks ago removed drives from server and checked with smart tool in windows to see if disks are broken but they have 92% health left

so there is no issues with disks hw or the raid

yeah i wonder that too

might as well raid 1 two 2.0 usb flashes then

before it took a few minutes to compile kernel ... now it take hours

same speed

before?

what was there before?

yeah a few months ago

hdd's?

ketas: the ssd drives were faster then

but how

ì dont know? that is why i ask here

there was older fbsd there?

im all out of ideas

nah

same as now

so now it's slow?

Yes

maybe ssd actually crapped out

or raid card

sadly you can't test now

crapped out?

well if it worked before

do you have spare ssd to replace and see if it writes faster ?

same hw same fbsd

hernan: nah i dont not yet

i'd suggest, on upgrade, do mirror in zfs

could at least debug more

but i checked the drives with crystaldiskinfo and i have 92% health left?

or switch with the windows raid and disks and see if there is any difference

clue is probably, it was fast before

give that windows raid equipment is similar, maybe you could take out the windows disks, replace with the freebsd disks and plug on the freebsd and see if that changes anything

that will tell you if the raid equiment has degraded (it shouldnt but..)

hernan: did not test speeds in windows with them... i only verified the drive health to be sure its not "broken"

get spare ssd

take old one out and r/w test it

yeah

and keep it as backup with health left goes down to 5%

better be prepared

what was read speed?

but write... hell knows

does it get hot?

they do throttle

some

how much writes it got?

i hope that raid allows smart passthru

i wonder if raid does trim...

so much fun with hw raid

yes they have benefits