meena: There's an easy way to be sure that it does: invidious.snopyta.org/watch?v=1S1fISh-pag

Title: Write in C - Invidious

looking

i don't know of a good reason why EC2 wouldn't be around with much of today's functionality intact in 20 years.

i've def. had my own code around and in use after 20 years

(despite my strong recommendation to replace it)

hello everyone

hi

EHLO

vkarlsen: builds with php81 but it doesn't work with it. remember php is an interpreted language. I made 1.36.32 work today thanks to a patch from the maintainer and submitted my patch to him, hopefully .32 will be up soon. he stopped updating because zm had issues and didn't work on bsd

hello

howdie

quiet in these dark dank and dangerous times.

Al203: i hope everything goes well for you.

vtbd0p3 ONLINE 0 0 0 block size: 4096B configured, 32768B native

that's a new one

i thought 4k was a sure thing heree

vtbd0p3 ONLINE 0 0 0 block size: 512B configured, 32768B native

reinstalling didn't seem to help

can ne1 tell me why my sectorks are 32k, plz my family is dying

rtprio: what happend to your family?

I have 5 rsync running on a fast desktop freebsd 13.1

amd64

and things were moving fast

1 disk is NFS mounted with ntfs-3g

and all of a sudden things slow to crawl

sonewconn: pcb 0xfffff801a83bc7c0 (192.168.1.4:51756 (proto 6)): Listen queue overflow: 5 already in queue awaiting acceptance (284 occurrences)

from dmesg

I also have qbittorrent running

whihc is doing fine

uploading 2mb/s

oh sudenly the rysncs qre working

does freebsd pause from time to time/

i wonder if the usb system is taxed

or somthing

chrome is locking up

plenty of swap free

lad av 0.34

i7 cpu rted over 8,000 in cpu benchmark site

sounds like your network card is taxed

no idea why chrome locked

hmm

when chrome unlocks there is soconn max or something I change with sysctl?

all my rsyncs are from local attached usb drive to another

3 usb disk total

qbittorrent uses network card yeah but never a problem

sonds like something to do with ntfs-3g id suspect

blame windows :)

kern.ipc.somaxconn this is mentioned on a forum

sysctl -a | grep doesnt show it

heh

I do not run dbus or hald

are those needed these days?

no i don't think so

sonewconn whats this

now its working ok

hmmm

maybe it needed to warm up?

lol

chrome still chuggin while laod oavg on liek 8 cpu box is 0.3

system in top 9%

is it out of inodes?

or something I remember inode starvation on linux casue funky stuff

Title: c - How do I center widgets in a RowColumn with horizontal orientation in X11/Motif? - Stack Overflow

sysctl kern.ipc.somaxconn

512

set to 2048

I use iceWM

not idea motif

no idea about motif

kern.ipc.somaxconn=2048 I can put this in /boot/loader.conf right?

not sure its helping

i belive so, yes

hello people i need please help i will try to install from cpan re2/dfa.cc:1737:12: error: cannot initialize return object of type 'bool' with an rvalue of type 'nullptr_t'

ohh sorry

this error i get after i will try install module 're::engine::RE2'

someone know about it in freebsd 13.1?

Paste the error log at bsd.to, or ask in #perl channel

Title: dpaste

ok thanks

this a error bsd.to/46jz

Title: dpaste/46jz (Plain Text)

Do you have "re2-20220601" installed?

Or, does the module bring it along?

"re2" port is "devel/re2"

this a from perl

The error message appeared in C++ code while compiling. (To answer own query, seems like the module brings it along)

I wonder why FreeBSD won't boot in under 96MiB of RAM

ASHER, Sorry I cannot resolve the issue

after i install the re2-20220601 now more error i have bsd.to/X5Mv

Title: dpaste/X5Mv (Plain Text)

this run freebsd 13.1

Reinhilde: on my zx spectrum it does not work either, and I have 80kB of RAM! :D

kolcon: you have an eight bit processor.

kolcon: too low. 80kB is peanuts. (with divide/divmmc you can get at least 512kB, with MB03+ at least 2MB, with eLeMeNt at least 4MB)

just trying to be funny, ignore me

kolcon: i'm funny, too! :-)

Everyone is a joker

except from those who are a bird.

No exception accepted. Some birds are jokers too.

(magpie, pet parrots; apparently geese are just mean)

kea are parrots, jokers, and can be *really* mean too

people still use perl?

wow

I learned all these cpan tools and I remember this one wasnt popualr but it was best

cpanm or something

dancer n monjolicious look cool

but the bovio online tranaction procesing system looksed sick

500 widgets n stuff

was bitchin

BOP

I noticed some postgres query tool used mojo

Title: New explain | explain.depesz.com

mojo is teh awesome, really

concrete_houses: people still use C

Hi, I have an issue with a jail not using the ip addresses set in jail.conf.

the jls utility lists the ip address, but when I jexec into the jail the ip is not in the ifconfig listing.

OpenBSD's package tools are entirely written in Perl.. so yep, people still use Perl. :)

Freebs 13.1 is the version I am running.

concrete_houses: whats wrong with Perl

khronos: vnet, or no vnet?

Wanting to use vnet.

Each jail ill receive its own ip address.

this way all hosts of the network can interact with each jail as if it were its own vm.

khronos: this is the setup i use: alpha.pkgbase.live/howto/jails.html#jail-config

Title: Howto: Setting up Jails

I have found it difficult to put each jail on its own bridge, but I've seen setups separating them with vxlan

Is there something new in 13.1 as the jail config in the handbook doesn't mention anything about networking not working properly.

or just vlans, not vxlan

I can see that the syslog in the jail works as it tells me that the networking part is not up.

even htough I have assigned an ip 192.168.12.3 static and on the host 192.168.12.2 static with router of 192.168.12.1

khronos: how many bridges do you have there?

None. The Freebsd documentation never told me I needed any to get things to work.

I have re0 configured for 192.168.12.2

khronos: not sure that makes sense

Title: Chapter 16. Jails | FreeBSD Documentation Portal

This doesn't mention anything about requiring special networking setup.

requiring

Basically it states that you set an ip for each jail and you hsould be good to go.

should be

If this is not the case then at some point the instructions should be updated to account for other possible requirements.

khronos: that's not a vnet setup, tho

Is this strictly for service jails?

Even still having an ip address in this given setup doesn't do anything as the ip address doesn't ping from even the host itself.

What does work is if I jexec into the tail and it fuctions like a virtual system with out networking support of any kind.

jail

It appears that in order to do what I am looking for the documentation in the handbook doesn't give me the knowledge base that I really need.

Also on the host system there isn't any log entries to tell me if the jail process has even started.

jls and ps should tell you

meena: Ok, stand corrected on my last statement.

I guess what I am saying is with out the link posted a bit ago I wouldn't have nearly enough information to do what I want by relying on the handbook.

The handbook doesn't mention anything about networking requirements to use the jail function.

Ok, after reading a little seems there are two kinds of networking.

vnete and something else.

vnet

What is the easiest for me to use?

I want each jail to have an ip address on my main subnet 192.168.12.0/24 for the home lan.

In other vm platforms this would be nown as bridge networking where there is a single subnet for everything.

khronos: first off: jails aren't vms

next: it's only with vnet that jails offer proper separation between jails (and host)

but, yeah, kinda sucks that VNET has been around since FreeBSD 10, and there's no handbook entry on how to get it working

khronos: you will probably find that vnet jail looks more like a VM in sense that you can have dhcp client, firewall, it's own arp table, ...

The way to use it is to have bridge interface acting like a switch and two epair interfaces per jail (think of it like ends of a cable, one goes to the bridge, one goes to the jail)

Ok, so to get VNET working will I have to do something in /etc/rc.conf or is this just a jails function?

khronos: that really depends on how you manage your jails

khronos: do you use epair or netgraph?

do you use a jail manager?

You can do everything through rc.conf/jail.conf or you can use jail managers (cbsd, bastille, pot, to name the few)

Not right now.

alpha.pkgbase.live/howto/jails.html#jail-config is everything fine thru jail.conf

Title: Howto: Setting up Jails

I am only using nano to edit my jail.conf manually and bsdinstall to create them.

This is my first go around with using jails in general.

you should also setup VLAN, bridges, etc on the HOST /etc/rc.conf

when I run 'top -w', I see a SWAP column but when I add all the numbers up they are much less than what the "Swap:" summary line says my swap usage is - what am I missing?

mage: slow down, let them get it working

A friend of mine likes jail.conf only solution, so here it is how he manages jails: mimar.rs/blog/host-your-own-service…ebsd-jails-part-1-prepare-jail-host

Title: Host Your Own Services With FreeBSD Jails Part 1: Prepare Jail Host | Mimar

it's 3 part tutorial

e.g., it says I have 593MB used but the column adds up to ~99MB

khronos: if it can help, this is what I have for my $HOME router/firewall/... gist.github.com/silenius/9a1ce1a1bd21bdf0e592a4372d0cf6c2

Title: gist:9a1ce1a1bd21bdf0e592a4372d0cf6c2 · GitHub

(fw1 is a VNET jail)

meena: I didn't know vnet is so poorly covered in the docs. My pick for docs improvement this year is sound and pf

I suggest reading freebsdfoundation.org/wp-content/up…s/2020/03/Jail-vnet-by-Examples.pdf

jeffpc: it's the difference vs anonymous vs private pages, iirc

See klarasystems.com/articles/exploring-swap-on-freebsd

Title: Exploring Swap on FreeBSD | Klara Inc

I'll give that a read

there are many ways to define what's in the swap column, but it surprised me that it was off by 5x

thanks

you'll also need wiki.freebsd.org/Memory

Title: Memory - FreeBSD Wiki

it's not a bad thing to have a bit of swap being used, it just means that freebsd has noticed that some of the memory used by applications hasn't been addressed for long enough that the space taken up by that memory is better used for something else

the problem is, firefox likes to run my system out of memory :)

and I mean run it out of memory - 100% swap use, processes failing to allocate more memory, and all the hillarity that ensues

mage: that could, with some edits, be dropped into the handbook

then you probably need more memory

meena: feel free :-)

I have 16GB which should be enough :)

there's no such thing as enough memory

and firefox is fine for days, and then suddenly spikes in memory usage for no apparent reason

you know what I mean

mage: I have different priorities: freebsdfoundation.org/project/freebsd-as-a-tier-i-cloud-init-platform

Title: FreeBSD as a Tier I cloud-init Platform | FreeBSD Foundation

I file a bug a while ago bugs.freebsd.org/bugzilla/show_bug.cgi?id=263436

Title: 263436 – www/firefox uses excessive amount of memory

meena: didn't know about cloud-init..

I've to admit that I'm using salstack and my own formula (github.com/silenius/jails-formula) to remotely provision my jails

Saltstack*

like last night, the system was happy (but with 30% swap use) for days, I stepped away for 2 hours and when I got back I found 65% swap usage; closing firefox brought it down to 29% where it remained after I started up firefox again

mage: it's the standard way to bootstrap cloud images into a somewhat functioning cloud server. Tho people also use it for hardware, too

just for context :)

do you have other things using memory?

right now I'm doing some transcoding, but not really: bsd.to/DkTe/raw

Title: DkTe

(that's sorted by rss)

Mem: 988M Active, 2847M Inact, 376M Laundry, 3508M Wired, 758M Buf, 172M Free

ARC: 1373M Total, 620M MFU, 245M MRU, 3839K Anon, 13M Header, 488M Other 484M Compressed, 882M Uncompressed, 1.82:1 Ratio

top alone can't be used to diagnose issues

I know

but it is a decent overview to answer "do you have other things using memory?"

grab people.freebsd.org/~debdrup/zones.pl then run it through sort -nrk2

meena: what's the added value of cloud-init vs something like terraform?

jeffpc: not necessarily

mage: terraform creates/modifies infrastructure

debdrup: I find it a good enough starting point (pointing at userspace anyway)

zones.pl is a script from.. decades ago(?) that uses vmstat -z to figure out what uses the most amount of memory based on uma(9)

anyway, got zones.pl

mage: but the actual resulting VM etc will be vanilla

jeffpc: but userspace doesn't know anything about memory allocation

mage: so cloudinit allows passing commands/requirements from the deploy phase *into* the deployed VM

mage: like ssh keys, runtime scripting, whatever

mage: and in most cases, the terraform module you use for this will likely require a cloudinit embedded in the image somehow

debdrup: sure; zones.pl output: bsd.to/ZHnC/raw

Title: ZHnC

dch: ok, that's exactly what I'm doing with Saltstack

jeffpc: i can't use that output now

mage: probably, yeah. how does salt get the fresh image to "do stuff" ?

i need it when your system is experiencing issues

rather, you

mage like, it won't even have python, already?

debdrup: fair, this can serve as a baseline

mage: btw if you have a blog, please make a short write up of how you do this. I use ansible for provisioning & updates, its cromulent but not delightful

jeffpc: better save it then

dch: I have a formula which does all of this, basically a minion is installed on the HOST and creates the jail, then a number of scripts could be run into the jail afterwards

already did

dch: check github.com/silenius/jails-formula

Title: GitHub - silenius/jails-formula: SaltStack FreeBSD jails formula

now I wait anywhere from hours to days :D

mage: so how does that initial minion install happen then? thats the bit cloudinit provides a generic mechanism to handle

thanks for the tip about zones.pl

dch: github.com/silenius/jails-formula/b…er/jails/files/scripts/jail_init.sh is executed

Title: jails-formula/jail_init.sh at master · silenius/jails-formula · GitHub

mage: I sense some confusion here. cloudinit is for h/w, not jails.

dch, mage: cloudinit.readthedocs.io/en/latest/reference/modules.html#salt-minion

Title: Module reference - cloud-init 22.4.2 documentationMenuExpandLight modeDark modeAuto light/dark mode

ah.. ok :)

cloudinit could be bent to work in jails, but usually its not the right approach.

nope. it's not

meena: I almost have hedgedoc ready to commit now.

I hope podman + buildah will become a standard <the fifteenth standard>

I want a new standard

say no to go

no kubernetes and docker clones

+

dch: can't wait for you to finish hedgedoc (and then take a look at TheLounge)

in other news, I threw a tantrum and local_unbound out, switched to knot-resolver, I couldn't be more delighted.

dch: the is confusing, my formula doesn't run _in_ jail, it runs on the host

s/the/there

dch: podman seems much less painful than Docker. and really, i don't wanna fiddle around with Puppet in jails any more. I just wanna rebuild a jail whenever it needs changing

mage: right. but when you deploy the host the first time, something has to install the minion

dch: maybe document the switch :)

dch: yes

dch: what knot-resolver? and why did local_unbound have to go?

knot-resolver.cz config in lua (whats not to like), noticeably faster than local_unbound from a user perspective

Title: Knot Resolver

unbound would just stop resolving, repeatedly. no errors in logs, no explanations, just SERVFAILs and similar. It has 2 jobs, resolve & cache.

I think when our DSL line gets flakey, or resolvers time out, it decided to go AWOL

I just wanted to setup log rotation for a service when I noticed that newsyslog seems to be installed on my systems, has config files but the service is not running. Is FreeBSD 13 using newsyslog or just syslog?

unbound is OSS, did you adapt it to your needs so it can emit things you want to log?

tct: syslog is for logging. newsyslog is for rotation.

idwer: no, its literally the result of `service local_unbound setup`

meena, I see. so how come that the newsyslog service is not running (by default)?

tct: because it's not a service. it's run by periodic

meena, thank you for that information :D

tct: look in /etc/crontab

periodic is different from cron, right?

or cron, i never know

Title: newsyslog(8)

any idea why I have a high load on my NFS server when the share is accessed through a wireguard interface?

mage, which process consumes "the most" CPU?

0 root 96 -16 - 0B 1536K swapin 3 52.2H 190.64% kernel 11 root 4 155 ki31 0B 64K CPU0 0 3304.1 181.88% idle

sorry, this one: 0 root 96 -16 - 0B 1536K swapin 0 52.2H 371.00% kernel

any idea how to get more "details" ?

I guess it could be crypto related ..

mage, I guess you could do something like top | nc termbin.com 9999 to provide a proper overview

(it just pastes the current output of top to a web server (careful!))

this is top -S

mage, how much traffic is flowing through this in that time? (eg. check systat -ifs)

mage: nfs4 or nfs3 ?

on second thought, the inability to run up the system with less than 92MB of RAM makes sense. With 95MB of RAM, I have to add a drum device to get anything done, and the system just seems slow, even over a local network.

dch: nfs4

tct: ~20MB/s

(it's a gigabit link)

mage, am I corect in assuming that you run wireguard as a kernel module?

yes

wireguard-kmod-0.0.20220615_1 on 13.1-RELEASE

mage, is this a production server?

is

yes

awww

the issue is more with nfs I think

not really production, it's used as an external backup (it's a Hetzner box)

I'm no expert and I haven't used wireguard yet. What I would have done is disabling the nfs service and observing the change in load/behavior. then bring it back up, see if the situation is the same and then disable wireguard etc.

with zrepl I get ~60MB/s through wireguard

so I think the issue is on the NFS side

I'm afraid I won't be of much help then. I only know NFS (v4) from the "setting it up and it just works" point of view :/

:)

mage, if I may ask: what is your experience with wireguard so far? I wanted to check it out but didn't so far.

wireguard-on-freebsd that is

Reinhilde: why are you trying to run on a system with 92mb of ram

rtprio: why not :)

(the answer, by the way, is that the bootloader doesn't know where to put the kernel with such little RAM)

huh

meena: sorry for dropping off yesterday, had a bunch of stuff to take care of. Here's the rc.conf and routing table for my jails host: gist.github.com/rtyler/cd29ae50e9acf235213fc6980c994f3a

Title: netstat -nr · GitHub

Reinhilde if you seriously want that small system, then you need to make some adjustments. Even for bootloader (it is using 64MB for heap).

tsoome: yeah.

did wireguard-kmod squeak into 13.1? :)

or was it in 13.0?

I assume those huge heaps are for ZFS stuff?

tsoome: for the first time in a long time, i signficantly trimmed town GENERIC and the resulting kernel was _much_ smaller

in loader? not directly, no

rtyler: can the jails and the host communicate?

hmm

rtyler: I reckon yes, because you said that gitea can talk to the Internet

for zfs support in loader can get away with less, depending on how large recordsize you must support.

but then again, with such an small system, you want to stick with ufs anyhow or maybe even something else. (zfs in kernel needs minimum 64MB for arc)

I am using UFS.

I had to mount a 48MB drum file to even survive with 95MB of RAM in the first place, and the whole system felt weirdly slow, in that way that a powerful system on hard drives does not feel.

and by survive, I mean that simple tasks like installing a package crashed with a failure to reclaim memory.

userland is probably swapping a lot.

yes.

meena: you reckon correctly

the weird slowness was like... it wasn't even responding to keypresses over ssh, and we're in the same single family (single occupant, in fact) house (albeit via a radio link)

not "not even responding," but just, slightly laggy

I guess I'm just giving y'all like, trip reports from lomembsd now, this isn't really a support question

rtyler: I feel like kp would be able to tell you what's up in a second

what's particularly frustrating is that I had this working on an old build of this machine but my backups don't seem to include the /etc directory =_=

sonewconn: pcb 0xfffff801a83bc7c0 (192.168.1.4:51756 (proto 6)): Listen queue overflow: 5 already in queue awaiting acceptance (41 occurrences)

how fix?

kern.ipc.somaxconn=8096 tried this not sure helping

concrete_houses, Pivot to a different question. Why is your service not accepting those queued connection attempts faster?

If your service accepted them faster they would not be overflowing the queue limit of 5.

But specifically AFAIK kern.ipc.somaxconn only sets the maximum allowed queue depth. Your program calls listen(2) with the "backlog" parameter for the dynamic depth.

The limit of 5 is the traditional limit for this queue though.

Also IIRC the client trying to connect upon receiving a ECONNREFUSED will likely retry. If the queue has drained sufficiently to make space then the retry will succeed.

service? its rysnc from ons usb3 drive to another

one is ntfs-3g

while qbittorrent is running 2mbit / s upload

has wireguard now been recoded from scratch after that stupid first attempt was reverted?

hmmm

wwt: its in ports, not src for one thing