hrm… can you rename a network interface on FreeBSD that has a name like re0:2, to begin with?

meena: yes

it feels like that would break rc.conf…

ifconfig_re0:2_name="eth0"

Oh, that you mean ;-)

Maybe `\:` but lets hope, it works without backslash. I preffere names like LAN and WAN

well, either way, i committed regex crimes today: github.com/igalic/cloud-init/pull/new/bsd/device_drivers

Title: Sign in to GitHub · GitHub

now it's time for sleep

or WIRE

Nighty night, meena

Time for a bit of freebsd love. I always liked the concept of the bsd family, but never thought it could be a full desktop. Boy, how wrong I was. I installed sway wm, and it really felt like a complete experience. I have decided to keep freebsd on my laptop ;)

meena: that link goes right to creating new PR; wanna spam canonical into accepting it?

aloha

I cant get headphone auto switching to work with device.hint, does anyone have some tips

I have the followed this guide freebsdfoundation.org/freebsd-project/resourcesold/audio-on-freebsd

Title: Audio on FreeBSD – Quick Guide | FreeBSD Foundation

i don't think you'd need to use hints for that

is there another way I'm not htinking of?

meena, rwp, there is no need for the jail to have privileges to mount datasets if its a jailed one, it will be jailed & that is sufficient

except for some reason now, it's ofc not working <grrrrr>

yuripv: I feel like it needs five minutes more design work, or a better way to determine what is network device is renamed from

dch, Technology will always fail you when you need it. :-/

rerooting into an MFS always blows my mind people.freebsd.org/~lidl/blog/re-root.html

Title: Using FreeBSD's re-root capability

That is pretty crazy cool stuff there.

yeah, I use it quite a bit to do things like converting a cloud VM from UFS2 -> ZFS

can i somehow "pin" a package to not be removed, even if its dependencies are?

(this is a custom package i wrote myself and never published, not even quite sure how i installed it anymore, don't want to remove the package just because the system upgraded to py39 since it was installed)

and yes, updating the port, rebuilding and reinstalling the package would be the clean solution, i know – i'm looking for the quick and dirty solution.

eh whatever, i'll redo the port when i get around to it… disk performance stats are more important than poudriere right now^^

phryk, you can always pkg lock package name.

so it doesn't get removed.

Joint: ah, good to know, too late for this package tho, it got ded :P

oh what a bummer but now you know at least.

yeah, also not the biggest loss. this was my poudriere_prometheus exporter and i currently don't even have poudriere running, so ¯\_(ツ)_/¯

also it's probably still running in memory so it might just keep working till i get around to fix it up.

yup, still running lol

how do i even create a network interface called vtnet0:2 ?

did we obsolete those?

i ran out of space on daemon's very, very, very big machine…

phryk: is it perhaps still in /var/cacje/pkg/* on whatever box you installed it into?

erm /var/cache/pkg/

meena: how big is "big" ?

dch: 1T

i filled it with packages…

RAM ? that's getting there

dch: storage

meena: pfffffffffft

dch: i deal with much much much smaller things usually

not like i need all of these snaps, would be cool if there was a way to mark one every week or so as "good"

anyway: network interfaces. do we still have stuff like re0:2, or is that a thing of the past?

meena: I tried to find these, I though it was vlan but thats a different syntax

what even is that from?

might be old linux syntax for multiple IP addresses on a single device

really don't know

but if we don't have it, and none of the other BSDs have it, it means a bit of simplification of code

github.com/canonical/cloud-init/blo…stros/parsers/ifconfig.py#L107-L123 not by much tho

Title: cloud-init/ifconfig.py at main · canonical/cloud-init · GitHub

but in other places…

ask on -net I don't find a mention of that in the man pages for ifconfig

yeah, me neither

dch: is there #freebsd-net on here? or #bsd-net over on the other network?

nothing here: wiki.freebsd.org/IRC/Channels

Title: IRC/Channels - FreeBSD Wiki

Finally got my touchpad working

not that i know python good (or at all), but "current ifconfig pops a ':' on the end of the device" seems to suggest it is just looking for e.g. "vmx0:" and not "vmx0:1"?

yuripv: aye i'll check

vmxnet3s0:1 seems to be solarish, at least when adding more than 1 address to interface does that for me on illumos

yuripv: no Solaris support in cloud-init yet, although there is support for the SmartOS cloud.

When a maintainer of a port updates it, how long does it take before it is moved to the packages

RoyalYork: I'm not sure of the answer, but I'd be interested in knowing. With that said, I'd think if you have your packages set to "latest" vs "quarterly" will have a big impact.

I can go, with the flow

Thanks guys

dch: nope its gone :P

RoyalYork: it depends on how busy builders are. rest assured that security patches get priority

RoyalYork and zykotick9, IME and based on what I heard, it depends on the port and the ports it depends on by default. For latest, not all ports work on all architectures and versions, and packages aren't provided for each architecture. In addition, some ports (*cough*Chromium*cough*Firefox*cough*) are often at or past the time limit allocated for compiling and packaging a single one, so it can take

lots ot tries before random factors let it succeed narrowly. For quarterly, unless the update fixes building issues or security vulnerabilities, it's unlikely to make it until the next calendar quarter.

V_PauAmma_V: thanks.

RoyalYork: the easiest answer is to go look at the poudriere instance and then try and figure out from context how long it took in the past.

For example if you go to pkg-status.freebsd.org/builds?type=package then you can see that main-amd64 (ie. 14-CURRENT for 64bit Intel/AMD processors) gets built on a server called beefy18 (by hovering over the little bomb icon, or checking links in the source view) - then you can go to pkg-status.freebsd.org/beefy18 and find the latest build to see how long that took, then the second

Title: Poudriere Index

latest and so on and so forth.

The entire ports tree needs to be built for it to start syncronizing to the distribution servers, and the first step takes roughly 100 hours for a full build of ~40k ports whereas the second step takes roughly 24-48 hours.

Each full build happens every 14 days, and in between that there's always an intermediate build running (ie. one is kicked off as soon as the last is finished)

So the answer is somewhere between ~10-15 and ~100-150 hours.

Poudriere may one day get distributed build support which may speed things up, but that depends on the people working on poudriere.

I'm tempted to say that quarterly will probably on average be quicker, because it only gets security fixes and build fixes - so things change less often.

yupp

However, I do think it's important to say that if you need updates quickly, you should probably be building them yourself.

With thin seeded repos (both of which are ~fairly new, as in maybe only a couple years old), you can get away with building only a very small number of ports yourself, if you absolutely need something updated.

I use poudriere to build my own packages, mostly because some of them need non-default configuration options. Also because I can build them quickly when I need an update.

Also, building your own means non-default options as dvl suggested and I was about to mention.

Is there a recommended PCIe enterprise nvme ssd card that works well with FreeBSD 13.1 ?

I was looking at the Intel P3608 but I can't find anything in the compatibility index

The compatibility index isn't really maintainable, considering the explosion in hardware OEMs and ODMs.

With with just a single ODM, Realtek, they publish such a broad swath of almost-but-not-quite-similar devices to make a computer do other things than beep, that it's basically impossible to keep track of.

With pkg, does -c imply /var/db/pkg/... in the chroot'ed environment, or does it use the host db?

I also noticed there's PKG_DBDIR that can be defined as well. Trying to figure out how best to do this.

I'm creating a jail using pkgbase and don't want to change the pkg db on the host, just within the jail. Not sure if -c (and possibly -r?) are required and/or if PKG_DBDIR needs to be defined.

_xor: would love to know the answer to that

how does freebsd differ from nixos in terms of included packages

freebsd is an operating system, nixos is a set of scripts to piece together parts of an operating system

sonotmelty: Different groups of volunteers. It might be better to check packages you know you want to make sure they exist. They probably will.

I use bsd but im just looking how different nixos is from freebsd. Sounds almost identical. Just not a bs

I could have sworn that they listed all the developper on the freebsd website not too long ago

it feels that they took it off I suppose for security reasons

sixpiece: you mean docs.freebsd.org/en/articles/contributors/? also /usr/share/misc/committers-*.dot

Title: Contributors to FreeBSD | FreeBSD Documentation Portal

yuripv thank you, it's not exactly the one I remember but it seems equally as good...

meena: Just -r is needed.

meena: termbin.com/g68by

There's the script I just whipped up to create the jail using pkgbase.

Some of that stuff is probably not needed (e.g. pkg-bootstrap), but it works.

_pkg_config is set to my copy of /usr/local/etc/pkg/... & /usr/local/share/pkg/...

sixpiece, docs and the website were split a year or so ago. Perhaps you remember something under www.freebsd.org?

perhaps thanks, it doesn't matter that seems like ample information

_xor: so I'm accidentally doing it right, alpha.pkgbase.live/howto/jails.html

Title: Howto: Setting up Jails

meena: Looks like it. Only diffs seem to be that I tend to be explicit about the repository to use and also I use "caroot zoneinfo fetch rc runtime" as the initial packages and then add anything else that's required for the service.

meena: I'm using this with OCI (podman). The current template image that I use to build+deploy services on my cluster is around 584mb compressed (zst) and 627mb uncompressed.

meena: That's built with poudriere-image. With pkgbase, the template image is 26mb compressed and 41mb uncompressed.

Quite nice and also reduces potential attack surface area.

_xor: how does podman work for you?

do you use to deploy / schedule stuff with podman, or are you just using raw podman?

debdrup: so how can I figure out if P3608 is supported on FreeBSD 13.1 ?

meena: Deploy+schedule, but right now I'm just getting buildah+skopeo+podman working.

meena: Now that I have the jail built using pkgbase, I'm going to figure out how to use buildah on it to create an OCI image, then launch it using both buildah and podman to verify it starts & stops fine.

meena: After that, I'm going to integrate it with my orchestrator (nomad) and setup my image registry to push/pull images. I already have the podman task driver for it ported and built.

Will probably use quay for the image registry, though I did update the docker-registry port in addition to creating sysutils/nomad-podman-driver

_xor, nomad seems pretty cool. is it very different from Kubernetes?

which seems… extremely over complicated.

They target the same use-cases, but with different requirements/implementation.

I've used both, and I prefer Consul+Vault+Nomad over K8S every time for myself. However, in a non-trivial sized company/team, K8S is going to be the practical choice due to available hiring pool.

Nomad tends to be more portable. It's not supported on FreeBSD, but easy enough to use with raw_exec, which just executes shell commands on the given host.

That's what I'm doing right now. I scripted a bunch of it up for FreeBSD usage and made template job specs (the file you write to deploy an app/service). Works really well so far.

The main things I use rc for on the servers are sshd and starting consul+vault+nomad. Beyond that, everything else is scheduled/deployed/monitored by nomad.

cool

Everything else includes syslog-ng, prometheus, loki, promtail, node_exporter/jail_exporter/zfs_exporter/etc, powerdns, isc-dhcpd, gitea, minio, ...and a bunch more that I can't remember lol.

The intent with podman is to remove the need to use raw_exec and just use podman with OCI images, which will be much smaller and won't require my scripts (for raw_exec).

ah, yeah

_xor: have you done any configuration management stuff before?

Yes.

I set "Xcursor.theme: redglass" in my ~/.Xresources. But I only get the redglass cursor inside X applications.

markburgess.org/promises.html yer man who invented the theory behind configuration management said a few years ago: Kubernetes is like promise theory but for workloads

Title: A Theory of Voluntary Cooperation

and I really like that way of thinking

If you're at all curious, I just took some quick screenshots of my dashboards...

Title: Imgur: The magic of the Internet

what's the difference between service and system?

That's the idea behind this type of clustering/orchestrations. You have resources configured into a cluster using a distributed concensus algorithm (e.g. Raft). Then you describe WHAT you want (e.g. "I want service A to be running 2 instances at all times") and hand it off to the cluster.

The cluster calculates HOW to fulfill it and bring the state so that it meets the requirements of your declarative job.

K8S is not my preference, but it's ok I guess. It gets the job done, but damn is it a pain in the neck sometimes.

Upgrades were a fear-inducing chore, at least it was a few years ago. They use etcd for their distributed key-value store, and it was fragile as heck.

Plus they would break enough things between upgrades to where I'd end up spending a day just fixing the workload specs to get it upgraded successfully.

I've found nomad+jails to be both simpler and more stable, but I can only speak for my own use-cases.

stage.freshports.org/sysutils/nomad what does it need php for?

Title: FreshPorts -- sysutils/nomad: Cluster manager and scheduler

It...doesn't?

i misread

People watching this port, also watch:: vault, php74, bsdstats, tmux

meena: yeah.. was about to say..

I should submit my port. That one is a bit stale (v1.3.2). I updated it internally to the latest version, v1.4.2.

_xor: yes, you should

Latest version has builtin key-values+secrets management (optional if you have Vault, but an out-of-box feature that can be used quickly).

I have a ton of ports that I should submit, just haven't gotten around to it yet due to time. FUZxxl was going to help me go through them and clean them up, but time is currently scarce.

At last count I think I had something like 80 ports or so left that I was going to submit :/

_xor: wouldn't it be cool if submitting ports was slightly less painful than bugzilla or phabricator?

Probably, but I suspect it's more of a human/process problem then a infrastructure/tooling problem.

I've configured root's PS1 to be just "\h:\w\\$ ". Is there any compelling reason to actually make it show "root@"?

dch: people.freebsd.org/~lidl/blog/re-root.html this was way cool

Title: Using FreeBSD's re-root capability

glad you liked it too

what's a good *bsd livecd distro?