ivy, try also setting the mask width the same as the non alias IP on the same subnet.

CrtxReavr: that's impossible as the non-alias IP is 127.0.0.1/8. but it works fine once i rebooted with a kernel that supports IPv4 :-)

Yes, I know what the ifconfig manpage says. . . and that used to be a requirement, but it hasn't been for many years.

I think that requirement went away in the 6.x or 7.x timeframe.

okay, maybe i don't understand your suggestion. there is no non-alias IP in this subnet, it's a /32

Try an appropriate mask width or subnet mask to a 1918 address then.

the appropriate mask is /32

Maybe.

no, definitely...

jb1277976, bluetooth is not something I've ever attempted with FreeBSD before, sorry.

rwp: you know, my terminal supports text attributes such as dim, underline, reverse, blink,... but it seems it only supports one at a time. like, you start a reverse text here, then back to normal, it looks as expected. but then if you want to do a blinking text later, now the text that was bold, is now blinking

hjf, which terminal software?

CrtxReavr: i.imgur.com/CRwrnTa.jpg

Oh, *THAT* kind of terminal.

heheh

your *THAT* appeared as reverse video

You have $TERM set to something the terminal supports?

yes i found a terminfo entry for something similar

ADDS 3A+

this terminal is configured to be a 3A+

That's not one I'm familiar with.

I've dealth with some vt10x TN3270, TN5270. ..

And if I go way back, some TVI312.

hmm. how would i go about debugging something that freezes the terminal?

this is a port causing the problem

What kind of port?

*posts a picture of a parallel port*

Is it IEEE-1284 compatible?

i've never heard of this

this is a real thing?

It was a cable spec for bi-directional parallel port support. . . wasn't actually well supported until the mid-'90s.

it says usb2.0 displaced it.

You had to look at the cable for 'IEEE-1284'

Indeed.

no loyalty among technology companies

one day i need to sit down and work out Forgejo never starts properly on boot. usually it just fails, this time it started but refused to connect to the database and never recovered, works fine after restart...

sounds like fun. can i tag along

Forgejo?

CrtxReavr: fork of Gitea, think github/gitlab but lighter

it's pretty nice when it works

Good bet it's a race condition.

well i thought that but the error was really weird

Tried using /etc/rc.d/netwait ?

2024/10/18 02:32:17 cmd/web.go:194:serveInstalled() [I] PING DATABASE postgres

2024/10/18 02:32:17 routers/common/db.go:36:InitDBEngine() [E] ORM engine initialization attempt #9/10 failed. Error: dial tcp: lookup postgres-1.svc.eden.le-fay.org on [::1]:53: server misbehaving

it's not a DNS problem and it's not a Postgres problem, and after restarting it works... so maybe it's cached some failure condition

Local DNS server?

yes, unbound (from ports) on localhost

I've not used unbound. . .

but the system gets its default route via BGP so sometimes networking isn't available early in startup even if rc thinks networking is up

(this causes problems with some other things, but usually i've been able to fix/work around them)

But I've had issues with bind giving me startup errors basically saying that IPs weren't (yet) bound on NICs it was set to listen on.

i might just move it to a VM, i've been meaning to do that anyway, maybe that'll work better

Or just add a 'sleep 10' to the startup script in the right place.

that is too hacky for me :-) for Kerberos (kdc) i wrote a new rc.d script that waits for openldap to be available before it starts, might need something like that here

I would edit this section of the startup script:

#!/bin/sh

# PROVIDE: named

# REQUIRE: NETWORKING ldconfig syslogd

# BEFORE: SERVERS

# KEYWORD: shutdown

And yes, those lines are supposed to be commented - rc.subr parses the, despite them being comments.

after adding HDD disk to a existed zfs Pool (ssd disk), how to save some file on HDD (/dev/ada1), and another in ssd(/dev/ada0)?

xxy: you can't do that, if you added the disk as another vdev (a stripe) zfs will automatically distribute the data between all disks

ivy: could i create multiple pool in one Os? each disk give one

xxy: yes, you can have (more or less) unlimited number of pools if you want

you can also use the ssd as a cache for the hdd (l2arc/zil) but this does not improve performance in all workloads, in a desktop/laptop type of system it's probably not very helpful

ivy: HDD is used to store big files like videos, after executing "zpool add zroot /dev/ada1" and followed execute "zpool remove zroot /dev/ada1" it prompt "out of space",

xxy: i've never tried removing a disk from zpool so not sure what's going on there, sorry

so i'm like, way-behind on upgrading a machine i've got (12.2-RELEASE behind) is the safest upgrade method still rebuild world from source, reboot, etc?

the packages i imagien would be quite messy being that behind, i imagine

gewt: you should be able to freebsd-update to 12.4 then 13.4 then 14.1 (do not skip 13.4 as you will run into a freebsd-update bug) but if you normally upgrade from source you should be able to go straight to 14.1 that way

i *think* the 12.4 freebsd-update files are still online even though the release is EOL... although i'm not 100% sure about that

seems that way

it didn't complain about specifying that release

after you've done those upgrades, pkg will be broken, use pkg-static to update the pkg package, then do a normal pkg upgrade

ah, dang. i think the host failed in a different way so this upgrade might be moot lol

seems the disks may have died

that could perhaps cause a problem with the upgrade

i can't ctrl/c ping and top won't launch

hjf, That matches my memory of those terminals. There were often several attributes available. But only one at a time.

how do you turn off the green text attribute

You turn off the green text attribute by removing and replacing the CRT with an amber one. :-)

amber is prettier than uranium, so that sounds like a win

In my museum piece collection I am keeping my amber 13 inch CRT monitor. You never know when you might need it.

i saw an amusing news story recently that a nuclear power plant had a failure of their Honeywell control computer, so they replaced it with the backup PDP-11

wonder if HPE still sells parts for those

It's good to have spare backup minicomputers.

Minicomputers that can be easily simulated on a Raspberry Pi these days.

i'm not sure rpi is approved for controller nuclear reactors though

boot drive is toast but my VM drive is intact

so i need to buy a new ssd lol

gewt: or two new SSDs, so you can set up a root mirror and a failed disk doesn't break your computer :-)

true!

i think i am out of drive caddies in the server tho

eh, 2'll be here tomorrow and i can take a trip to the datacentre and swap 'em in

I mirror everything possible these days. It's just so much less hassle for when something fails. Just replace it and keep going.

guess i don't need to upgrade the OS after all. :P

yeah

it's possible the raid card failed...but the other virtual drive's intact so i think it's just the boot drive

If it is just the boot drive and the rest of the data is intact then you should be able to replace the boot drive, install a new system, and mount the data and be good to go.

ivy: i once read a website claiming in the nuclear apocalypse the z80 would be the dominant computing device due to its sheer amount of mass production

did an `fsck` of the vms virtual drive without issue

rebooted and the raid card finally failed the boot ssd out

The Z-80 might do acceptably well in a large radiation environment due to the large geometry transistors used in it too. Not radiation hard by any means. But as compared to this year's silicon? Yes.

kevans: ran into a weird issue where ever time i type 'kevin' on the serial console (ttyu0) the kernel panics, any idea what's up with that?

just says 'panic: you sneaky shit'

i installed FreeBSD UNIX years ago on a virtual private server (VPS) with a company that says they 'support' that, maybe starting with v11. I upgraded every version since then, and asked them to allow newer .ISOs, which they said they would then never did. Now they say people can't 'afford' newer .ISOs (they're not English speakers, the VPS is in Moldova) despite all the .ISOs are free. FreeBSD 14.1 barely fit on this VPS because I used the full DVD ISO.

next time can I use the CD or USB or mini-USB ISO and have it not take so much space?

darwin: not really cler what you're asking -- using the 'full DVD ISO' has nothing to do with how much space freebsd takes up once it's installed

so the other methods just download a lot of stuff and install the same amount of software?

yes

the 'full' image is not a bigger version of freebsd, it just includes more stuff on the distribution media

then I guess I'll be stuck not upgrading or switching VPS or installing something else on it. OpenBSD UNIX still fits on a CD.

a base installed of freebsd is something like 200MB, which would also fit on a CD if you wanted to install the OS on a CD for some reason

i just installed a new VM here and it's using 1.5GB (with a bunch of ports installed), if your VPS has less than 2GB space it really might be time to upgrade

it's taking up 8.2GB on mine, but I only installed binutils, gcc, gcc13, gettext-runtime, gmp, indexinfo, liblz4, mpc, mpfr, nano, pkg, screen, tcl86, tcltls, zstd, and compiled eggdrop. I don't think any of this even takes up 1GB nor close

so, have you tried to use 'du' to find out what's taking up the space?

/usr, var are almost 4GB+ each

a lot is /usr/src

okay, so freebsd itself definitely doesn't put 4GB of data in /var, you have something using up the space there, use du to find out what it is

and don't install /usr/src if you're on a tiny VPS

can I uninstall /usr/src ?

if you're not using pkgbase (you would know if you were) you can just rm -rf /usr/src

/var/db is almost 5GB, which I don't use

ok, so look in /var/db and see what's using the space there

/var/db/freebsd-update is 4.4GB

you can simply delete everything in there, assuming you aren't in the middle of upgrading

excellent

it only stores some files there in case you want to rollback to the previous release

great; now I got it down to 2.3GB so can keep using it years/decades... still 800GB larger than yours but I guess a few things I installed could be bigger (I also have just cleared all the compressed logs)

Today bugs.freebsd.org/280186 hit me, again. Would anyone feel up for suggesting how to either improve my patch or suggest some other path to make that part of the handbook more helpful?

|cos|: usually if people ignore your patches the best option is to mail the appropriate mailing list to attract attention -- for src, i would also suggest github, but i don't think doc is using that (yet?)

ivy: that seems like a good idea. thanks!

my jail system blew up again

hate this webchat

luke_jobless_sb: you know you don't need a webchat to be here right? you can use a normal irc client

ivy: i mean it is a pretty interface. well-made client but I just like traditional way

ivy: my bouncer is lost in his(her) jail

ah i see

Hello, if I have a server running jails, can I build ports in one jail and in another jail just install the resulting software (i.e. not having to build in the second jail)?

Yes

Pionat: you probably want poudriere for this

dh: is there any guide on how to do this?

it can build packages to a pkg(8) repository that you can serve from any http server

although it's a bit fiddly to get it running in a jail since it uses jails to do the build, i've never tried that (but i know it is possible)

basically everyone who builds from ports source uses poudriere nowadays, except the 3 people who still use portmaster

and those people have grey beards and live in strange huts in the middle of the forest

Well, simple way would be 1) build and install packages in one jail 2) run "pkg create -a" on that jail 3) copy created package files to other jail 4) install packages in the other jail

Spent two hours tinkering with podman on FreeBSD inside a jail, to find out it can't run inside a jail because it uses jails itself. Poo. :°)

Alver: i don't know anything about podman but nested jails are a thing, just not enabled by default

(if you know that and it still doesn't work, no idea)

ivy: I did not know that (apart from a vague suspicion based on a single comment on a forum) but I don't think I'm in any position to dive into that :)

podman on FBSD is already pretty experimental from what I read, so adding another level of complexity wouldn't be a smart thing to do I think

Podman already requires the host server to have the various linux-compat kernel modules loaded which makes me uneasy enough as it is

oh is that for running linux containers? i thought podman was freebsd native but maybe i'm thinking of something else

Yes, it allows to run unmodified Linux container images

Which would be a nice thing to have - but not if I have to do it straight on the host that also runs the jails with my "important" services

podman does not need to load the linux kmod unless you want to run linux containers

maybe one day I'll know how to use xargs without looking up the man page every single time

darwin: make sure you empty /var/cache/pkg/ in particular (IIRC its `pkg clean -a` but I just `rm -rf` it) this can waste a lot of space

Hmhm. So nested jails are a thing, and I'm trying to get podman to play inside one. I cannot get my jail to be able to manage a ZFS dataset, though.

I've added the mount.zfs and related options; statfs=1; devfs_ruleset=4.

Still getting permission denied when trying to zfs create something inside the jail.

zfs jail myjail path/to/dataset too in exec.created.

Alver: post your config? I think you need to have devfs expose /dev/zfs* stuff or similar too

dch: yes, that's through the devfs_ruleset=4

But I think I've got a lead (mounted status preventing jailed=on to be set, possibly)

no thats not required

this is what I use in production

and look at the end of git.sr.ht/~dch/ansible-jails/tree/main/item/README.md for how the datasets are created

IIRC you can create them ready jailed, `doas zfs create -o jailed=on -o mountpoint=/var/db/eden zroot/jailed/eden` for example

Alver: just make sure that you actually have devfs mounted inside the jail(s)

dch: yeah, that was done. It appears that bastille isn't keen on doing jailed=on on datasets

If bastille sees the jail, the dataset is mounted and I can't set jailed=on. If I set jailed=on manually, bastille forgets the jail exists.

Guess I can try with a separate dataset.

kevans: our lord and saviour Sir Mark Johnston fixed my vtnet bug bugs.freebsd.org/bugzilla/show_bug.cgi?id=282168

ivy: woot

but as usual the fix came while i'm in the middle of a ports build so i can't test it

wonder why poudriere is building qt6

when you have pourdriere build ports do you build them all?

i had this idea. if you could somehow create a graph of like "most used packages" in ports. So then you could build those packages only, that way building most packages would be fast

johnjaye: you can build small amount, or only one port with poudriere, but all the depencies will be built too, so it won't be one pakcate

*package

ah right.

i guess my idea was like. common libraries everything uses and just build those

johnjaye: why would you build "most used packages" when you could just build the packages you want?

is there no command to build the deps of a project?

linux has something like that called build-dep weirdly

ivy: to make it fastest to build a random package i suppose.

johnjaye: I don't know what are you going to do, but poudriere wants just a list of ports. If one of them was dependent on for example devel/cmake-core, then extending this list by another port requring this tool will use already built devel/cmake-core

right. i was thinking of a pkg type command from linux land that installs prereqs

so basically for package X you get all the prereqs you need to build package X

hm... are we talking about pkg(8) command or poudriere builtds ?

mzar: what he is trying to explain is a function of apt & dpkg in the debian world

build-dep is a build dependency step of package building

it is not "linux" specific as much as it is dpkg/apt

So it sounds like to me poudriere already has that depemdency functionality built in when building a package

s/depemdency/dependency

(I haven't used poudriere in years so I can't remember personally)

Tenkawa: I am not much Linux user, but it works for me. Thank you for the explanation

johnjaye: yes, with poudriere you can build and commit to the repo all the dependencies but not package itself using command poudriere testport. Does it fulfil you needs ?

kevans: am i doing phabricator wrong, i had to download the raw patch and apply it to my source tree using patch(1)?

i mean that's okay but this is an odd workflow when git is involved

i do like how quickly bhyve+nvme writes crash dumps

ivy: there's arcanist and some tooling that markj wrote around it called git-arc that might be helpful here

devel/freebsd-git-devtools

oh i know about that but i didn't know it could do this

i only used it to submit phab reviews

can't we just use github smh my tree

you're specifically looking for arc-patch/git-arc-patch

patch is broken

/data/build/src/freebsd/lf/main/sys/kern/subr_bus.c:2579:20: error: incomplete definition of type 'struct prison'

2579 | CURVNET_SET_QUIET(TD_TO_VNET(curthread));

| ~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~

(updated the PR)