oh that's a tad annoying sort --file0-from takes the output of find -print0 but not as a pipe only as a file.

doesn't take "-" as a filename?

nope

or /dev/stdin?

just double-checked whether my friend's FreeBSD box supported that

not tried that

that's a workaround I use for some stuff

it's evil and nonportable but sometimes it's all you got lol

seems odd to me anyhow.. you can put filenames on the command line but as an option it takes a filename as an argument.. im guessing it was patched in and whom ever did it never considered adding a switch to change the delimiter... :/

but you can use xargs here?

how do you mean?

find -print0 | xargs -0 sort, if I understand correctly what you are after

yuripv, no.. find /path -type d -print0 | sort -r | xargs -0 rmdir

that doesn't tell me what you are trying to do

oh.. hang on.. of course i should RTFM .. the files0-from is for a list of filenames to sort the contents of not the output

im trying to get a list of directories nul delimited (ie -print0) from find, reverse the order of the output then pipe that over to xargs

Michelle: sort -z then?

as i said i needed to RTFM.. thanks anyhow..

(i just did --help)

anyone having trouble with latest ports/packages mate-panel? There's a recent bug that may have an upstream patch. When I get far enough along I'll do a patch for it, otherwise the fix would likely be updating ALL of mate to 1,28.2 from 1.28.0 and a patch is faster. The details are mentioned here I think, look at 1.28.1's entry. github.com/mate-desktop/mate-panel/blob/1.28/NEWS

lol, on account of my ancient Debian I'm still rocking 1.24.1

so I guess when I get to a point I can, "git" that source, compare, patch, test. Hopefully just that. then bug+submit fix of course

For the most part find can do everything xargs used to be needed for. What's wrong with: find tmp -depth -type d -exec echo rmdir {} +

Michelle ^^

When I post an example that uses rm or rmdir I always put an echo in front of it for safety. :-}

^

good practice - I do that as root a lot

sometimes copy/pasta what was echoed into a script too

nodnod

I someone is really on the ball they can claim a race condition there and that I should use -execdir instead. I just couldn't roll that off the top of my head immediately.

random: when I wrote my own implementation of rm(1), I deliberately nerfed it

*If someone...

no more "rm -rf W.T.F.???" moments

it won't let you remove more than one file from / at a time

and I think if you do try to do "rm -rf /" it'll say "rm: cowardly refusing to remove '/'"

is worst if you load something filled with symlinks on a Linux system from a tarball to look at it and you need to rm the tree but the links point to things you stilkl need!

oy oy

"rm: cowardly refusing to remove multiple files from /" was the other error I wrote

Probably this using -execdir is the safer race condition way. But IMNHO rmdir is fairly safe and we know who is running on our systems: find tmp -depth -type d -execdir echo rmdir {} \;

yeah whaat was that "do not follow links" switch again...

-P maybe?

rm -rf does not follow symlinks by default.

too late if you firget it.

it wouldn't make sense to

on linux I believe it does

Nope.

it _shouldn't_, at least. I'd have to dig into the GNU code to see what it does

well I just remember that happened to me somehow. was a bit of a pain to fix. ended up force-reinstalling all packages from the cache

What happens is that people do things that push through symlinks, like "rm -rf /*/" and then the command does not see the symlink but the shell has expanded past them.

that actually worked

I did try replacing GNU's rm with a crippled version of BSD rm, but apt-get sometimes complained

rwp yah makes sense

You would have needed to install a dpkg-divert of rm in order to have it stick across upgrades.

it was years ago

that wasn't the problem, lol. the problem was that the package manager assumed a GNU-specific switch was available

GNU rm also won't rm -rf / unless --no-preserve-root is explicitly given.

Ah..., yes, portability of options. You would have needed to have installed a wrapper around the rm and have it accept all of the GNU options. Too much work for me!

lol

I actually think the GNU versions are bloated piles of issh, but that's just me

(not sure how far is acceptable to go with language here. some channels are run by puritans, so I default to self-censorship)

I once got banned from a jebuschat for saying "OMG"

anywho

the safe route is, if Posix doesn't specify a switch, assume it doesn't exist (as opposed to "don't assume it exists")

Years ago someone in the lab ran rm -rf / as them on a network connected system with our work NFS mounted files. It removed some huge amount of it before it was stopped. I remember Doug ran into the server room and started unplugging network cables to stop the process! It took weeks to recover because backups were broken.

The problem was rm -rf $var1/$var2 and the variables were not set. Later management tried to mandate an rm2 which would be safer. But if you were savvy enough to know to use rm2 then you were also savvy enough to write this in a safer way from the start.

x.x

my version does -dfIiRrvx, but -I is undocumented

as is -r

(you're "supposed" to use -R rather than -r, apparently)

You can thank POSIX and the Austin Group for that.

:P

I also wrote a few comments

one was that I didn't implement "4.4BSD -P" because I felt that with modern drives it was meaningless

making -f and -i turn each other off has the comment "Posix me harder", meaning I did what the docs said I was supposed to but I thought it was stupid

The -P is implemented as a noop on FreeBSD too. The original behavior is not useful now.

nod

there's also a "Posix me harder" comment when I return 0 if -f is used with no file parameters

instead of showing a usage diagnostic (which is what I would think is the saner behavior)

A lot of scripts do "rm -f $listoffiles" and the list of files might be empty.

nod

the Posix behavior is treat it as a NOP

I would display the usage and return an error ;p

And to exit 0. No error allowed there. Which would be really annoying if it did error there.

We will just have to disagree on that point. :-/

heh

-dxv are described as extensions to various BSDs

there are several points where I implemented stuff more to prove I could - and when I actually try implementing an OS with this issh, I'll prolly just drop in BSD code instead

Seeing your -f -i comment I don't do this myself but I do see people alias rm='rm -i' and then they need rm -rf to turn that off.

I think that just builds bad habits.

Seeing your -f -i comment I don't do this myself but I do see people alias rm='rm -i' and then they need rm -f to turn that off. There, fixed it. Grr...

I'm not sure what I would've done. prolly "rm: -f and -i are mutually exclusive"

If they are mutually exclusive then it breaks people who want the default -i behavior and use am alias rm='rm -i' to get it.

nod

I remember Red Hat used to do that

for root

That does sound vaguely familiar. It's a bad idea. Root scares people though.

I think using aliases as mollyguards is daft, personally

I'm also not a fan of how a lot of systems alias "ls" with default switches

The thing is though that even though I don't think it is a good idea the OS is going to be used by everyone and we want it to be possible for everyone to customize it to their liking. Even if we don't agree that certain things are a good idea for them to do.

;p

I guess Microsoft prolly thought similar to add "SET DIRCMD"

to change the defaults for DIR

hello what is everyone's thoughts on freebsd, i'm considering switching to it from gentoo but i'm

curious as to what people here think about freebsd vs. gentoo

i ask specifically about desktop usage and i'm just looking for a minimal operating system with decent package support

I personally would push Debian, but I _would_ push Debian lol

it's what I know

apathetic

unrealapex, I have no experience of Gentoo and no opinion of it, but I switched from Ubuntu to FreeBSD and never had cause to regret it.

unrealapex, I switched my desktop from Debian/Devuan to FreeBSD and I am happy with it. But I don't have any laptops running FreeBSD yet.

V_PauAmma_V: what do you like and dislike about freebsd?

Anyone here knows much about implementing and using FreeBSD ktls?

rwp: no, my package went from 8.6.0 to 8.9.1 from an upgrade

hello, i'm considering switching to freebsd from linux, do configuration files for programs follow the xdg specification like suggested on linux or should they exclusivly go in /usr/local? also what should i do for programs that put their config files in ~/.config by default?

unrealapex: Anything in ports probably also exists on Linux, so will have roughly the same behaviour as there.

/usr/local/ is where you'll typically find the libraries and binaries, but for per-user programs the config will go in your home directory.

a lot of Linux developers "forget" that other *xes exist

imo, a properly written Linux program should trivially or even effortlessly port to FreeBSD, OpenBSD, NetBSD, illumos and OSX

when I wrote Marduk, I never did anything "Linux-specific"; it took only minor tweaking to support Windows, and none at all to support OSX

Licca: good job

but then maybe it's because I see Linux as "a *x", and not "a non-Windows"

I feel like Linux, the BSDs and illumos are multiple (basically 3 - GNU, BSD and System V) different implementations of the same thing, i.e., "Unix", even if they don't actually use the _name_

Licca: i just see linux as a nest for breeding systemd infections.

LOL SYSTEMD

systemd was clearly written by people who have never understood Linux

*cough* who never understood UNIX

by not understanding Unix, they fail to understand Linux.

even though Linux is *not* Unix, it is by design a workalike

iirc, "bsdinit" and "sysvinit" are ultimately two variations of a common system

systemd? it's a monstrosity. it's a malignant tumor

what's worse, is for all the grandstanding of projects like debian being "democratized" and a "meritocracy", one jerk made all the linux distros install his malware in a surprisingly short amount of time

still not sure how that coup was achieved, other than the "omg we have to make gnome happy" vibe

GNOME, schmome

the program calling itself GNOME hasn't been for years. MATE is that

as someone from the Linux side whose influence is mostly SVR4...yeah. I don't always agree with the BSDs on this or that. but frankly I think the BSDs tend to get it "more right" nowadays

the three most direct free descendents of the original Unix are illumos, NetBSD and FreeBSD.

so what I've been attempting - it probably shouldn't be any wonder that it kinda looks like I threw BSD and illumos into a blender

freebsd is a breath of fresh air.

I wanted to build a SVR4 clone on top of a BSD kernel. why? because I damn well couldn't write my own kernel, and the BSDs were most suited to my goals

i'm so glad i moved. base system in under 1 G? eat that redhat.

ZFS as a core first class integration? amazing.

what I call "herpetic" licenses makes me twitch. but then again, for some major applications I really have no choice in the matter: that's as free as you're gonna get

there's also what I call "semi-herpetic" licenses. like LGPL. I'm not sure if EPL and CDDL fall in this group, I *think* they do?

also, I don't know how Sun managed to get the OK to do what they did and frankly that scares me: Oracle couldn't put the genie back in the bottle, but it's possible that another company still could - i.e., whoever bought out the company that bought out the company that bought out Novhell

so I'm like...let's be safe...let's just theseus this stuff out

and I'm like, well, as long as I'm already rewriting the damn thing, I can use whatever the hell license I want.

so I decided, for my own part of the project, I'd use a 2-clause BSD-style license. because that just made sense to me

and I think it'll help me SysV-styled crud blend in a bit better with BSD-styled stuff taken from, well, *the BSDs*'

lol

so when I was creating the basic userland stuff I was like, I need something that ends in -ix that's different. and I came up with "Strix"

and then when I needed a name for the overarching concept, I decided to come up with something that looked like "Unix" but wasn't, and settled on "Unitas" - which I felt was a good suggestion of what the goal of the project was (it's Latin for "unity"), since it would be basically a second merger of BSD into the System V line.

just that in this case the System V stuff is, by necessity, fake

antranigv_: hey will you teach me how to set up a mail server on my freebsd?

mane sure

cool

what should i begin with

i'm trying to send a mail to one of my gmail accounts but it doees not get delivered

ouch, google needs lots of antispam setup on your side before that'll work

:s

i've lost track of all the layeres. DKIM, some dns records, etc

mane: Actually settings up DKIM should suffice. There were some tutorials over there I think: davd.io

It is using opensmtpd but the basics should be the same everywhere. Although it is very easy to get hooked on opensmtpd. ;-)

yeah, i pay a webhost to host my site for https and email so i don't have to chase these constantly

I have run my mail on FreeBSD in the past, though I just copied a Debian config

i'm trying to use smarthost on dma

port 587

but it's not working neither

I'm trying to update my bootcode after upgrading to 14.1 and using zfsroot

the manpage at gptzfsboot says GPT bootcode for ZFS on BIOS-based computers

and I have machdep.bootmethod: BIOS

but my first partition is EFI: pastebin.com/AWbFTNeH

slightly confused whether I'm using bios or efi

last1: you probably used GPT(BIOS + UEFI) in the installer

i'm a bit lost, I just installed www/nextcloud on 14.2, on an rpi4. the install finished, but IDK now I now start it. I would ahve assumed that I just start some service named nextcloud, but I can't find one, or a binary

aaah it installs it in a dir, and I have to point a werbserver to that it seems

bpa.st/PIGA <--so long story short, a misbehaving serial terminal server indirectly messed up my /tmp when logged in as root. i can't use rm as it says bad file sewcriptor. but if i try to use mv, either on the native install or booting an installer thumb drive, this is the error i get. can this ever be repaired?

yuripv: yes, I installed on the boot part and it was fine

ufs file system

If I saw that I would run fsck on the file system soonest.

I don't see how a terminal server misbehaving in any way could corrupt a file system though. Instead I would be thinking about bad RAM or flaky disks.

rwp: fsck returns clean

Bad RAM? Bad disk interfaces? Something is definitely not happy. Now the problem is to find out what.

this was an interesting situation. with power flickers, whatever serial characters ended up seting a command that put that specific port in "menu mode" and the menu text ended up being flooded when root was logged in

the ram is new as is the nvme disk

it ended up creating all of these bank files with names of each of the terminal server's menu items

blank

New RAM is most suspect. It hasn't been burned in long enough to know if it is going to be long term stable or not. And unscrupulous mfgs might bin for speed producing flaky product.

I might try booting the install media as a live shell and then mounting the UFS disk and then doing cleanup on /tmp that way. If it works then you have cleaned it up. If it fails then you know another data point.

i did try exactly that. i could not cd or rm. it keeps saying bad file descriptor

personally I always set clear_tmp_enable="YES" in /etc/rc.conf so that /tmp is cleared on a reboot too.

i should add that

With that datapoint then I am thinking bad ram is less likely to be so repeatable. Maybe something wrong with the disk controller hardware path?

I would then be thinking about doing a backup of my data. Then totally destroying the disk image and rebuilding it fresh from scratch. Restore my data on it.

forums.freebsd.org/threads/panic-ufs_dirbad.74953 <--i had this issue too. i had to run fsdb and clear the inode

random: I wonder if it's reasonably possible to daft up a minimal build tree which is just a kernel, libc, clang and sh, and pretty much just that

rwp: the disk controller is an nvme controller, and works otherwise

but i wonder if there is a way to repair /tmp or at lease delete it and create another one

samba uses /tmp and i'm mainly using it as a file server

Heya. I am setting up Let's Encrypt and I seem to be having trouble getting ddns with nsupdate (bind9) to work. Likely I've made an error with authentication or usage, though things look right to me. Does somebody have a tutorial I can follow that actually explains what I have to do?

CCFL_Man, I am still in shock that a live boot could not move /tmp to /tmp.BAD without a panic there too. That's pretty extreme. I would destroy and recreate the system at this point. Backup and restore it. In between the backup and the restore I would do a serious test of the integrity of the NVMe destructively.

Who is the NVMe vendor by the way?

yeah, i was thinking the same. it's an intel NVMe.

I have never had an Intel or a Samsung die on me. Not yet. I have had other vendors fail off. Just had a Kingston fail for example.

i don't think it's bad. this terminal server created over 100 blank files named menu items and shell prompts in /root and /home/user

gh00p, I only very sparingly ever use dynamic dns but I will ask what upstream you are trying to update? Is this your own bind9 named? Or is this a dynamc DNS vendor named?

CCFL_Man, Those would just be files on disk though. On a working system I could create tens of thousands of randomly named files filled with random data and the system would continue to work okay. I would not expect a kernel panic from a file system mv action.

these were call blank files

all blank

in whatever home directory as i was logged into

So the blank file consumes one inode just for the file name, user, group, mode, and so on. Even if it ran you out of inodes that would not kernel panic.

And you said that an fsck of the file system returns clean on it too.

And you said you booted a live boot image and it also kernel panic'd when using that live boot image.

Did you disturb the grave of some powerful Egyptian priest and now have taken on a curse?

there wasn't too many files, like over 100

fsck had to be run and it repaired it after i cleared a bad inode

could clearing that inode have done it

I expect that the fsck would account for that okay. It would treat it as if it were any other data structure integrity failure and take corrective action.

is there a wat i can check what inode belongs to /tmp

How did you run the fack? I would run it with the -y option.

"ls -ild /tmp" will return information about /tmp which includes the inode number.

rwp, I'm trying to set up Let's Encrypt, my own DNS servers, and I'm going this route because I haven't set up the web server yet. But I also just want to grok this properly, it seems to me it shouldn't be this problmatic. I used dnssec-keygen to make a shared secret, I added the "allow-update" section to the zone statement in named.conf, and I

still get NOAUTH errors when I connect.

ls: /tmp: Bad file descriptor

i can do a backup and reinstall freebsd, but i hate to do that if i don't have to

Sorry, SERVFAIL now. yet named is running

the terminal server was a lantronix ETS16PR

it was in service since 2008 and recently retired last month

i even replaced the power supply. it was manufasctured in 1998

gh00p, As far as setting up Let's Encrypt I use and recommend the dehydrated client. It's simple. It works excellently. I previously used the EFF Certbot and switched away from it because it is a HUGE monstrous pig and I did not like it.

If you want a certificate where getting the certificate is the main thing then I would simply assign an address instead of trying dynamic dns just to get the address going and then use it to get a certificate. Yes that's a hack because it can't autorenew when the dynamic address changes, until you manually update it. But it would get things going. And you can chase the dns problem later.

I have no experience with nsupdate to BIND's named and so don't have any suggestions there. Sorry. Good luck!

CCFL_Man, Seeing that "ls -ild /tmp" returns "ls: /tmp: Bad file descriptor" that would send me into a "mind-panic" too! Not to mention a kernel-panic. I am still surprised that an fsck does not do anything for it. I would try "fsck -f -y" on it again and maybe a third time just to be sure.

If that doesn't improve things then a backup, rebuild, restore seems like the path I would use to get to a good final state.

And then I would be hoping that did actually solve the problem. And that it did not simply return in some other weird way.

I share you pain and feel for you that this good hard working terminal servers did something untoward that triggered all of this. That's pretty strange. I just can't imagine how.

Looking now I see those are selling used for $19 on Amazon right now. How things have changed! Those were really quite expensive new.

Oh, and it's new old stock, not used. I am thinking about it myself now!

my Nabu as I got it was NOS

...of course, no way you're gonna run FreeBSD on a Nabu lol...it's Z80!

rwp, I'm trying to set up LE with dovecot and sendmail at the moment. I'm using acme.sh, which seems nice. I'll have a look at dehydrated, as it also has few depencies. My problem remains authenticating the certificates though. I'll take your suggestion though and set it up manually this time to give myself a couple of months.

it doesn't have any dependencies but the error messages are fairly inscrutable

the lack of dependencies is perhaps a bit of a misnomer, given it's a rather large shell script

the openBSD ACME fetching program seemed to have the most helpful error states, but the submission to ports didn't get maintained and it's no longer available

with ACME there are often issues with what external observer (the certificate authority) can see, and what you can see locally, and there is also rate limiting + a bunch of different error states that the authority can return

also, because it's a shell script there's a relatively high risk of return values being interpreted by the script. this was actually exploited by a certificate authority to do local checks outside of the ACME spec!

hm, I recently upgraded from stable/13 to stable/14 and I ran delete-old-libs and now when I run make installworld it fails with Required library libdialog.so.9 not found.

well, I have libdialog.so.10 already, any idea why it is looking for old version?

hm...maybe this is the reason make[1]: "/usr/obj/usr/src.14/amd64.amd64/toolchain-metadata.mk" line 1: Using cached toolchain metadata from build at hostname on Fri Aug 2 20:06:32 CEST 2024

duncan, zoiks, the dehydrated client was exploited, or acme.sh? I've been using the latter to fetch certs (in staging at least) and it seems to work well. I'm running it on a jail host with separate jails for web and mail, so I like avoiding yet another language on the host.

how do i disable that :)

DanGer, is it a port that is failing on libdialig?

og

gh00p: no, it's failing on make installworld (willing to upgrade one of the jails)

debdrup: so not /usr/local/etc/ or i'm guessing that's for system files that aren't a part of freebsd

rwp: i'm giving that a try

debdrup: i finally good a sas hba and a disk shelf. that's for all of the help with my nas

gh00p: acmesh-official/acme.sh #4659

this is crazy in so many ways

whoa. Thank you. Time for a re-think.

I don't have any way to judge whether acme.sh is better or worse than other clients, and they did fix it

It's also a question of why anyone would use a malicious CA to sign their certificates. If a CA is doing things like that then I don't want any association with them as the smell would affect any certs signed by them too.

CCFL_Man, If you are rebuilding then consider using zfs instead of ufs for the rebuild. You might have noticed that almost everyone who uses zfs becomes an advocate for zfs. Come join us! :-)

zfs gang

If uskerine were in the channel I would say that ffmpeg, firefox, and ungoogled-chromium have apparently been able to build because I was offered an upgrade from firefox: 129.0_1,2 -> 129.0.1,2 today.

seems avahi-daemon cannot be launched

nmz, Sounds like a feature to me.

probably

nmz: I've had ordering issue with avahi-daemon and samba but never seen it not start.

maybe I'm doing something wrong

service avahi-daemon start ;Starting avahi-daemon.;[root@ifrit ~]# service avahi-daemon status ;avahi_daemon is not running.

rwp: maybe I'm misunderstanding, but I just want to ssh into a local linux box, is there a better method than avahi?

ssh?

I mean I just "ssh machinename" and log into it.

I cannot

Log into it by IP address? "ssh 192.168.12.34"

yeah, that's what I'm doing

and editing /etc/hosts

You can always create individual ~/.ssh/config sections for hosts too. "Host mischief\n Hostname 192.168.93.185\n hostKeyAlias mischief\n"

I also do that

but I want to... stop

So I guess I must assume that you are working on a network that you do not have control over and therefore can't set up static IP addresses and can't set up a DNS name for these? Is that the limitation?

After you try to start avahi-daemon what messages are logged to /var/log/messages if any? If there is a problem starting there should be information posted there.

If I were on a home LAN, or even a coffee shop LAN, with a couple of devices I might be inclined to dual-home those systems. Simply pick an network subnet that does not collide and assign a secondary static IP address to all systems using that secondary subnet. Then can always connect over the LAN to those other systems.

sorry, I barely know what I'm doing

WARNING: Detected another IPv4 mDNS stack running on this host. This makes mDNS unreliable and is thus not

recommended.

so that was it

lol... stupid weechat

sorry

Weechat is the other mDNS stack?

WARNING: Detected another IPv4 mDNS stack running on this host. This makes mDNS unreliable and is thus not recommended.

That still seems like just a warning and not a reason it would exit. It might be crashing after printing that warning message.

Failed to create server: No suitable network protocol available

Hmm... I am unfamiliar with the details of that error. I might guess but I don't know the solution. Time to search the web for it.

Regarding my hack suggestion of using an overlay subnet what I mean is that if you are using DHCP on a random LAN let's say it gets a 192.168.1.101/24 address/subnet. Or a 10.0.0.101/24 subnet. That would tell me that 172.16.0.0/16 is not being used.

I might add a ifconfig_em0_alias0="inet alias 172.16.44.11/24" example to /etc/rc.conf for em0 device and 44 being a randomly selected subnet in the range and 11 being an address I just picked in that range, to one device, and .12 to another one.

That would allow those coordinated systems to talk to each other using those statically assigned subnet addresses without conflicting with the DHCP assigned subnet addresses. It's a hack sure. But it's one I have used often. Think of it as a truly "poor man's" vlan because though it isn't private but it connects the devices without interacting with DHCP.

nmz: Is this inside a jail?

no

It seems I need to edit avahi-daemon.conf and set disallow-other-stacks=yes

Though I do not understand it I'll save your suggestion rwp

Do you have anything else mdns installed?

pkg info -x mdns

nss_mdns

when did I install this...!?

I don't think that's what causing your problems.

no it is not

sockstat | grep 5353

It would help me understand your network a little better if you described in a short number of words what systems you have that you are trying to communication between.

skered: termbin.com/3hks

If you close chrome and start it you still get the warning and does it start?

screw this, I'll stick to /etc/hosts

warning and error

skered: yes, that was it

I'd guess the conf might be needed then

At we found the other stack.

how bizarre

why is chrome taking over mdns?

I suppose it used a random port

welp

Appears to be some security feature for not leaking data.

rwp: The network is a router that connects to 2 linux boxes and a tp-link wifi extender thingamagig which is where all the wifi's are connected into, the tp-link is not in AP mode so that should cause troubles

also the main router, when it goes offline, which is often, takes down the network with it, so I can't really depend on it at all

3rd world country problems

but anyway, I'll just have to read a networking book because idk what I'm doing nor what I should ask

If this is a NAT router and you have your own private LAN there for your two linux boxes and the wifi extender then almost certainly the router has the feature of DHCP Reservations where it always gives the same MAC ethernet address requesting client system the same static IP address assignment.

Then with those static assignments the addresses never change on the LAN. And never changing it means you can either use /etc/hosts statically which is fine, or you could set up your own domain name server with a private domain name zone for your LAN.

In which case there is no need for mDNS/Avahi because all of the addresses are always static and always known.

Since you have just a couple of machines I would start with /etc/hosts because that's straight forward easy and no stress about it.

I think the feature that would make things nice for you is DHCP Reservations on your router.

It does have and I do use reserved addresses via MAC, however when it loses connection to the ISP, it starts rebooting constantly, so I lose access to everything

meanwhile I switch to another router

but that one isn't configured, but I should do that

Is this router being talked about the one that connects to the ISP? From the description I thought it was a secondary one that was then connected to a router connected to the ISP.

Also I don't know why the LAN would be disrupted if the modem loses connectivity to the ISP. That should not happen.

¯\_(ツ)_/¯