Devuan 1.0 Makes It To A Release Candidate: Debian Without Systemd

[TheBlackCat]

Apple isn't just *Nix, unlike GNU, Linux, or FreeBSD, Apple is actually POSIX certified, so OS X is real UNIX.

Yes, that is exactly what I said.

[microcode]

Hea, if a mass populous wants to throw shit at the wall till something sticks instead of designing a good system, who are we to complain?

Nobody forced Debian to go systemd, nobody forced Ubuntu, nobody forced Arch. Nobody's throwing shit at the wall, the shit is being brought and smeared by the owners of the wall, and they all like it enough to stick with it.

[nigratruo]

Most of you guys are idiots. Really.

Says the idiot himself. Do you expect to get any kind of support and agreement after calling us stupid?

Why do you make systemd or not into something that is almost religious? If you use it you're a windows sellout working with Red Hat, and if you don't use it you're one of thirty people who hate anything new. When Debian went systemd the pro side said "well do you own thing without systemd then?", and now when they do they still whine. On the other hand calling systemd names like systemdodo isn't in any way less childish.

We are not making systemd into something religious, most of us are relaxed about it. It is a new system, which in IT is a big deal (SARCASTIC)....of course we constantly have innovations and things that change for the better, that is what Linux ist. That is why it now has a better startup system, a better update system, a better packaging system than Windows, we did many required reforms and systemd was one of them.You thought that systemvinit was not broken, we disagreed: It had many failings and needed fixing.
Since migrating to systemd on 1 laptop and 2 servers, plus 5 raspberry pis I have had ZERO issues. One was a system not starting anymore, because systemd is structured to either consider a harddisk optional or required for startup and I removed one and did not know that it was marked required, so no big deal. In a new system, you need to learn how it works, that is to be expected.

The issue lies with people that HATE systemd for some very very strange reason. It is like an antithesis for them, a red flag that they stampede towards like a furious bull, projecting all their hate and frustration on.
I know that I'm an open minded individual and that means to not have preconceived notions and prejudices, this is technology and IT and progress is CONSTANTLY happening, whether you like it or not. Fighting this progress is silly and dumb, how do you except a company to hire you if you attribute ideology to a innocent and simple technology like systemd? Linux turning into Windows? Linux turning into a monolithic proprietary OS? All not happening, all very strange assumptions that have no base. Linux betraying its UNIX roots? Well, too late, did that many years ago, and successfully cleaned out the old garbage. How many true and real UNIX systems do you still see in use today? Not many, all of them are legacy. Linux took over, Linux has the advantage that it constantly improves and does so PRAGMATICALLY, so to say, following what users, admins and developers need, not what some quasi religious figure like RMS or other thinks it should be. We get Linux to work, we make it adapt whatever it needs to do, that is how it took over the world and will keep doing that.
It is YOU that make this a religious thing, all we others, the silent majority, are fine with systemd. get out the documentation and figure out how to make it do what you want. I just checked out systemd-analyze and wow, I like this! Pretty nifty:
# systemd-analyze
Startup finished in 16.326s (kernel) + 11.022s (userspace) = 27.348s

I can tell you that my old Debian sysvinit could not do that. Systemd actually has TOOLS that you can use to analyze logs and such.
I always thought that having to cat or less logs was always kind of dumb, (and that saying, I have used Linux since 2001, so I'm hardly a newbie) a logging system should have proper tools that you can utilize.

[nigratruo]

Apple isn't just *Nix, unlike GNU, Linux, or FreeBSD, Apple is actually POSIX certified, so OS X is real UNIX. Few modern operating systems can make this claim. Linux and GNU implement most of the POSIX API calls, while FreeBSD maintains the spirit of the UNIX Community from the golden age, and is a direct port of the Berkeley Software Distribution by its original authors.
.

HAHAHAHA, that is the joke of the day. So a proprietary OS that is closed is supposed to be more UNIX than Linux? Dumb and silly, I have to just laugh at that.
Well, you know what? That makes UNIX exactly what it is: A theory without working and wide spread application, an OS idea that people swoon about that has died, died deservedly by not keeping up with the times and sitting on its laurels for decades. Every UNIX that I know of got replaced with Linux and it was a relief and breath of fresh air to usability and performance.
I'm a quite experienced Linux admin and I have to say I never liked UNIX, it was always something obsolete, so I never use the term, I refer to it as Linux, that is the proper incarnation of it and, unlike most UNIXes, Linux is actually open source and open (anybody that knew the real UNIXes back then knows that they were the most proprietary of software, as BSD can attest when it got sued for copying the precious code and then was found to be not guilty)
So I'm very happy with my LINUX and still won't use the meaningless term UNIX in the future. ;-)

[nigratruo]

Then lets get to the technical operation of systemd. It works, it works well, and traditional sysvinit and scripts where a giant fucking mess. There is not a greater weak point than relying on shell scripts for your start up. its a convoluted mess, and makes process tracking nigh impossible. Stuff like pid tracking is done automatically, and doesn't rely on depending on someone to write a shell script to do it. Systemd units are far easier to read than shell scripts, and there is far less room for fucking one up. Systemd is very modular and doesn't need all modules to run, but it provides enough modules that alone it can boot a working system, hence replacing the role of init scripts in their entirity(yes, init scripts did previously set up the network).
.

I like systemd, because it finally got some structure and organization in the whole startup process, allows parallel startup and you are right, sysvinit scrips were a horrible mess.

.

The worst part of previous init scripts is that every distribution had completely different ways of managing init, which means doing something simple like changing the network configuration was different on every distro. So if you write daemon software, you will need to re-write and test units for every distro, or let them write their own which often leads to failure of sub optimal scenarios. Before systemd, a unified cross distro init system was needed. of the lot, systemd is the best, most roboust, for carrying GNU and Linux into the 21st century and offer the best competition to windows and osx.
.

Agreed. You know, Linux is so sucessful, runs on ANY hardware, because it is so flexible and has been reformed for decades. In Windows, the naysayers, the non technical people always get their say and they don't want Windows changed (or improved for the matter), they don't know what is wrong with it, that it is a terrible mess of ancient technologies that were kept in the desperate wish to keep everything compatible back till 1999.
In Linux we improve these things. I'm an expert in Linux audio and for many years, Linux audio was such a gigantic mess, that I could recommend it to no one ever, ALSA, OSS, Jack, ESD, Arts and the other seemingly millions of audio systems made it horrible to use to produce sound.
Windows and Mac Users laughed at me using Linux for audio "you know, we just have a system that WORKS, you con't have to know what it is, because it is reliable and robust. JUST ONE SYSTEM!" and they were absolutely right. There are areas where standardization is super important, where having 10 different competing systems is useless and even harmful. Pulseaudio is not perfect, but now all software supports it, all but the ones that use Jack, which is for professional audio, so now we got two systems only (and bare metal ALSA, which is not a contender, because it does not allow soundcard sharing by several processes) and this is a huge improvement.
The same was done with systemd and the same will be done with wayland replacing the X11 server, which at this point is so old and a convoluted mess, a steaming pile of dung, that is starts to stink really bad.
And we will have the exact same fight with you guys, you oldtimers, that want to hold on to the old and never change a thing, because you think it is perfect. Well, it is not.
I know, IT is very unforgiving: It is a field that knows no antiquities, no "beautiful old stuff" that can be cherished, everything we do gets always better, faster, more efficient and leaves no room for old stuff, that gets outclassed with everything. You old guys have the freedom to stay with the old stuff, keep it alive in a specially made distro, but the rest of us will move on and we need this stuff to work, on a large scale, to be maintainable for many many servers, and insisting on some obsolete technology seems the same idiocy as somebody that does not like TCP/IP and instead would rather use a Token Ring network. Well, if you do this at home, have a lot of fun with it, but doing this in a company where you work, well..... you get the point.
.

Devaun is going absolutely nowhere. Despite the outcry over systemd, most users, most administrators, and most developers either don't care or actually like systemd. I'd say I'd wish them the best on devaun, but watching them fail when they realize they don't have enough talented people to maintain it, and it, and the industry just ignores them is going to be funny. To anyone who jumped ship over systemd butthurt, don't let the door hit you on the way out. The linux community will be better without your baseless conspiracy theories.

I would think that Devuan is not going to be very successful. Truth be told, the systemd haters where not many to start with, but they clamoured very very loudly. I remember that they did not even want to show their names when they started the new project and that already felt very odd, seems like people that are either not convinced or sure of what they are doing, don't want to stand behind it or are conspiracy nuts, that think that systemd was done by the US government (tm), to destroy everything and everybody, that the Vatican is of course behind it, all orchestrated by the Illuminati and that chemtrails were specially forumlated to make us cow-like supplicants that accept systemd without properly rebelling have been brainwashed into liking it.

I'm afraid that the Devuan makes are rebels without a cause, they just don't like something new, are critical, but fail to provide any better alternative.
After saying all that, I believe that it is important that Devuan exists and that people can find refuge there, if they feel attacked. It is freedom that makes Linux great, not coercion and force. As they say with free speech "I might strongly disagree with your opinion, but I will fight for your right to say it!". And it is only free speech really, the issue is only true when somebody does disagree with you.

[nigratruo]

Nobody forced Debian to go systemd, nobody forced Ubuntu, nobody forced Arch. Nobody's throwing shit at the wall, the shit is being brought and smeared by the owners of the wall, and they all like it enough to stick with it.

What I find odd and strange is that some people pretend that it is shit. In my case, the new stuff is a lot better than the old stuff, while the old stuff was manure, this one here looks like fertile compost. I think a lot of people that glorify the old sysvinit way are forgetting all the nasty issue that that system had and how hard it was to troubleshoot.
There was no methodology, nothing to teach a newbie, except: "Here, you can do whatever you want, no rules, you can write your own scripts and contribute to the mess. Faster parallel startup? You don't have to worry about that, lad: Just does'n work." ;-)

[bnolsen]

You tried troubleshooting with systemd analyze-blame?

Also, what is a crossover cable network? Gigabit NICs don't need crossover cables.

Duhh, gigabit is auto detect so it doesn't matter. but everyone knows what a crossover cable does.

on archlinux seems they like running a mandb update on start which sucks the most, but there's tons of other crap that goes crazy doing things. As I mentioned before I don't like the overengineered mess that is systemd which adds in another full layer of unnecessary complexity and 10s if not 100s of thousands of lines of code for systems that should take thousands at the most. I prefer spending my time debugging why services don't start, not debugging an overly complex init system. What systemd is reminds me of some mistakes I made when I was a junior programmer which was to massively overdesign stuff mostly for the sake of overdesigning something at the time.

Regardless I prefer void linux and will install that in the future, or a system like that which probably will have a system similar to runit/s6/whatever else.

[jrch2k8]

My 2 main dev systems (athlon2 x4 for term and dual xeon e5 ivy bridge) are archlinux. Their boot times have also degraded with the systemd switchover and I have some problems with the term wireless and occasionally with the dual xeon's crossover cable network now occasionally stalling the whole boot process. I used to run an nfs server which wasn't much fun with systemd either...

I've moved my other systems over to void linux which is easily far more straightforward with how everything runs, and it's wicked fast. But definitely missing some polish and completeness. Poor overloaded maintainers...

systemd seems to me to be an overengineered arbitrary artificial layer which is just more crap to have to figure out. I'm a KISS type guy and systemd always sets off all kinds of bells and alarms for me. Sorry, just a crotchety old guy who's been running linux since fall 1991...

my slowest server:

Opteron 4130 (4 cores)
8 Gb 1333 ECC DDR3 (4 + 4)
root filesystem:
ZFS mirror pool 1 TB (2x 1TB seagate barracuda) with DEDUP and LZ4 compression ON
Archlinux latest (pacman -Syyu as of today)
systemd-swap with ZRAM
systemd-networkd

Software:
2 KVM virtual machines (OpnSense 17.1 + Archlinux 802.1q router)
5 Archlinux Containers on ZFS subvolumes(nspawn) and sandboxed with extra seccomp and resource limitations [Samba4 PDC(30 pcs), Databases(mongodb, mariadb, postgres), Apache2 + php7 +memchaced server, myowncloud system, DistCC server]

systemd-analyze
Startup finished in 8.048s (kernel) + 5.545s (userspace) = 13.593s

in my main workstation

Xeon E3-1231V3
16 GB 1866 ECC (8 +8)
C232 MB GigaByte
Radeon R9 280 3GB OC
root filesystem:
ZFS mirror pool 240GB (2x 240GB transcend SSD) with DEDUP and LZ4 compression ON
2x ZFS mirror pool 2TB (2x 2TB WD NAS) with DEDUP and LZ4 compression ON
Archlinux latest (pacman -Syyu as of today)
systemd-swap with ZRAM
systemd-bootd UEFI
Gnome 3.24 Wayland

systemd-analyze
Startup finished in 2.048s (kernel) + 1.328s (userspace) = 3.376s

Note: with ZFS my journald never gets corrupted ever or suffer slowdowns, if someone was interested to know lennart was right

Note: if your times are way slower than mine check with systemd-bootchart since some service file must be messing you up(probably old combat sysV files still around), also don't use BTRFS since i noticed it make systemd slower depending the kernel version, systemd by default force fsck on boot if your root filesystem is getting constantly checked or is slow scanning check your SMART or move to XFS or ZFS

systemd-bootchart and systemd-analyze are your friends to make sure

[eigenlambda]

systemd-analyze to find out when "startup" finished and how much of that time is "kernel or userspace", a singularly bad time to be making those decisions.

The old way would be to type dmesg and find the point at which you think the system is started up and which second that was on.

[jrch2k8]

systemd-analyze to find out when "startup" finished and how much of that time is "kernel or userspace", a singularly bad time to be making those decisions.

The old way would be to type dmesg and find the point at which you think the system is started up and which second that was on.

at least for my systems it seems accurate enough when manually checking systemd-analyze total time against a chronometer