I didn’t use Linux two decades ago (I started on a systemd distro a few years ago), and the init system I use (runit on Void) is both simple to use and boots faster than any systemd-based distro I used; which is what I personally care about. I never used any “advanced” features of systemd other than timers and user services, and these have many alternatives.
I think systemd is fine, and has generally proven itself as reliable, but that doesn’t mean the current alternatives have no merit.
I would be interested in understanding how it boots faster considering that systemd init is fairly small and the features of runit is parallel process starting, which is also something that systemd provides.
Are you sure it’s not just that void itself has less crap to load?
This feels like I’m going to have to set up a couple of Arch installs to compare for myself.
I would be interested in understanding how the hell it’s such a big issue, honestly. Even if we generously assume that runit boots 5 times as fast as systemd, on modern systems it makes like a few seconds difference, which… who cares? Who goes around constantly rebooting their shit so many times a day that those 5 seconds they save per boot add up to any significant amount of time?
If you don’t care, good for you; I do care though.
And you can generously assume whatever you want but with systemd I sometimes (across multiple OS’s and machines) had random boots which took way over 5x my current boot time (not to mention extremely long shutdowns, more than 2-3 minutes sometimes compared to less than 2-3 seconds now).
I know this stuff can be analyzed and fixed with tools provided by the systemd project, but I never had to fix anything with runit.
I’m on Fedora so I’m unlikely to be switching from systemd for my main system, I have very little interest in switching init’s and I’ll use the one that comes with my distro but people talking it up makes me wonder.
As for rebooting, my couch gaming system gets started and shut down for each session because it could be days/weeks between use so shaving 10 seconds off the start up time is an interesting concept. Although I could probably do better just replacing the grub delay that HoloISO added for some unknown reason.
Why are you so invested in systemd that you have to project your values on other people?
Besides, different approaches can be educational and solutions to improve performance can always be reincorporated into systemd to improve life for everyone.
Even you asking this question means you should not. Systemd is perfectly fine for most people and the people that dislike it have their very specific reasons. Just use what your distro gives you, you shouldn’t have to care about these things.
Just use what your distro gives you, you shouldn’t have to care about these things.
To expand on this: OP, if you want to try a different init, you're most likely better off switching the whole distro to one using that init. Unless of course your current distro explicitly supports many different init systems (like Gentoo). Don't try to replace it by hand in a distro that expects to run on systemd, it's a recipe for disaster.
Gentoo is great because they give you step by step directions for systemd and openrc so you can see the differences in action. (If you choose to read both sets of directions)
Openrc at least is sysvinit based. Pre systemd the options were really sysvinit and upstart. Upstart was even worse than sysvinit such that afaik no one has bothered to try to revive.
just another one of the holy wars within Linux – for the average user, it’s not going to make any difference – most of the mainstream distros switched over a LONG time ago so if you want to avoid systemd, you have to do a little hunting (ex. Devuan, Void, Gentoo, etc.)
It used to be that everything in Linux was a file, ideally a text file, so if you could find the right file you could access or change what you wanted. Systemd is a big program that manages a bunch of stuff and creates unique commands within its programs for doing so, which moves away from that principle and turns system management into what feels a bit more microsofty (like the registry editor program vs editing config files, etc) and a lot of people don’t like that. But to its credit, it does solve a few problems with cobbling together a modern system that doesn’t suck.
Actually it doesn’t really move away that much from the “everything is a file” principle. For example a “service” is a file describing how to properly start a particular service and if you enable a service, then a symlink is created to your service file, …
It used to be that everything in Linux was a file, ideally a text file
Yes? The entire Systemd configuration is done with files. With a very well defined structure called units that you can use to configure, boot, service startup, networking, containers, mount stuff, open sockets… that’s exactly the point Systemd provides a cohesive configuration file format for a system.
Yeah but I’ve interacted with it a lot and most of my interaction is commands sent through one of their programs. Versus scripts like init.d whose contents I can easily inspect and modify. Init scripts aren’t config files, they’re directly executable code.
As a Windows app developer, I wish Windows service management, boot control, and logging were more like that of systemd. What we have is so much more janky and Sisyphean to work with.
Yes and no. Let me start by saying I use systemd and have never seen any of this as a problem big enough for me to switch.
The Unix philosophy is that a program should do one thing and do it well, the old init.d system did exactly that, but systemd does more than just start programs, for example there’s the systemd-logind service which is used to authenticate users. Why is this a problem? Because some people don’t want to use systemd, therefore they don’t have logind, so if something were to depend on that, like GNOME’s GDM did for a while, it would be impossible to use it without systemd. So in a way people complain about programs becoming dependant on systemd, and systemd grabbing more and more responsibilities to the point where it would become an integral part of Linux, it’s not that systemd is a problem now, is that it has the potential to become so ingrained to everything that it becomes impossible to remove. But in the meantime systemd does provide some advantages, including parallelism which makes the computer boot faster.
Should you switch? Not really, this is more a philosophical debate on what Linux is and should be, I agree with all of the philosophical points and don’t think programs should depend on systemd, but I don’t think that’s directly systemd’s fault.
For desktops it doesn’t make much sense, but now everything’s so oriented to systemd its actually starting to affect embedded Linux applications… so lets try our best to keep the alternatives alive.
I’m on Guix which doesn’t use SystemD. Running on an MSI GS66 Stealth laptop with an Nvidia RTX¹ 3080. I browse the web, develop, run Stable Diffusion on the GPU, game using Steam and Lutris (Chivalry 2 and Diablo 4), Zoom calls for work.
Systemd is a large piece of software. There are ways to make it smaller and disable various modules for it, but usually by default it’s very heavy.
With a traditional init system, it’s just an init system, and you’ll use other other programs to do the other things. This basically means a chain of interconnected bash scripts. Perhaps you’ll run into some integration issues. Probably not though. It’ll be mostly the same.
There is no real advantage to this from a user perspective beyond a philosophical one. Systemd works quite well at doing the things it tries to do, but it’s the Unix philosophy to “do one thing and do it well,” and some people care very deeply that systemd does not follow their interpretation of that philosophy, and that’s certainly a fair reason to not use it.
However, if you’re not having problems with using systemd, I’d say don’t bother switching.
Systems is not the Unix philosophy, at least, not to me. It tries to handle so many different things and use cases. “One thing” normally means a small thing, and initialising everything you could ever think of is not a small thing.
Is the experience of trying new distros very cool? yes
Should you abandon systemd? no
systemd is not bloated and it’s not insecure. If you don’t have any problems, don’t switch (unless you wanna have some fun trying new things, if you do, run a vm).
If a distro that doesn’t use systemd ends up booting much faster or being much easier to configure, maybe those are features you care about. But switching away from systemd in this case is merely an implementation detail. What you’re really doing is moving from a distro to another one that serves you better.
Otherwise, the choice of init system has very little impact to the average user. Maybe it’s worth it to switch init systems if you hate the syntax of unit files and/or the interface of systemctl/journalctl and you use them often enough to warrant the effort. The people who want to use alternatives to systemd without having such a practical issue with it are doing so for philosophical reasons.
These are terrible sources. 3 random CVEs and opinions of randoms on the internet. The “sources” conflate arguments about systemd as an init system with the non-init parts and with criticisms of Poettering, and a lot of it is “this is bad” with no argument or, worse, incorrect arguments. If there is anything in there that actually proves something, link directly to it. I’m not going to shift through mountains of garbage to find it.
Basically, if you do not see any reason to switch from systemd then you should not. The thing with systemd is that it is really big and complicated. If you just use defaults of your distro systemd works just fine, but if you want to (or have to) change something fundamental, then dealing with this monstrosity becomes a bit of pain. You basically end with the situation where you are in a war with your own PC. After some time of this, dealing with an init system that does exactly what you tell it to do feels refreshing. There is also the part, where some init systems (sysVinit and runit) boot faster then others (openRC and systemd), but it is not that significant. I use runit BTW. With my setup I spend much less time dealing with runit then I used to with systemd. That being said I still miss some of systemd features.
The biggest benefit would be to learn more about how unix systems work from the ground up. I’d say if you’ve had no problems ever with systemd then just stick to it. My linux usage predates systemd (by a lot) and I just want options kept open so I’m never forced onto it against my will.
I say this as someone who doesn’t use systemd. There’s not much benefit to it. It’s cool to do if you’re an enthusiast or experimentalist, but from a practical stand point, systemd is most practical.
I don’t see any fundamental reason why systemd would be insecure. If anything, I would expect it to be less prone to security bugs than the conglomerations of shell scripts that used to be used for init systems.
The bloated argument seems to mostly come from people who don’t understand systemd init is a separate thing from all the other systemd components. You can use just the init part and not the rest if you want. Also, systemd performs way better than the old init systems anyway. I suspect many of the those complaining online didn’t really have first hand experience with the old init systems.
If a different init suits your needs better, then sure go with it. But for the vast majority of typical desktop/server stuff, systemd is probably the best option. That’s why most distributions use it.
If anything, I would expect it to be less prone to security bugs than the conglomerations of shell scripts that used to be used for init systems.
Not sure. In the end the shell script were just an easy and consistent way to start/stop programs. If the programs were secure (read: checked the input and sanitize it, did the check for permissions and so on) there is not a big difference.
Also, systemd performs way better than the old init systems anyway.
In what regards ? Boot faster ? Fine, but on a server it does not mean anything, a server does not reboot that often; for a desktop it not that the 5 seconds you gain are a fundamental gain.
One problem I see is with the logs: it is true that the format is documented, but a text format is always readable while a binary format… (been here, done that 🤬 )
In what regards ? Boot faster ? Fine, but on a server it does not mean anything, a server does not reboot that often; for a desktop it not that the 5 seconds you gain are a fundamental gain.
Are you sure it doesn’t mean anything? It means to a LOT of people.
Are you sure it doesn’t mean anything? It means to a LOT of people.
Fine, still not understanding why something that I should run once in a while (on a server) or it is not that critical seems to be so important. Look, I had way bigger gain moving from a HDD to a SDD than switching to Systemd from the old init.
I refuse to belive that for a desktop user a 5 seconds longer boot time is that important. I could understand on a server where, if you work with it, you can have fines for downtime but even in this case it is a thing that could be handled in different ways.
systemd-analyze isn’t only about reducing your boot time by 5 seconds, it’s about when you’ve problems knowing exactly what is happening and when and also about having a clear view of dependencies between services.
I’ve never personally had any problems with binary logs.
I had it and I am sure that I could have solved the problem faster if I could have solved it faster if I did not needed to first understand how to access the logs on a damaged system.
You could always forward to a different logging daemon if that’s a concern.
This does not solve the problem, it only move it to somewhere else.
Add comment