Systemd

14 replies [Last post]
vsud
Offline
Joined: 05/29/2014

Would it be possible for you to cover the pros and cons related to Systemd.

I would of thought that Richard Stallman would of published his opinion on systemd.. I would also like to find out if Trisquel will be adopting systemd (if it has not done already)..

The following is a description from a hacker friend who is an experienced admin for Arch.. I have also added a link to a developer called Ignorant Guru who does not like Systemd; I would ask for you not to brush this under the carpet as I have tried to pose this question to three people who are "authorities" in this area but they have not responded..

"IMHO, SystemD is the most infuritating thing to darken the Linux world I have ever encountered. It's like somebody decided Entropy was a winning design philosophy and ran with it. The initial selling point was greater improved boot speed by allowing dependency based parallel starting of system daemons. Now booting any system with it feels like a chore as it chews through things with random delays waiting for other things that it decides are unhappy when they're fine and who knows, because debugging it not as simple as just reading the log file... you've got to futz around with its syntax and modules that throw convention out the window. Creating your own init scripts for custom shiz is no longer as simple as just sticking a short text file in a standard to every system location and throwing the switch. It now involves creating multiple files and learning a whole new syntax and paradigm that seems to exist only to make the task as complicated as possible. I've yet to meet anybody who's looked at the code and honestly been able to say they understand whats going on under the hood. I've heard that its the personal baby of some aspie whos beyond even sharing its terrible secrets with those outside of his immediate earshot. It's flying completely against the modularity and interdepency principles of the POSIX model. Swallowing up everything thats in its way, to turn the entire stack into an interdependent spaghetti nightmare. It's an approach of DaemonX Doesn't work in a way that fits the devs mental paradigm, so DaemonX is bad and must be surplanted with a puppet Daemon that tows the party line. If it were a government... It would speak to you in Italian."

The link that I mentioned http://igurublog.wordpress.com/2012/10/14/gentoo-systemd-torvalds/

Michał Masłowski

I am a member!

I am a translator!

Offline
Joined: 05/15/2010

I use Parabola on a server. It previously had Arch's initscripts: the
most buggy, unreliable and hardest to use init system that I know. It
required big shell scripts for starting a daemon, often failed stopping
them and left old ones running, so configuration changes wouldn't apply
to the running ones. They would also die while running and fail to
restart. Order of starting daemons was completely manual, the next
tasks waited for the previous one unless it was asynchronous (it usually
couldn't reliably be since it was used by other services).

Systemd has simple configuration files for services (several lines in a
well-documented format to start a daemon), manages dependencies, starts
services when the dependencies started, supports socket activation
(makes e.g. handling FastCGI services with nginx easy), is the same on
different distros (I don't need to know completely separate init script
systems for Debian and Parabola; while Debian uses many existing init
scripts written for sysvinit with systemd). Speed is imo a sideeffect
of handling dependencies, parallel service activation, socket activation
and using single-task C programs and service files calling programs
directly instead of shell scripts. Journal has a nicer interface that
what syslog daemons that I knew had, journalctl does what I needed less
and tail for with additional features like per-service filtering.

Ubuntu has systemd since utopic (Debian since Wheezy while it's vastly
improved in Jessie), Trisquel should get it in 8.0.

vsud
Offline
Joined: 05/29/2014

Thank you for the response. I use Arch and to be honest I have not had a problem with it. I was just a bit worried about the comments and feedback from PCBSD forums highlighting issues with systemd (please see the following link)..

http://forums.pcbsd.org/showthread.php?t=22422

Magic Banana

I am a member!

Offline
Joined: 07/24/2010

systemd is free software (under the GNU LGPL). That is probably all rms cares about. As Michał Masłowski tells, systemd brings many new interesting features and is a definite improvement upon the old System V init, which was not designed for modern computer (with hot-plugging, multi-core processors to start init scripts in parallel, etc.).

The discussion was more about systemd vs. Upstart (developed by Canonical). systemd won and Upstart will soon stop being developed. All mainstream GNU/Linux distributions have now chosen systemd: Fedora (and the next RHEL 7), Arch Linux (by the way: you probably want to take a look at Parabola), Gentoo (although it is not the default yet), openSUSE, Mageia, and, more recently Debian and Ubuntu (that admitted the failure to push Upstart after Debian, its base, decided to switch to systemd). Trisquel, based on Ubuntu, will follow. Slackware, which is known to be very slow to adopt new technologies, is, to the best of my knowledge, the only major distribution that did not announce the switch. They will certainly make it later.

jxself
Offline
Joined: 09/13/2010

Don't forget that GNU has DMD: http://www.gnu.org/software/dmd/
They're not using systemd. If you think of every distro as a fork of GNU, they're probably the most mainstream of them all ;)

kpengboy
Offline
Joined: 09/08/2013

> Slackware, which is known to be very slow to adopt new technologies, is, to the best of my knowledge, the only major distribution that did not announce the switch. They will certainly make it later.
Then again, Slackware doesn't even use Sysvinit (they use a BSD-style init) for their own reasons, so no saying if they'll adopt systemd soon.

GustavoCM

I am a member!

Offline
Joined: 11/20/2012
jbar
Offline
Joined: 01/22/2011

Just to add more info about systemd, some common myths about it

http://0pointer.de/blog/projects/the-biggest-myths.html

vsud
Offline
Joined: 05/29/2014

Hi jbar, thank you for the link. It is old.. and things have moved on.. I personally love using Trisquel but do not trust systemd. I don't say this lightly as I have been an Arch User for 18months as well.. but I when I posed the question to Arch Developers, I was shot down..

I am aware that systemd is "cutting edge" and Linux as compared to BSD is up-to-date with lightening fast booting.. but the changes that are being brought in are not being explained fully to users. Linux is about choice or am I wrong?

Allow users to run alternative init system, much like Gentoo..

One the reasons why I have brought this subject up here is because there should be a discussion of choice.. I am not against systemd..

Magic Banana

I am a member!

Offline
Joined: 07/24/2010

Writing init scripts for various init system is a lot of work. You are free to do it. GNU, and not particularly Linux, is about freedom, not choice.

And the way a system starts with systemd is far more understandable than the way it does that with the old System V init. It is a declarative syntax to specify a nice dependency tree vs. a bunch of Shell scripts with spaghetti communications between them.

You can find many recent talks of Lennart Poettering on the Web to learn more about systemd. For instance: https://access.redhat.com/site/videos/403833 and three videos in http://meetings-archive.debian.net/pub/debian-meetings/2013/debconf13/webm-high/

vsud
Offline
Joined: 05/29/2014

Hi Magic Banana, thank you for posting these links. I have seen these before - very interesting and informative. But there is still resistance to Systemd.

Systemd is a good concept but through out the videos the audience are not entirely convinced.. In the Why_Debian_should_or_should_not_make_systemd_the_default.webm file Lennart is actually rude with an audience member..

So for me it is important to be able to listen to users. I have not seen a proper article which does not steer away talking down to users..

If you can show me an article covering a layman's explanation of systemd that would be good.

Systemd is being implemented no matter what, that I am aware off but at what cost?

Magic Banana

I am a member!

Offline
Joined: 07/24/2010

The audience member is not a user. He is an Upstart developer. Another init system that could have replaced the System V init (but now lost against systemd).

But yes, Lennart Poettering can be rude... and so are many outstanding programmers (rms, Linus Torvalds, Theo de Raedt, etc.).

The "layman" does not even need to know the name of the init system he uses! System administrators and people developing/packaging daemons/services do.

vsud
Offline
Joined: 05/29/2014

I would tend to agree with you that the average user does not need to know what is happenong under the hood.. eh why bother? But I, like a large linux user base, are curious about this.. (that is the reason why I got involved in the first place with Linux).

I am of the opinion that systemd is important and that any init daemon with strong parallel capabilities is a necessity esp with the development of technology.

I have used arch on the raspberry pi as compared to raspbian that is developed specifically for it.. Arm Arch uses systemd and there is a clear cut difference in the performance of the OS on the Pi..

There is a brilliant article in Ubuntu User Summer 2014 mag that gives an interesting take on systemd.. I know that Ubuntu does not follow the Trisquel philosophy but as the distro will be adopting systemd, this article is important.

https://www.ubuntu-user.com/

I am to understand the trisquel would not work on the raspberry pi due hardware conflict..

salparadise
Offline
Joined: 09/08/2013

The argument that anyone who resists change is "anti-progress" has left us with genetically modified crops, nuclear power plants built on or near fault lines, a whole raft of technology that has done nothing other than to increase poverty and unemployment whilst increasing the concentration of wealth into a smaller group of hands and global pollution of a near overwhelming nature. In short, that argument was used to silence sensible debate in favour of short-term profiteering.

One wonders where the argument that anyone who hates systemd is anti-progress and 'stuck in the past' will lead?

salparadise
Offline
Joined: 09/08/2013

Got sent a link to this today.
From Mr Poettering's blog.

We want a unified solution that ultimately can cover updates for full systems, OS containers, end user apps, programming ABIs, and more. These updates shall be double-buffered, (at least). This is an absolute necessity if we want to prepare the ground for operating systems that manage themselves, that can update safely without administrator involvement.

http://0pointer.net/blog/revisiting-how-we-put-together-linux-systems.html