10 ms·
The Tragedy of Systemd [video]
- creshal 8y agoThe real tragedy is that so many complain about systemd and insist it's awful, yet nobody ever bothers to actually make a better init system. It's either 1980s crap that's totally unsuited to modern environments, or systemd with all its bugs and design flaws.
- Bluecobra 8y agoI thought that Sun did a pretty good job with SMF in Solaris 10, it’s too bad that Oracle did everything they could to ruin Solaris.
- sgt 8y agoThey say Larry Ellison is still burning Solaris and other Sun books over the weekends. At one point his Japanese bamboo house almost caught fire.
- onli 8y agoThere is runit, used in void. And OpenRC in Funtoo/Gentoo. And the systems used in alternative Operating Systems, Solaris just got mentioned. There are a bunch of better alternatives out there.
- Accacin 8y agoI believe in the talk he goes into why this (in his opinion) isn’t the case.
- onli 8y agoI don't think so? I skipped only the questions, maybe there? I didn't hear him mention those other systems, might have missed that. You can argue that he does that indirectly: He likes that systemd provides something like a service layer, which the tools I mentioned don't really do. Maybe you mean that. That would work, I just happen to think he is wrong there. That's not the job of init, making it a plus point of an init system is some kind of mental gymnastics I don't agree with (though I liked that talk).
- isostatic 8y agoWhat modern environments? The problem with systemd is that it's replaced systems that met peoples needs with a single system that many people consider to be worse. Perhaps there are some niche environments that systemd is needed, and that's great -- but why break things for everyone else?
- loeg 8y agoTo flip this around, if you replaced systemd with the old rc init in e.g. Fedora today, you could make the exact same argument: > The problem with [old init] is that it's replaced systems that met people's needs [systemd] with a single system [old init] that many people consider to be worse. > > Perhaps there are some niche environments that [old init] is needed, and that's great -- but why break things for everyone else? So the argument is kind of meaningless without a tie-breaker of some kind; perhaps popularity vote. I think you're under the impression that your views of systemd's relative suitability are held by the majority of users, but I believe that is incorrect.
- isostatic 8y agoChange is something many people are naturally averse to, so to change something there should be a good reason for it. If you sell something and people choose to use it to solve a problem, that's great. If you force change on someone, and not only not deliver any benefit to them, but actually make things worse, expect pushback. Remember too that most people's systemd impact wasn't just replacing init, it was a whole raft of changes from log formats to dns. Imagine you had a working systemd setup and someone came along and replaced everything. They claim that there are use cases it's better for, but you don't see yourself with those use cases. You would be averse to that change. You then find out that pretty much every distribution going has got rid of systemd. I suspect you wouldn't be happy.
- keypusher 8y agoNo, the tragedy is that you didn't watch the video, because the author isn't complaining or insisting that it's awful. Instead, he (a well-known FreeBSD developer) goes through the history of init systems and offers some justifications for why systemd is the way that it is, and what challenges might need to be overcome to make a "better" init.
- colpabar 8y agozing!
- Gby56 8y agoGreat talk imo, my TLDR would be : > systemd is not the best it could be, but it's there. > distros patting themselves on the back because they don't use it means they haven't listened to actual needs and stayed behind. > death threats over software that nobody else tried doing are ridiculous
- episodeiv 8y ago> death threats are ridiculous Fixed that for you...
- mondoshawan 8y agoDon't forget the gaslighting with "software has bugs" and "you just dont like change", while glossing over legit complaints about the outright narrow minded design of the system that is actively harmful in some cases (systemd-timesyncd, I'm looking at you). Personally, the video made me rather angry because all he does is set up straw men and knock them down using the same terrible pro-systemd arguments that get trotted out every time someone has a legit problem with its design.Some of the distros that use systems like runit and openrc are using those sustems because systemd doesn't fit their technicals needs, and likely won't ever because of designs that are made without community input. Oh, sure, he mentioned the Debian developer resignations, but outright failed to point out the reason why they resigned: systemd was being railroaded into the system as a political move rather than being judged on its technical merits.
- theamk 8y ago> Oh, sure, he mentioned the Debian developer resignations, but outright failed to point out the reason why they resigned: systemd was being railroaded into the system as a political move rather than being judged on its technical merits. No. Ian resigned because he tried to ignore systemd's technical merits and tried to used political moves (GR process) to to kill it. As a result, most of Debian's community told him to top messing around with politics. Ian's resignation letter [1] explicitly says that: > I should step aside to try to reduce the extent to which conversations about the project's governance are personalised. [...] > The majority of the project have voted to say that it was wrong of me to bring this GR at this time. [1] https://lwn.net/Articles/621895/ https://lwn.net/Articles/621895/ Please stop saying systemd was being railroaded. I have no idea what was going on in Redhat (I do not follow their mailing lists), but at least in Debian, it got on its own technical merits, and Poettering had no special influence there. (That is not to say it is without problems; there are plenty of them. Still, the lesser of evils...)
- timw4mail 8y agoThe tragedy of having to watch a youtube video instead of reading a transcript...
- loeg 8y agohttps://news.ycombinator.com/item?id=19023232 https://news.ycombinator.com/item?id=19023232 LWN does a reasonably faithful summary.
- ilovecaching 8y agoWhat problems do people have with systemd? I actually think it’s pretty great. They also made some tools like busctl that are really nice for Linux developers.
- isostatic 8y agoSomething that annoyed me 10 minutes ago Old fashioned way, when I restart apache having made a typo: Output of config test was: AH00526: Syntax error on line 2 of /etc/apache2/sites-enabled/mysite.conf: Invalid command 'blah', perhaps misspelled or defined by a module not included in the server configuration Action 'configtest' failed. Now [....] Restarting apache2 (via systemctl): apache2.serviceJob for apache2.service failed because the control process exited with error code. See "systemctl status apache2.service" and "journalctl -xe" for details. failed! The former is far more informative
- gh02t 8y agoI mean I agree that it should just show you the output of "systemctl status" in the error message, but your comparison isn't entirely fair. The commands systemd error message give will show you the same thing as the generic sysv output.
- isostatic 8y agoBut it's that type of unhelpful behaviour that leads to dislike and accusations of arrogance (how dare you type a command that doesn't have the word "systemd" in it)
- kbenson 8y agoIs it unhelpful, or just different? In the case of systemd, isn't it logging all the errors in the journal so you can see prior occasions it had a problem and errored? What is the error output in the jouirnal? Is it better, worse, or the same? Systemd makes the process two steps, which you think is worse, but you've failed to address whether the possible benefits of those two steps (error retention, possibly more or different levels of error reporting) have been taken advantage of, and how that looks. There's a difference between something being worse and just different. So far all I've seen is different, and possibly worse, but you've left out the information that would be required to determine that.
- antirez 8y agoThe tragedy of systemd is that it's the wrong solution to a real problem. Because most people against systemd will keep saying that the old init system was ok, they don't move forward trying to provide a true alternative. So systemd became a standard despite the issues, because it solves certain real problems that people have, even if not great.
- kfwhp 8y agosystemd became a standard because the guy works at Red Hat so he can make GNOME require it (there is no way to use GNOME without systemd; distros that allow that such as Gentoo do so by patching GNOME, which is costly.) I agree that a more modern init system is necessary, but the idea of systemd is meh and the implementation is subpar.
- tapoxi 8y agosystemd became a standard because Debian adopted it after a lengthy discussion, partially because big Debian users like Spotify expressed an interest in systemd. https://lists.debian.org/debian-ctte/2014/01/msg00287.html https://lists.debian.org/debian-ctte/2014/01/msg00287.html
- nickik 8y agoConspiracy nonsense. It became standard because the existing solutions were crappy and nobody else bothered to do anything about it.
- kfwhp 8y agoI don't believe this to be a "conspiracy". It's nothing shadowy, really. What they did is completely natural. That doesn't mean I support the idea though.
- kevin_b_er 8y agoThe intentional and willful breaking of screen and tmux was to fix a GNOME bug of GNOME not closing up as it should when the user logs out, so systemd was changed to mass kill processes. The interplay between GNOME and systemd in backroom dealings is a major pain point.
- ardy42 8y agoPreviously: LWN article: https://news.ycombinator.com/item?id=19023232 https://news.ycombinator.com/item?id=19023232 (1/29/2019) This talk: https://news.ycombinator.com/item?id=19157293 https://news.ycombinator.com/item?id=19157293 (2/13/2019) Another youtube version of it: https://news.ycombinator.com/item?id=17778202 https://news.ycombinator.com/item?id=17778202 (8/16/2018)
- isostatic 8y agoI think that the section on change at is great. The first 1500 seconds are about systemd any why it is and what it's trying to do, but that doesn't really get to the core of why I complain about it. https://youtu.be/o_AIw9bGogo?t=1526 https://youtu.be/o_AIw9bGogo?t=1526 Systemd is change. Massive change that's being forced on us by a twat. Unlike Linux, which was a kernel delivered by a twat but was take-it or leave-it, this is being forced on us. It's untrustworthy and threatening and unknown. In some cases it's an attack on our identity -- "the way you did it in the past is rubbish, do it this way". Clearly we are going to be defensive about it. This isn't unique to people who've used linux for the last 3 decades. Douglas Adams knew this 20 years ago: 1) everything that’s already in the world when you’re born is just normal; 2) anything that gets invented between then and before you turn thirty is incredibly exciting and creative and with any luck you can make a career out of it; 3) anything that gets invented after you’re thirty is against the natural order of things and the beginning of the end of civilization as we know it until it’s been around for about ten years when it gradually turns out to be alright really.” His original article (in a newspaper - remember them?) is at http://www.douglasadams.com/dna/19990901-00-a.html http://www.douglasadams.com/dna/19990901-00-a.html He mentions John Humphrys and his sneering view of this new fangled internet. Humphrys was (and still is) the bbc radio 4 breakfast presenter. At the same time Terry Wogan was the radio 2 breakfast presenter, a similar age (about 60 at the time). His attitude was far more opening, energizing a whole generation of middle-aged Brits to technology and the internet. Change was something to be embraced, not feared. This talk has made me realise that my attitude to systemd is far more like Humphreys than Wogan, and that scares me.
- syn0byte 8y agoI ask this in every systemd thread but I genuinely would love a real answer. Can anyone give me exact real-world examples of where systemd is objectively better suited than any number of other init systems? The fact that I still have yet to see any such examples in the last half a decade seems pretty telling. Wouldn't and shouldn't there be tons of such examples given the rhetoric?
- foldr 8y agoIt makes it a lot easier to set up a service than System V init. Writing a service file is much easier and less error prone than writing shell scripts.
- JdeBP 8y agosyn0byte explicitly said "any number of other init systems", and was clearly trying to avoid people trotting out, as you and several others have, the tired old fallacy, called out by the Uselessd Guy long since, of assuming that only systemd and van Smoorenburg init+rc exist ... * http://uselessd.darknedgy.net/ProSystemdAntiSystemd/ http://uselessd.darknedgy.net/ProSystemdAntiSystemd/ (https://news.ycombinator.com/item?id=8488235 https://news.ycombinator.com/item?id=8488235) * https://blog.darknedgy.net/technology/2015/09/05/0/ https://blog.darknedgy.net/technology/2015/09/05/0/ * http://jdebp.eu./FGA/run-scripts-and-service-units-side-by-side.html http://jdebp.eu./FGA/run-scripts-and-service-units-side-by-s... ... and then calling the latter "System V init", when AT&T System V had actually adopted a new system with separate system and service management several years before Linux was even invented. * https://news.ycombinator.com/item?id=18788974 https://news.ycombinator.com/item?id=18788974
- foldr 8y agoI took syn0byte to be asking for an example of an init system that systemd is better than. I can't really see any other interpretation of their comment. >and then calling the latter "System V init", This is what everyone calls it in the context of Linux init systems. You may not like that particular piece of terminology, but please don't give me a hard time for following the standard usage.
- cr0sh 8y agoWhat I have come to believe is the "Tragedy of Systemd" - and if I am wrong on this, please let me know - is that it was a virtually unilateral decision to make the change, by one person, without any input from others. It was like all at one shot - here's this huge and largely non-transparent init system created by seemingly one person who it felt like was shoving it down everyone's throat. My way or the highway style. That, on top of what felt/feels like a lack of transparency about how it works (I could look at init scripts and understand them fairly easily and how they worked together - systemd seems almost a mystery to me; I just have to trust it works - so far so good I guess). Maybe I am wrong on all of this - again, if I am (and there's a good chance I am I feel) - please let me know.
- JdeBP 8y agoYou are wrong. Read https://news.ycombinator.com/item?id=14735100 https://news.ycombinator.com/item?id=14735100 .
- fridsun 8y agoI just must let this one thing out. > 34:02 binary logs are not a bad thing as long > 34:06 as you have the tools to pull them apart Really? Did he just "as long as" the biggest downside of binary log compared to text log, that is way fewer tools work on a binary log due to its special encoding? With the same logic, flu viruses are not a bad thing as long as you are immune to them.