by Anonymous Coward writes:
on Monday October 30, 2017 @04:30AM (#55456287)
Read Slashdot since 2003, but I almost never post. Sure, I'm still posting as AC. Posting here in the (hopeless) hope that somebody will read this and reconsider...
I've been an (advanced but uncertified) RHEL and CentOS system administrator and user for 15 years now. My first work started with Redhat Linux 9. I am now holding on to RHEL 6 for dear life despite spending significant effort evaluating RHEL 7. I find systemd distributions to be buggier, less reliable in general, and harder to administer. I won't go into details in this post, because they are amply present elsewhere on the web. I find the community surrounding systemd, especially with respect to "principle of least surprise"-violating behavior to be disheartening to say the least.
I viewed RHEL 4, RHEL 5, and RHEL 6 with excitement, and I embraced the changes that came with each distribution as mostly improvements. I am not afraid of learning a new system for the sake of improvement. With RHEL 6 end-of-life fast approaching, however, I am preparing to switch all the institutions and environments I support to FreeBSD. This isn't meant as any sort of threat but just as a fact that I have sadly resigned myself to accepting. In the early days of RHEL 7, I vehemently hoped that RHEL 8 would move off systemd. As the years marched on without any signs of such a decision and an apparently doubling down on systemd, I feel like I have no options for a reliable init system without the wealth of bugs and breaking decisions systemd developers seem to feel comfortable routinely making.
With all due respect, and choosing my words very carefully, I think you are somewhat mistakenly perceiving the nature of the broad acceptance of systemd across distributions. I think its somewhat component-viral nature has forced its adoption in many cases, and it seems quite apparent that this has happened to the effect of far greater antagonism and technical pitfalls than of advantages.
This is deliberately not written as a technical critique. Those can be found elsewhere. This is written as an entreaty from a random anonymous coward on the Internet on behalf of many anonymous cowards that, in my experience, systemd may not be as well-received as it seems. Definitely too late for RHEL 8, I know, and I don't approach a FreeBSD transition lightly. I know how silly it must sound to transition from the Linux to BSD philosophy and environment over an init system, but reliability is key for me and for those I serve.
A professional tool is more forgiving than a crap tool, every professional knows that. A crap tool that you are forced to use is a special kind of bad.
Meanwhile, the rest of us have adapted, find systemd works just fine and have moved on with our lives.
No you haven't, you just haven't been in a situation where systemd fucked you. Here is an example taht happened the other day to me, svchostd, erh systemd, remaps the keyboard in some distributions (Linux Mint 17) so that the pipe character isn't accessible from the keyboard when you use a console, like in a system recovery situation.
Linux Mint is developed by assclowns. Stupid security breaches, stupid package name conflicts, and their attempt at a rolling-release Debian Edition was a complete clusterfuck. Also, no one takes your desktop issues seriously. If that's all you use Linux for, suck it up and buy a Mac. The rest of us are concerned with the world's foremost server platform, and our ideas about stability and reliability are in direct opposition to your shiny-UI mentality.
Also, "it remaps the keyboard" => "I don't know how t
And yet they still manage to patch root kernel vulnerabilities months before Redhat lazily get around to it. The time to patch and critical exposure to known vulnerabilities on Redhat is appalling.
Yeah.
It is unfortunate that Linux MINT went the systemd route.
You can look at their blogs how they praise systemd.
Of course they never gave specific examples as to how this makes:
a) their life easier
and, more importantly
b) the life of users easier
The usual "argument" is "we use it because all distributions use it". And, honestly - that is not a good argument at all.
I don't think it's fair to put it mildly to so quickly characterize OP as a bad workman. OP claims to be seasoned and claims a history of welcoming adaptation. Hell, 15 years is longer than I've done sysadmin work but its mostly gotten easier over the years.
I'm in almost the same situation as the OP with EL 6 death. Considered other distros but have no confidence that SystemDeath won't penetrate them too. FreeBSD is the safest choice long-term and thats where I'm headed. If you personally have never been sc
True. A good workman, in contrast, picks the correct tool as the first step in working on any job and sidesteps a lot od the problems that a bad workman faces.
You don't shoot an IMAX movie with iPhones. You can say a bad workman always blames his tools...but when he's a carpenter and he's handed a piece of rebar and some chewing gum to measure, cut, and hammer, he has to go out of his way to learn to become MacGyver to work with the garbage tools he's been handed. Even then he's still going to produce some pretty crude work.
systemd is rebar and chewing gum in the hands of a carpenter.
Except that, if you want Windows, you take what Microsoft offers, which includes built-in IE. (It's perfectly usable to download a decent browser. Don't knock it too much.)
Every Linux distro is a different but similar OS. It's entirely possible to keep systemd out of a distro, but nobody actually does. If it were really that horrible, someone would have a non-systemd distro that would become popular. Until this actually happens, I don't see how systemd can be that bad.
As init systems go, I actually like systemd, far more than Upstart or, especially, Solaris SMF. The XML-laden can of worms known as SMF is particularly something I hope I never have to work with again (then again, with Solaris being barely on life support now, that's a pretty good bet). The only thing I'd wish is for systemd to confine itself to being an init system. Tying important system components tightly into systemd, on the other hand, is something I think is a Bad Idea.
Systemd dies if there is no cgroup support in the kernel. Poettering: "To make this work weâ(TM)d need a patch, as nobody of us tests this"
R!/dir/.* destroys root. Poettering: "I am not sure I'd consider this much of a problem. Yeah, it's a UNIX pitfall, but "rm -rf/foo/.*" will work the exact same way, no?"
Systemd dies if there is no cgroup support in the kernel.
Poettering: "To make this work weÃ(TM)d need a patch, as nobody of us tests this"
And this very first entry made me stop reading. You're a moron.
In order to be able to constrain services' resource usage, systemd needs cgroup support, that's why it is, among other things, a cgroup manager. Obviously running without cgroups is not a supported configuration right now, as who in his right mind would run a fucking cgroup manager without cgroups?
If you mean Poettering's responses to these issues are moronic, I agree
If you are saying the poster is moronic, please explain, citing history as to why those are not valid criticisms? I'm pretty sure the actual history and his quotes are accurate.
Systemd dies if there is no cgroup support in the kernel. Poettering: "To make this work weÃ(TM)d need a patch, as nobody of us tests this"
R!/dir/.* destroys root. Poettering: "I am not sure I'd consider this much of a problem. Yeah, it's a UNIX pitfall, but "rm -rf/foo/.*" will work the exact same way, no?"
Perfect is the enemy of good. For all of my use-cases, good is good-enough. Then again, my use-cases are not your use-cases.
Now, you have piqued my interest. What sort of use-cases do you have that are dependent on a specific init system? The only thing I can think of is something so integrated into the low-level OS stuff, like starting/stopping the process, or system logging, that it will require a lot of development work to rewrite for systemd.
Care to fill me in?
[A computer is] like an Old Testament god, with a lot of rules and no mercy.
-- Joseph Campbell
Never comment but... (Score:5, Informative)
Read Slashdot since 2003, but I almost never post. Sure, I'm still posting as AC. Posting here in the (hopeless) hope that somebody will read this and reconsider...
I've been an (advanced but uncertified) RHEL and CentOS system administrator and user for 15 years now. My first work started with Redhat Linux 9. I am now holding on to RHEL 6 for dear life despite spending significant effort evaluating RHEL 7. I find systemd distributions to be buggier, less reliable in general, and harder to administer. I won't go into details in this post, because they are amply present elsewhere on the web. I find the community surrounding systemd, especially with respect to "principle of least surprise"-violating behavior to be disheartening to say the least.
I viewed RHEL 4, RHEL 5, and RHEL 6 with excitement, and I embraced the changes that came with each distribution as mostly improvements. I am not afraid of learning a new system for the sake of improvement. With RHEL 6 end-of-life fast approaching, however, I am preparing to switch all the institutions and environments I support to FreeBSD. This isn't meant as any sort of threat but just as a fact that I have sadly resigned myself to accepting. In the early days of RHEL 7, I vehemently hoped that RHEL 8 would move off systemd. As the years marched on without any signs of such a decision and an apparently doubling down on systemd, I feel like I have no options for a reliable init system without the wealth of bugs and breaking decisions systemd developers seem to feel comfortable routinely making.
With all due respect, and choosing my words very carefully, I think you are somewhat mistakenly perceiving the nature of the broad acceptance of systemd across distributions. I think its somewhat component-viral nature has forced its adoption in many cases, and it seems quite apparent that this has happened to the effect of far greater antagonism and technical pitfalls than of advantages.
This is deliberately not written as a technical critique. Those can be found elsewhere. This is written as an entreaty from a random anonymous coward on the Internet on behalf of many anonymous cowards that, in my experience, systemd may not be as well-received as it seems. Definitely too late for RHEL 8, I know, and I don't approach a FreeBSD transition lightly. I know how silly it must sound to transition from the Linux to BSD philosophy and environment over an init system, but reliability is key for me and for those I serve.
Thanks for your responses.
Re: (Score:1)
SYSTEMD IS SHIT.
Re: (Score:0)
A BAD WORKMAN ALWAYS BLAMES HIS TOOLS
Re: (Score:1)
A BAD WORKMAN ALWAYS BLAMES HIS TOOLS
A professional tool is more forgiving than a crap tool, every professional knows that. A crap tool that you are forced to use is a special kind of bad.
Meanwhile, the rest of us have adapted, find systemd works just fine and have moved on with our lives.
No you haven't, you just haven't been in a situation where systemd fucked you. Here is an example taht happened the other day to me, svchostd, erh systemd, remaps the keyboard in some distributions (Linux Mint 17) so that the pipe character isn't accessible from the keyboard when you use a console, like in a system recovery situation.
oh no, blame your distr
Re: (Score:0)
Linux Mint is developed by assclowns. Stupid security breaches, stupid package name conflicts, and their attempt at a rolling-release Debian Edition was a complete clusterfuck. Also, no one takes your desktop issues seriously. If that's all you use Linux for, suck it up and buy a Mac. The rest of us are concerned with the world's foremost server platform, and our ideas about stability and reliability are in direct opposition to your shiny-UI mentality.
Also, "it remaps the keyboard" => "I don't know how t
Re: (Score:0)
And yet they still manage to patch root kernel vulnerabilities months before Redhat lazily get around to it. The time to patch and critical exposure to known vulnerabilities on Redhat is appalling.
Re: (Score:1)
Re: (Score:0)
I don't think it's fair to put it mildly to so quickly characterize OP as a bad workman. OP claims to be seasoned and claims a history of welcoming adaptation. Hell, 15 years is longer than I've done sysadmin work but its mostly gotten easier over the years.
I'm in almost the same situation as the OP with EL 6 death. Considered other distros but have no confidence that SystemDeath won't penetrate them too. FreeBSD is the safest choice long-term and thats where I'm headed. If you personally have never been sc
Re: (Score:3)
A BAD WORKMAN ALWAYS BLAMES HIS TOOLS
True. A good workman, in contrast, picks the correct tool as the first step in working on any job and sidesteps a lot od the problems that a bad workman faces.
Re: (Score:2)
systemd is rebar and chewing gum in the hands of a carpenter.
Re: (Score:2)
systemd is 'popular' for the same reason IE is. They're included with the OS. Neither would be chosen on their own merits.
Re: Never comment but... (Score:0)
Unfortunately, every major distro maintainer disagrees with you. Should make you think.
Re: (Score:2, Funny)
Yes. Every member of the Debian technical committee voted for systemd -- it was a unanimous decision! /s
Re: (Score:0)
Nobody voted for sysvinit. Though Devuan was started to provide only sysvinit as a supported init system (choice is bad!).
Re: (Score:0)
You might want to review the relevant discussion on the Debian mailing lists.
The so-called TC "vote" was anything but.
Re: (Score:2)
Except that, if you want Windows, you take what Microsoft offers, which includes built-in IE. (It's perfectly usable to download a decent browser. Don't knock it too much.)
Every Linux distro is a different but similar OS. It's entirely possible to keep systemd out of a distro, but nobody actually does. If it were really that horrible, someone would have a non-systemd distro that would become popular. Until this actually happens, I don't see how systemd can be that bad.
Re: (Score:0)
Enjoy BSD, it's a fine OS family, and you can take your Bash scripting skills with you to the grave.
For me, systemd itself isn't the problem. (Score:2)
As init systems go, I actually like systemd, far more than Upstart or, especially, Solaris SMF. The XML-laden can of worms known as SMF is particularly something I hope I never have to work with again (then again, with Solaris being barely on life support now, that's a pretty good bet). The only thing I'd wish is for systemd to confine itself to being an init system. Tying important system components tightly into systemd, on the other hand, is something I think is a Bad Idea.
I've ripped-and-replaced several
Re: (Score:0)
I'm adding this here as the most technically egregious offenses related to systemd
Part of the reason why this award was given:
https://www.csoonline.com/arti... [csoonline.com]
Systemd dies if there is no cgroup support in the kernel.
Poettering: "To make this work weâ(TM)d need a patch, as nobody of us tests this"
R! /dir/.* destroys root. /foo/.*" will work the exact same way, no?"
Poettering: "I am not sure I'd consider this much of a problem. Yeah, it's a UNIX pitfall, but "rm -rf
Processes owned by a user with a leadin
Re: (Score:2)
And this very first entry made me stop reading. You're a moron.
In order to be able to constrain services' resource usage, systemd needs cgroup support, that's why it is, among other things, a cgroup manager. Obviously running without cgroups is not a supported configuration right now, as who in his right mind would run a fucking cgroup manager without cgroups?
Re: (Score:0)
Re: (Score:1)
I'm just being efficient. If the very first point of a longer post is moronic, I can expect the rest to be the same tired FUD, so I just skip it.
Re: (Score:1)
Don't worry, the others are moronic too.
Re: (Score:0)
If you mean Poettering's responses to these issues are moronic, I agree
If you are saying the poster is moronic, please explain, citing history as to why those are not valid criticisms? I'm pretty sure the actual history and his quotes are accurate.
technical issues with systemd (Score:0)
I'm adding this here as the most technically egregious offenses related to systemd
Part of the reason why this award was given:
https://www.csoonline.com/arti... [csoonline.com]
Systemd dies if there is no cgroup support in the kernel.
Poettering: "To make this work weÃ(TM)d need a patch, as nobody of us tests this"
R! /dir/.* destroys root. /foo/.*" will work the exact same way, no?"
Poettering: "I am not sure I'd consider this much of a problem. Yeah, it's a UNIX pitfall, but "rm -rf
Processes owned by a user with a leadin
Re: (Score:1)
I viewed RHEL 4, RHEL 5, and RHEL 6 with excitement,
You need to get out more.
Re: (Score:1)
Perfect is the enemy of good. For all of my use-cases, good is good-enough. Then again, my use-cases are not your use-cases.
Now, you have piqued my interest. What sort of use-cases do you have that are dependent on a specific init system? The only thing I can think of is something so integrated into the low-level OS stuff, like starting/stopping the process, or system logging, that it will require a lot of development work to rewrite for systemd.
Care to fill me in?