Gentoo Resisting?

Posted Nov 30, 2014 18:58 UTC (Sun) by jb.1234abcd (guest, #95827)
In reply to: Gentoo Resisting? by vonbrand
Parent article: The "Devuan" Debian fork

"How do you know "sysadmins dont't want systemd"?"

I think that the Debian fork in progress should be a good example:
http://debianfork.org/

They describe themselves as "Veteran Unix Admins".
You can assume with a high degree of probability that they know their stuff. And that includes a healthy dose of confidence in their own abilities and understanding of "what works".

They put a reference on their site to a comparison between typical service
startup by means of a shell script and systemd.
There it is:
https://web.archive.org/web/20141020161905/http://forkfed...

I look at both ways side by side and have a sympathy for these guys.
They want to know what is *executed* by reading the script and if there is
a problem they can react almost instantly. It gives them confidence !
On the other hand, the systemd's declarative config file seems to be easier to read (even to compile), but are they able to comprehend
instantly what actually happens when those directives are processed behind the scene by systemd ? Do they have the same confidence ?
What if they want or need to make changes beyond config file ? What if they want to try something (a fix, a prototype) ? Prospect: C code files and program compilation.

There is much more to this then the above example.
I think they are right.
jb

Gentoo Resisting?

Posted Nov 30, 2014 20:32 UTC (Sun) by vonbrand (subscriber, #4458) [Link] (15 responses)

I could pass myself up as "veteran Unix sysadmin" too by setting up a website, or as "well-seasoned queen of Sheba" for that matter. Let's see who they really are (in names and numbers).

As sysadmin, I have very little sympathy for "want to know exactly what is being run", I want it to work, period. Today's machines (and software running on them, and services offered) are just much to complicated for having the whole source in your head. Maybe in Lions' day, not anymore for a few decades. Configuration has to be simple, transparent, and foolproof. It is just not aceptable if common functions have to be written into each and every daemon (where it will be done exactly the wrong way around a few dozen times), that common operations work on the "wait a bit, cross your fingers and try again" principle, or starting and stopping setvices is done with no regard to dependencies.

Gentoo Resisting?

Posted Nov 30, 2014 20:41 UTC (Sun) by dlang (guest, #313) [Link] (14 responses)

> Configuration has to be simple, transparent, and foolproof.

you won't get disagreement from many people on this statement.

What you will get is disagreement on what "simple" "transparent" and "foolproof" mean.

the systemd config looks simple, but it's far from transparent and foolproof is questionable (the fact that there are experienced people running into problems with it indicates that it's far from foolproof)

shell scripts can be simple or they can be complicated. It depends on many things, including how comfortable you are with shell and what helper functions you have. The RedHat approach of init scripts that call other scripts that call other scripts ... that parse config files is not what I call simple. but that's not a requirement of sysvinit, it's just how RedHat opted to setup their scripts. In my opinion, they hide too much in layers of scripts and there is a lot that could be simplified with appropriate (clear) helper tools.

I prefer the approach "make simple things easy and hard things possible"

Gentoo Resisting?

Posted Nov 30, 2014 21:00 UTC (Sun) by vonbrand (subscriber, #4458) [Link] (1 responses)

Hundreds of lines of shell scripts doing some very un-intuitive hopping through hoops is nearly antonymous with "simple" in my book. Almost each time I had to mess with the "simple" configuration I had to make sure to have a nearby tty with assorted man pages open while wading through huge files setting up assorted shell functions, most of which had no bearing on the issue at hand. And this when the mess was finally somewhat tamed by setting up said common infrastructure. Not my idea of "transparent". Once a miscreant managed to slip in a few lines setting up a backdoor, due to the "flexibility" shell scripts offer. All those I can do without, thank you so very much.

Gentoo Resisting?

Posted Nov 30, 2014 21:04 UTC (Sun) by dlang (guest, #313) [Link]

> huge files setting up assorted shell functions, most of which had no bearing on the issue at hand

That is a problem with the distro that is setting up and maintaining those scripts, not with the idea of the scripts themselves.

I agree with you that init scripts that go that route are far from transparent. I hate them as well. But if you look at the different distro families, you will see that some are far worse about this than others are.

Gentoo Resisting?

Posted Nov 30, 2014 22:00 UTC (Sun) by jb.1234abcd (guest, #95827) [Link] (7 responses)

"the systemd config looks simple, but it's far from transparent and foolproof is questionable (the fact that there are experienced people running into problems with it indicates that it's far from foolproof)"

I agree.

This is quite telling:
$ apropos systemd |wc -l
155

Just browse thru all this stuff. Look for clues when your system is in trouble. Many settings are default and not present in the config files, thus out of sight.

I remeber my first encounter with systemd. It was like this:
systemd: please stop trying to take over the world :)
https://lists.fedoraproject.org/pipermail/devel/2011-June...

Few things have changed since for the better, quite many for worse (systemd-*d being one of the offending additions).
People feel that this stuff can be intimidating to learn, not to mention to handle in complicated work setups and emergency situations.
Quite few people think it is an overkill in some respects.
Some people think it is suitable for desktop machines, but not for servers.
Others resent an exclusive and/or default status afforded by the distros.
That's why they rebel.

Gentoo Resisting?

Posted Dec 2, 2014 19:39 UTC (Tue) by ms_43 (subscriber, #99293) [Link] (6 responses)

> $ apropos systemd |wc -l
> 155

Agree with you, the systemd documentation is really quite good and thorough and sets an example for other projects to follow.

Gentoo Resisting?

Posted Dec 3, 2014 17:08 UTC (Wed) by flussence (guest, #85566) [Link] (5 responses)

It's as if you actually think lines of code or documentation correlate to quality.

Systemd has a long way to go before it catches up to the gold standard, the MS OOXML spec.

Gentoo Resisting?

Posted Dec 3, 2014 20:47 UTC (Wed) by ms_43 (subscriber, #99293) [Link]

> Systemd has a long way to go before it catches up to the gold standard, the MS OOXML spec.

Heh, i've not found the equivalent of the baby pacifier border style in systemd's documentation yet. Maybe you could file a RFE against systemd?

Gentoo Resisting?

Posted Dec 4, 2014 16:31 UTC (Thu) by nye (subscriber, #51576) [Link] (3 responses)

>It's as if you actually think lines of code or documentation correlate to quality.

Wait, are you trying to claim that documentation is not correlated with quality?

Gentoo Resisting?

Posted Dec 4, 2014 19:43 UTC (Thu) by dlang (guest, #313) [Link]

> Wait, are you trying to claim that documentation is not correlated with quality?

If the OP isn't, I am.

Gentoo Resisting?

Posted Dec 4, 2014 23:38 UTC (Thu) by mathstuf (subscriber, #69389) [Link] (1 responses)

I'm of the camp that no docs are better than wrong or bad docs. Though I can't say systemd has had either (maybe harder to navigate, but it's certainly close to completely documented).

Gentoo Resisting?

Posted Dec 8, 2014 12:32 UTC (Mon) by nye (subscriber, #51576) [Link]

> no docs are better than wrong or bad docs.

Sure, bad things are bad.

But all else being equal, code that's properly documented is - as a general rule - more useful, and likely to be less buggy. Plus it often makes it easier to understand why some code is buggy (which all code inevitably is).

My experience has been that the kind of person who thinks that their code is so good that it doesn't need to be documented is - once again, as a general rule - the same kind of person that writes incomprehensible bug-ridden spaghetti.

But the difference between having no/little documentation and extensive documentation is even wider in cases like this, where the documentation describes glue layer interfaces and behaviour. I'm reasonably confident in asserting that it is an objective fact that documented APIs and protocols tend to be better than undocumented ones.

Overall, I don't think anyone can seriously claim that there's not even a correlation between the amount of documentation of some code, and its quality, unless they have some extraordinary evidence to support that claim. Not that they are *equal*, but *correlated*. Having tonnes of documentation is certainly not a guarantee of good code, but looking at the documentation *is* an extremely strong indication. I'm honestly surprised that anyone's even disputing this.

Gentoo Resisting?

Posted Dec 9, 2014 8:42 UTC (Tue) by cyronin (guest, #99592) [Link] (3 responses)

I want to like systemd, but in my experience so far with it, its just a bit too opaque and immature for the kinds of issues I end up having to troubleshoot with it. for example, I have been needing to troubleshoot an issue with a raid controller inconsistently mounting in initramfs, and aside from a pretty ascii diagram in a manpage and a few brief mentions on the website and mailing list, there really isn't a whole lot of accessible information about the boot proccess that allows me to proceed further with troubleshooting this...

Overall, its an intriguing system to say the least, but It seems to need a little bit more time in the oven before every distro decides to rely on it.

Gentoo Resisting?

Posted Dec 9, 2014 8:54 UTC (Tue) by Cyberax (✭ supporter ✭, #52523) [Link] (2 responses)

What initramfs has to do with systemd?

Gentoo Resisting?

Posted Dec 9, 2014 10:14 UTC (Tue) by cesarb (subscriber, #6266) [Link] (1 responses)

> What initramfs has to do with systemd?

At least on some distributions, the initrd nowadays runs systemd. Run "journalctl -b" and look for a "systemd[1]: Running in initial RAM disk" line; if you see it, your distribution is running systemd within the initrd. You should also see a "systemd[1]: Switching root" line just before the switch to the real root.

Gentoo Resisting?

Posted Dec 9, 2014 20:08 UTC (Tue) by johannbg (guest, #65743) [Link]

Shortcomings of systemd setup in initramfs is irrelevant to systemd itself ( other than it being used ) and bugs against upstream implementing this ( dracut, mkinitcpio etc ) should be filed so they can address the issue.