This systemd thing is really out of hand.

[cartman82]

For the end user, the advantages are advertised as:

• Faster boot time
• Better control over forked processes (eg. you can't have a CGI process stay alive after you kill apache)
• Supposedly easier configuration (no scripting needed)

[Eldelshell]

I haven't read much about this since I've moved away from SysAdmin stuff but from your summary up there, I would have hated this solution if the biggest gain is faster boot times.
Anyway, let them play along and when Systemd crash & burns, another contender will replace it. It's not like it's merged with the kernel and other distros won't be using it.

[Eldelshell]

@blakeyrat said:

It's a couple dozen noisy whiners on Slashdot

Exactly. As some around here whining and telling others how to do something they have been doing for 20 years and surely understand better.

[accalia]

@rad131304 said:

IIRC you can force a surface to reboot by holding the power button and the window button below the screen. You shouldn't have to wait for the battery to drain.

@accalia said:

stop responding to everything, including keyboard and hardware buttons

tried that. it really was completely locked up to the point that nothing got through

[blakeyrat]

@hhaamu said:

What problem(s) is systemd trying to solve?

Linux took about 36 times longer than OS X and Windows to boot up. Linux had no standardized way to restart crashed services without the user writing a new shell script specifically to monitor the service-- if you had 48 services you needed to restart, you had 48 monitoring shell scripts. The shell scripts are a bitch to deal with, it takes like 50 lines to do something in them you can do in a 4-line systemd configuration. Etc.

@hhaamu said:

And since I tend to be rather conservative with my software, yes, I'm rather miffed by this sort of upcoming change.

Nobody's holding a gun to your head. If you don't like it, don't upgrade to a distro that uses it.

[hhaamu]

@blakeyrat said:

Linux took about 36 times longer than OS X and Windows to boot up.

It takes about one minute on my decade-old hardware (that IIRC doesn't even parallelise its startup), and a big chunk of the minute is consumed by the BIOS checks.

Linux had no standardized way to restart crashed services without the user writing a new shell script specifically to monitor the service-- if you had 48 services you needed to restart, you had 48 monitoring shell scripts. The shell scripts are a bitch to deal with, it takes like 50 lines to do something in them you can do in a 4-line systemd configuration. Etc.

I don't know about you, but I'd rather keep the service down and investigate the problem, than fill the system's hard disk with 'XXX crashed, but no worries -- I'm restarting it now' log messages. (A full HDD will lead to a bigger crash.) And I recall there already being some software handling the monitoring thing, but since my daemons don't spend that much time crashing, I don't recall its name.

@hhaamu said:

And since I tend to be rather conservative with my software, yes, I'm rather miffed by this sort of upcoming change.

Nobody's holding a gun to your head. If you don't like it, don't upgrade to a distro that uses it.

There's this small thing called 'security updates', which I view as a pretty big gun.

[blakeyrat]

@hhaamu said:

I don't know about you, but I'd rather keep the service down and investigate the problem, than fill the system's hard disk with 'XXX crashed, but no worries -- I'm restarting it now' log messages.

Ok?

Systemd doesn't force you to reboot services, it provides that feature. You don't have to use it if you don't want to. If you don''t like it, turn it off. Or not on. I don't know which way it defaults for each service.

Did you... did you think that was a good argument when you typed it?

Also if you get a full HD from a crashing service, well, then I guess the logging infrastructure in Linux sucks shit because good OSes don't do that. Maybe... fix the bug? Instead of removing the feature that can potentially expose the bug?

@hhaamu said:

There's this small thing called 'security updates', which I view as a pretty big gun.

Well I'm not your fucking kindergarten teacher, you're going to have to stop crying, change your diaper, and decide on your own whether you prioritize getting security updates or using an ancient boot system that sucks ass.

[hhaamu]

@hhaamu said:

From my point of view, it seems that Poettering convinced Gnome to have systemd as a hard dependency, [...] So it pressures [the distros] to adopt systemd.

And I found somebody else making the same conjecture

[hhaamu]

@blakeyrat said:

Did you... did you think that was a good argument when you typed it?

No, not really, but I don't see it as a strong point for systemd either. You're exaggerating the need to restart things to the point of hyperbole.

@blakeyrat said:

Also if you get a full HD from a crashing service, well, then I guess the logging infrastructure in Linux sucks shit because good OSes don't do that. Maybe... fix the bug? Instead of removing the feature that can potentially expose the bug?

I'm just saying that there are circumstances where services repeatedly crash and that is likely to lead to a bigger issue if ignored long enough.

At least, if there's no exponential backoff with the restarting logic.

[Eldelshell]

And @blakeyrat right now is an expert in Linux systems administration, configuration and OS development.

Actually, with all the nonsense he's saying, no.

[blakeyrat]

@hhaamu said:

No, not really,

Right; you couldn't have, unless you're really stupid. You knew that argument would be demolished in like 1/5th of a second. So why the hell did you type it? I wanna get inside your head, man.

@Eldelshell said:

And @blakeyrat right now is an expert in Linux systems administration, configuration and OS development.

Blakeyrat is an expert in everything.

Since you didn't provide any explanation for how anything I've said here is "nonsense", reading your first sentence without the sarcasm is really the only option here. Thanks for the compliment.

[dkf]

@blakeyrat said:

Linux had no standardized way to restart crashed services without the user writing a new shell script specifically to monitor the service

That depends on the service. Things that started from inetd (“simple” services) would get automatic resilience, because each servicing instance would be isolated from all the others; sessions would correspond to processes. (Windows uses something fairly similar for some of the services it handles.) It's the other services which try to have a persistent core that need to be made more carefully.

In general, there's a trade-off. Increasing robustness will tend to be antagonistic to improved efficiency of resource usage unless correspondingly more effort is put into development. It's the same sort of reasoning that lies at the heart of the old engineering dictum “Good, Cheap, Timely: pick at most two”.

[rad131304]

@accalia said:

@accalia said:

stop responding to everything, including keyboard and hardware buttons

tried that. it really was completely locked up to the point that nothing got through

Ah; the home(?) button always at least vibrates for me even when it's locked up. I've never had my surface be that fubar.

[accalia]

it's only happened once, but it was damn annoying when it did.

[flabdablet]

And of course there was enough diagnostic information left behind in the Windows Event Log to pinpoint the cause of the failure. Because Windows is a modern OS that doesn't suck shit. Apparently.

[Polygeekery]

@PJH said:

Anyone else hear those goalposts move?

They made quite the ruckus as they moved by in a flash.

@PJH said:

I've had windows 8 bluescreen because of problems with the swap file. Of its own making.

I have had Server 2008R2 crash, while running virtualized in Hyper-V. So in that case MS made the (virtual) hardware also. But I suppose that @blakeyrat does not believe that happened either because Windows never crashes, and MS produces flawless software...

[ben_lubar]

I still haven't figured out why my computer is reducing its audio volume by 90% randomly and then putting it back to normal. The color correction corruption hasn't happened since the reinstall of Windows, at least.

[ender]

@Intercourse said:

I have had Server 2008R2 crash, while running virtualized in Hyper-V.

Happened to me yesterday. All other VMs are fine, but one bluescreened out of the blue.

[cvi]

@accalia said:

it's only happened once, but it was damn annoying when it did.

I can't quite reliably repo it, but connecting/disconnecting the SP3 from its dock with the "right" timing seems to lock it up more or less completely.

[Medinoc]

What do you mean by "NFP" here?

[dkf]

Now that's unusual. Everything that Google turns up doesn't fit with the context at all. Non-farm payroll? No. Natural family planning? No. Not for profit? TDEMS.

[Cursorkeys]

@Medinoc said:

What do you mean by "NFP" here?

Not Fit for Purpose.

@dkf said:

Now that's unusual. Everything that Google turns up doesn't fit with the context at all.

Interesting. I used that because I've come across it in military/aviation (definitely not NFFP which Google says is common).

Edit: It's either brevity or because dropping the extra F allows for an uncouth overload I guess.

[VinDuv]

@dkf said:

OTOH, changing the logging format isn't a great idea. Adding log corruption into the mix… abysmal.

I don’t see why binary logs would be more corruption-prone that text logs. Maybe the current implementation in systemd is broken, but after all databases store their data in binary format, and they usually are more resilient than flat text files...

[dkf]

@VinDuv said:

I don’t see why binary logs would be more corruption-prone that text logs.

There's no reason for them to be. Provided they're written in strict append-only mode.
@VinDuv said:

Maybe the current implementation in systemd is broken, but after all databases store their data in binary format, and they usually are more resilient than flat text files...

Are you aware of how much work databases go to in order to make that true, and how long it takes for a write to become committed (prior to which, it logically hasn't happened)? You don't want that in a logging system. Instead, you want something fast and where you can recover from any problems that end up in there, probably by using a self-synchronizing envelope format in the log (the simplest of which is in a text file: look for the next newline to find the next record). POSIX O_APPEND FD semantics are very useful for this sort of thing, and it's a thing that I wish Windows had (you can't correctly simulate it with a seek/write SetFilePointer/WriteFile pair due to the inherent problems with race conditions unless you do lots of locking, and that's problematic for other reasons).

The benefits of changing the logging are really very dubious, and corruption (well, non-recoverable corruption) is Bad.

[accalia]

hmm... don't have the dock. i have my beast of a desktop if i need more monitors.

that might be why it's only happened to me once.

[cvi]

@accalia said:

hmm... don't have the dock. i have my beast of a desktop if i need more monitors.

that might be why it's only happened to me once.

Normally, I'd use my desktop too, but being in the middle of relocating, the dock is somewhat handy. (Can't wait to get my desktop back, though. As nice as the SP3 is, doing actual work on it is no fun, especially without an extra monitor. And for gaming .. well, let's say that I'm maintaining a backlog ☹)

I've also noticed that the Wifi driver barfs all over itself every now and then. Sometimes it's fixable by briefly turning on Airplane mode, sometimes it actually requires a reboot.

[flabdablet]

@VinDuv said:

I don’t see why binary logs would be more corruption-prone that text logs.

They wouldn't be, necessarily. The point is that the purpose of logs is to communicate diagnostic information to humans, and humans are actually very good at extracting meaning from corrupted text; from corrupted binary blobs, less so.

Even the way text is corrupted can often communicate useful information to a skilled pair of eyeballs.

In my opinion, if Lennart were truly all fired up about wanting to be able to log chunks of binary data as-is along with the usual kinds of log message, he'd have been better off using something like base64 to embed those into a text log. If I'm looking at a text log and I see non-text bytes in there, I can immediately tell I'm looking at the result of something gone wrong; if non-text is supposed to be part of log files by design I have less to go on.

Systemd's append-only-and-rotate-on-crash model is good, as far as it goes, but it would be better still if the stuff being appended had the kind of resilience that the rather extreme redundancy of text would naturally give it.

Binary blobs lend themselves better to the kinds of structured filtering and presentation offered by something like the Windows Event Viewer. However, in my experience, the Windows Event Viewer is much less useful and far more annoying than /usr/bin/less run against any text file inside /var/log.

And I suspect that Lennart agrees with that as well; the default view offered by systemd's journalctl command is pretty much indistinguishable from running /usr/bin/less against /var/log/syslog.

[Weng]

@flabdablet said:

the Windows Event Viewer is much less useful and far more annoying than /usr/bin/less run against any text file inside /var/log.

Event viewer is fantastic for seeing long-term patterns. It's pretty shit for short-term diagnostics.

[Polygeekery]

@ender said:

Happened to me yesterday. All other VMs are fine, but one bluescreened out of the blue.

But, but, I thought that Windows doesn't crash?

[accalia]

@Intercourse said:

But, but, I thought that Windows doesn't crash?

...

here, let's do an experiment. Take your laptop up to the roof and drop it into the nearest skip... the sound it makes as it hits bottom, that is most definitely a crash is it not?

[boomzilla]

@Intercourse said:

But, but, I thought that Windows doesn't crash?

Hang on, you're not one of these sub-second-grade-readers who also groks metaphor, are you?

[Polygeekery]

@boomzilla said:

Hang on, you're not one of these sub-second-grade-readers who also groks metaphor, are you?

What I am replying to was not metaphor, or sarcasm, or satire, or tongue-in-cheek. At least not the way I read it.

So...maybe?

[flabdablet]

@Weng said:

Event viewer is fantastic for seeing long-term patterns.

Meh. Anything I can do with Event Viewer I can do faster and easier with grep and /var/log.

[VaelynPhi]

@asdf said:

That "argument" is mostly brought up by people who claim that Bash scripts are to be preferred over C code because they're supposedly easier to debug. I'm glad to hear that's not what you meant when you quoted it, though.

Depends on who you are. For system admins, BASH is almost certainly preferred. I'm sure sysadmins who can and want to debug C, especially poorly documented or undocumented C, exist, though. This is the difference between having to adjust some Javascript in your browser for a web page to work correctly vs having to recompile your browser in the context of systemd. The "startup" they're referring to, AFAIK, is the startup of systemd itself, not the startup of any individual component it manages.

I doubt that making sysadmins debug C is something the systemd devs intend; it may, however, be something they have caused.

@cartman82 said:

So you might see why there's such a huge drama over all this.

As for the technical dimension... is systemd a good idea overall, in the long run?

The is a good summary of the issues. Overall, I think systemd is a good idea; it's the implementation that needs some work. (That binary logging thing really rubbed me wrongly.)

@hhaamu said:

So, um, what problem is systemd trying to solve?

The clusterfuck that is sysvinit is the motivating problem. In this respect, it is like the attempt to replace the X server: difficult, and the proposed solutions are lacking in key areas. The issues seem to pop up in that the project basically bloated to the point of trying to replace everything else...

@dkf said:

The benefits of changing the logging are really very dubious, and corruption (well, non-recoverable corruption) is Bad.

Exactly my sentiments.

@flabdablet said:

However, in my experience, the Windows Event Viewer is much less useful and far more annoying than /usr/bin/less run against any text file inside /var/log.

My decade+ of experience suggests similar, unless looking for a pattern. Then, the major hassle is dealing with log rotation, which WEV handles in advance. As for searching the logs themselves, I'll take grep over WEV any day.

@flabdablet said:

Meh. Anything I can do with Event Viewer I can do faster and easier with grep and /var/log.

Or more robustly by fiddling with the logging daemon. WEV is certainly a passive tool in this respect, but perhaps that has more to do with the ecosystem it's logging than the tool itself, and so is an unfair criticism.

[flabdablet]

@VaelynPhi said:

the major hassle is dealing with log rotation

Meh again. ls -rt /var/log/syslog* | xargs zcat | grep blah | less gets the job done, and even if I'm working on a box where I don't have that defined as gl() it takes less time to type than to get Event Viewer up and wait for it to sort itself out.

[KillaCoder]

I can't even comprehend the minds that get upset about this sort of thing... minorly "ugh, that's annoying" sure, but the crazy religious war thing is... odd.

[rad131304]

@cvi said:

I've also noticed that the Wifi driver barfs all over itself every now and then. Sometimes it's fixable by briefly turning on Airplane mode, sometimes it actually requires a reboot.

I've not had this problem in a while, though I can't imagine they've only fixed it for the SRTv1?

[blakeyrat]

@boomzilla said:

Hang on, you're not one of these sub-second-grade-readers who also groks metaphor, are you?

The real problem is these idiots are all putting words in my mouth, then saying I'm wrong, then patting themselves on the back for being so clever by "proving Blakeyrat wrong".

Well, they've proved their fake version of Blakeyrat that said something I've never said before wrong. So... congratulations? Here, I'm going to prove some fictional entity wrong too, so I can feel good: "hey reincarnated Teddy Roosevelt, the Chevy Nova DID have a V-8 production model!" Woo I am so great.

[boomzilla]

@blakeyrat said:

The real problem is these idiots are all putting words in my mouth, then saying I'm wrong, then patting themselves on the back for being so clever by "proving Blakeyrat wrong".

Keep telling yourself that, real version of Blakeyrat.

[Polygeekery]

I get it now! Reading his precise replies on this thread is "putting words in his mouth"!

[blakeyrat]

I never said Windows never crashes. In fact, I said pretty much THE EXACT OPPOSITE OF THAT.

[ben_lubar]

@accalia said:

...

here, let's do an experiment. Take your laptop up to the roof and drop it into the nearest skip... the sound it makes as it hits bottom, that is most definitely a crash is it not?

That's a hardware problem. If you dropped the OS, it wouldn't crash.

[Yamikuronue]

@blakeyrat said:

Windows doesn't crash.

@blakeyrat said:

I never said Windows never crashes. In fact, I said pretty much THE EXACT OPPOSITE OF THAT.

OK then.

[ben_lubar]

@blakeyrat said:

I never said Windows never crashes. In fact, I said pretty much THE EXACT OPPOSITE OF THAT.

@blakeyrat said:

Well maybe you just need better developers. Windows doesn't crash.

?

[powerlord]

Damn it, I was just about to post that.

[accalia]

@blakeyrat said:

Well maybe you just need better developers. Windows doesn't crash.

@blakeyrat said:

I never said Windows never crashes. In fact, I said pretty much THE EXACT OPPOSITE OF THAT.

hmm.... one of these posts is wrong....

tangential: WTF: how do you have 4k posts in a topic that itself only has 141?

[blakeyrat]

Yup now keep reading, get to the relevant part.

[Polygeekery]

@blakeyrat said:

I never said Windows never crashes. In fact, I said pretty much THE EXACT OPPOSITE OF THAT.

So, did someone hijack your account and say the following?

@blakeyrat said:

Well maybe you just need better developers. Windows doesn't crash.

@blakeyrat said:

Stated a fact?

Yes. Yes that is what I did.

@blakeyrat said:

All OSes crash when paired with unreliable broken hardware. That's the only reason Windows NT crashes, and has been for... a solid decade now.

@blakeyrat said:

Because there's an OS that can magically run on faulty hardware? How the holy fuck is that goalpost-moving? Maybe "goalpost defining".

When you said that the only reason Windows crashes is because it is on faulty hardware.

@blakeyrat said:

I don't believe you.

When @PJH pointed out that he has had Windows crash regardless of hardware.

Maybe it was your shoulder aliens that posted it?

[blakeyrat]

@Intercourse said:

When you said that the only reason Windows crashes is because it is on faulty hardware.

Right.

@Intercourse said:

When @PJH pointed out that he has had Windows crash regardless of hardware.

That's because I don't believe him.

I also have no evidence that any of the other crashes in this thread (except perhaps the HyperV-related ones) are due to bugs in Windows.

[Yamikuronue]

OK, I kept reading:

@blakeyrat said:

Well maybe you just need better developers. Windows doesn't crash.

@PJH said:

I see what you did there...

@blakeyrat said:

Stated a fact?

Yes. Yes that is what I did.