initng

slapo · 2005-05-05 19:10:41

Hallo,

has anybody tried this: http://jw.dyndns.org/initng/ with Arch?
It definitely looks promising. If somebody has a test box (which I don't) that could be used to time current boot time and the one using initng, could you give it a shot and post it here?

Thanks,

Slapo

sud_crow · 2005-05-05 19:22:21

lol, i was just going to post about the same thing, just to expand a little on this thing, here is the OSNews comment on the issue:

Jimmy Wennlund has been doing to Linux what Apple has done to Tiger: Make it boot faster. Jimmy wrote initng, a replacement for the Sys-V style "init" application. It allows for better service dependency checking and will start services in a highly parallel fashion, dramatically speeding up the Linux boot process.
initng , like Apple's launchd, breaks somewhat from the traditional Sys-V style init scripts. It is however very easy to configure and setup. Unlike launchd, it does not use XML for it's configuration files.
Initial tests show that initng can dramatically speed up the boot process on single processor machines, and should result in even higher gains on multi-CPU machines. initng is currently considered beta-quality software and will require additional users, documentation writers and developers to become a very viable alternative for Linux users.

phrakture · 2005-05-05 19:34:53

in order to implement that in arch, it would require wild changes to the init scripts, as well as changes to each and ever daemon script

there have bee 100s of tries to replace the sysvinit system over the years, and none have ever caught on... they've all been pretty good though - this initng really isn't anything new...

http://john.fremlin.de/programs/linux/jinit/
http://smarden.org/runit/
http://www.fefe.de/minit/
http://www.energymech.net/users/proton/

jerem · 2005-05-05 21:44:20

Arch's initscripts lack dependency checking, don't they ?

And that background service initialization stuff (@daemon) is incomplete... We do not know if a service failed or not...

Arch has been good at bundling together best features of good distros and giving them the Arch touch... Why not continue ?

I also find the ideo to let the user choose his init system rather interesting.
A post-install script could determine what system is used and put the right service startup file in the right place.

Today, Arch's initscripts do not have a good wireless support and background service startup is still experimental and sounds to my ears like a quickly-hacked piece of code...

Nevertheless, Arch's initscripts are very simple and easily editable. One could say "you can just rewrite the initscripts if you're so clever", but why reinvent the wheel when good alternatives already exist ?

After all, that's how open source is supposed to work : take a piece of code here, take another there, fix bugs, improve it, release it, use it.

phrakture · 2005-05-05 22:01:25

jerem wrote:

Arch's initscripts lack dependency checking, don't they ?

that's up to the user... once again it's the "non automatic" feature of arch - I personally don't want my init process checking daemon depenadancies, because *I* put them in the proper order once... it's more efficient that way

jerem wrote:

And that background service initialization stuff (@daemon) is incomplete... We do not know if a service failed or not...
< snip >
and background service startup is still experimental and sounds to my ears like a quickly-hacked piece of code...

It's not incomplete - that's what you get from backgrounding - that's the point... it's a tradeoff - gain speed and lose the ability to know if it failed or succeeded (you can always check the run directory for running daemons).
It's also not quickly hacked... bash has the inate ability to background processes... this "quickly hacked piece of code" is a bash builtin...

jerem wrote:

I also find the ideo to let the user choose his init system rather interesting.
A post-install script could determine what system is used and put the right service startup file in the right place.
< snip >
Nevertheless, Arch's initscripts are very simple and easily editable. One could say "you can just rewrite the initscripts if you're so clever", but why reinvent the wheel when good alternatives already exist ?
After all, that's how open source is supposed to work : take a piece of code here, take another there, fix bugs, improve it, release it, use it.

So um, don't worry about the fact that instead of a package maintainer making one small (simple) init script for his package, he now has to worry about creating 5 or 10 just so a user can pick which init system he'd like? Would you like to contribute and do that? That is the way open source is "supposed" to work - not "take a piece of code here, take another there", but "I found a bug, I fixed it, here's the patch"

It's too bad people never think about the workload when they bring these things up - you want this? then help out... hell, you can donate to the developers if you can't code...

jerem · 2005-05-05 22:50:46

Sincerely, I did not mean to include every single init system in the world. I just thought that Arch's devs could use what has already been coded, instead of slowly develop a new solution.

I'm not a developer, I just give my ideas and thoughts. The developers are free to use them or not.

I like Arch because it is fast, simple, unintrusive, and I feel "in control".
If a faster init solution exists, I simply *think* it would be a good idea to try it.

I also introduced the idea of "one choosing its initscripts" because there are people who like the actual, and others who could like another or a new one.

Would such a choice need much more work ? Yes, of course. But I'm not a dev and I'm not the one who makes the roadmap.

Dealing with the way how open-source is supposed to work, I assume everyone has its own point of view. I know you are a frequent contributor(which I thank you for, especially for your improved wireless support), but I think a roadmap and development should not be made according to the amount of work it would generate, but rather according to what's on the wishlist.
Other users may think like me. Others may certainly not. I simply proposed an ideo that would satisfy both : let the user choose.
Next to that are the devs, who discuss about what is possible and what is not yet possible.

It seems like distros today are too bound to what they were supposed to become in the beginning. That's why there are so many forks. Instead of making one very flexible distro, people make hundreds of copies.

But what is a distro without its users ? Nothing but useless art...

phrakture · 2005-05-05 22:54:29

I'm not against using a different initscript method, but I don't think the devs should be the ones to do it - you can go ahead and install the package (I can make a PKGBUILD + instructions if you want), but it's going to be up to the user to create the "service" scripts - as well as creating all the dependancies and things...

phrakture · 2005-05-05 22:58:29

Hmmm, it looks like initng provides initscripts for most things... which is a plus - maybe later I'll try and sort it out and attempt to package it

jerem · 2005-05-05 23:13:51

I'm glad you took it easy. This was not aggressive criticism, I'm just a hard to please user :-)

Thanks, I'm already trying to tweak around to install initng, but I'm rather like a barbarious and occasional code tweaker...

phrakture · 2005-05-05 23:16:16

jerem wrote:

I'm glad you took it easy. This was not aggressive criticism, I'm just a hard to please user :-)

Yeah, with an influx of users recently, alot of people suggest things which are just too much work for the dev team, so I got a bit flustered. I apologize.

I'll look into getting initng working... maybe tomorrow or something (see this post for what tonight will be spent doing)

stonecrest · 2005-05-06 00:04:18

The_Fiddler wrote:

I find that most linux distro's now these days display more used ram after a boot up into X , then my XP I used to have. I have been a linux user etc since 97, so I have watched them bloat out. I boot up with Arch, slackware and into X, Its used 150 meg of my ram, if I was to do that in XP it would use up 70 meg.

I doubt that, this is more due to what DE/WM you are using than the fact that you are running linux (or which distro). I boot into X with fvwm and I'm using 40mb of ram. The whole point of Arch is to use only what you want or need to use - if you're using that much ram, it's because you made the conscious decision to run those processes/apps that are bloated.

Anyway, to keep it back on topic.. phrakture, I can't wait to see what you manage to do with this

shadowhand · 2005-05-06 03:19:18

I just saw the post on OSNews and came here just to see if someone had caught it.

I'd definately be down for this to be a default in Arch. Hell, we already use BSD init, why not take it a step further?

twpvw5 · 2005-05-06 03:34:42

WARNING A little OT:
speaking of memory usage is great I'm using fvwm-crystal w/ plenty of eyecandy turned on and I'm only using 71MB of my 1024
I like my lightweight wm

phrakture · 2005-05-06 05:04:59

ok, so in case you didn't see what I posted - I got careless... got hacked... and am now set back a bit - tomorrow though, my girlfriend has a benefit to go to, so I'll have to whole night to "geek out", as she puts it

jerem · 2005-05-06 11:28:59

Good luck for fixing your box :-)

A firewall would have helped, because people rarely have ips beginning with 212.

phrakture · 2005-05-06 14:21:29

jerem wrote:

A firewall would have helped, because people rarely have ips beginning with 212.

I left ssh open, and was careless... 8( already reinstalled... just gotta get my data I was able to save, then I'll get initng together

darkcoder · 2005-05-06 15:04:13

phrakture wrote:

so I'll have to whole night to "geek out", as she puts it

That's a comprehensive girlfriend

darkcoder · 2005-05-06 15:08:10

phrakture wrote:

there have bee 100s of tries to replace the sysvinit system over the years, and none have ever caught on...

Probably because most distros are based on RedHat or Debian and still use the whole SysV init process.

Since Gentoo and Arch use BSD style init process, it will be easier for them to implement initng.

Gullible Jones · 2005-05-09 01:49:25

Is it just me, or does Arch already boot up fast enough, especially considering that most users probably reboot once a week or less?

Edit: well, I suppose that stable, feasible improvements should be implemented when possible... And initng probably counts as such. Still, I don't think it's necessary to rush this into Testing right away.

(Maybe we'll let the Gentoo folks try it first. That's one of the nice things about Gentoo: if there's something wrong with a new package, you'll probably be able to find it on the Gentoo forums before it reaches your distro... )

darkcoder · 2005-05-09 05:18:20

I give it a try, the speed for what I was able to compare is incredible, probably like 1.5x or 2x faster startup. The problem is that most initng scripts were Gentoo standard initscripts backported. And they make references to files like /etc/conf.d/hostname (an example) that we do not have. Dunno if the missing files are the problem, but the system stops at 80%.

Acording to initng home page, there is also a group of Mepis users trying to port it to its distro, which of coure will include most if not all debian based ones.

jerem · 2005-05-09 11:32:14

Yes, mine also stops at 80% loaded.

If you browse through the files in the system section, you'll see that the files are easy to edit. They're most gentooish but easily understandable.

I'm currently working on a heavy copy/paste/hack.
What's nice with this package is that it can exist without hurting the previous init system.
You could even choose your init system on boot, because it's an option passed to the kernel.

And you won't need to edit each package which has a startup script, because you can simply add all those scripts in the initng package.

slapo · 2005-05-09 11:49:18

http://forums.gentoo.org/viewtopic-t-33 … art-0.html

shadowhand · 2005-05-10 08:43:51

I also installed it, and ran into issues with references to /etc/conf.d/* and references to agetty and tty*.

It appears that some of the scripts need to be patched/rewritten to work with how Arch is set up. I'll try taking at them if I can find time in the next couple of days.

darkcoder · 2005-05-11 00:48:19

it looks for some configuration files on /etc/conf.d. Even on Gentoo, the files are available if you are using unstable (~x86) init scripts.

Most of the services already available were ported from Gentoo current initscript system, so we could do the same.

slapo · 2005-06-03 20:08:55

Any luck with initng so far?

Arch Linux

#1 2005-05-05 19:10:41

initng

#2 2005-05-05 19:22:21

Re: initng

#3 2005-05-05 19:34:53

Re: initng

#4 2005-05-05 21:44:20

Re: initng

#5 2005-05-05 22:01:25

Re: initng

#6 2005-05-05 22:50:46

Re: initng

#7 2005-05-05 22:54:29

Re: initng

#8 2005-05-05 22:58:29

Re: initng

#9 2005-05-05 23:13:51

Re: initng

#10 2005-05-05 23:16:16

Re: initng

#11 2005-05-06 00:04:18

Re: initng

#12 2005-05-06 03:19:18

Re: initng

#13 2005-05-06 03:34:42

Re: initng

#14 2005-05-06 05:04:59

Re: initng

#15 2005-05-06 11:28:59

Re: initng

#16 2005-05-06 14:21:29

Re: initng

#17 2005-05-06 15:04:13

Re: initng

#18 2005-05-06 15:08:10

Re: initng

#19 2005-05-09 01:49:25

Re: initng

#20 2005-05-09 05:18:20

Re: initng

#21 2005-05-09 11:32:14

Re: initng

#22 2005-05-09 11:49:18

Re: initng

#23 2005-05-10 08:43:51

Re: initng

#24 2005-05-11 00:48:19

Re: initng

#25 2005-06-03 20:08:55

Re: initng

Board footer