Beta test New Init system, realy improved boot time.

Message

Tronic · Post by **Tronic** » Sat Jul 16, 2005 2:22 am

How to use ifplugd? ngc -u daemon/ifplugd/eth0 works, but ng-update add daemon/ifplugd/eth0 doesn't. I'd prefer to just use ifplugd for all interfaces automatically (instead of specifying each separately). Is this possible?

The ctrlaltdel.i doesn't seem to be doing anything useful (such as reboot the system).

It would be very nice if there was some kind of debug keystroke that would halt initng and allow seeing what exactly it is doing at a certain moment (kinda like Ctrl+Alt+Delete with the task manager in Windows NT5.x). Optimally this would allow spawning shells or something to allow fixing the system in case of b0rked init scripts (this happens quite frequently).

Is it possible to log the shutdown process? Mine halts for a few seconds after [stopping] system/clock. I'd like to know how to speed this up.

Tronic · Post by **Tronic** » Sat Jul 16, 2005 3:22 am

Hmmh.. Is there svn ebuild for this?

How's the precache stuff? I'd love to have whatever my desktop environment & Mozilla needs cached while I'm writing my username & password on xdm prompt.

neuron · Post by **neuron** » Sat Jul 16, 2005 12:54 pm

ctrlaltdel.i does work, this is what is executed when you hit ctrl + alt + del.

and you can't log the shutdown process (yet), as it unmounts your harddrives. Chances are a process needs killing (SIGKILL) instead of sigterm (for example getty). I've considered a SIGKILL_ON_SHUTDOWN variable for those, so we can speed it up a bit.

Tronic wrote:How's the precache stuff? I'd love to have whatever my desktop environment & Mozilla needs cached while I'm writing my username & password on xdm prompt.

You can look at the readahead plugin if you want to, although I haven't tried if that works in a while.

Tronic · Post by **Tronic** » Sat Jul 16, 2005 1:59 pm

neuron wrote:ctrlaltdel.i does work, this is what is executed when you hit ctrl + alt + del.

When I hit Ctrl+Alt+Del, it prints the message, but doesn't actually reboot (with 0.1.3).

and you can't log the shutdown process (yet), as it unmounts your harddrives.

Couldn't it do this in RAM? ngc -x would then do the trick.

You can look at the readahead plugin if you want to, although I haven't tried if that works in a while.

How are the plugins used? I didn't find any documentation on that.

neuron · Post by **neuron** » Sat Jul 16, 2005 2:39 pm

Tronic wrote:
neuron wrote:ctrlaltdel.i does work, this is what is executed when you hit ctrl + alt + del.
When I hit Ctrl+Alt+Del, it prints the message, but doesn't actually reboot (with 0.1.3).

and you can't log the shutdown process (yet), as it unmounts your harddrives.
Couldn't it do this in RAM? ngc -x would then do the trick.

You can look at the readahead plugin if you want to, although I haven't tried if that works in a while.
How are the plugins used? I didn't find any documentation on that.

how would logging in ram help? It'll disapear on reboot anyway.
And dont look at 0.1.3, use current svn, there is a MASSIVE amount of changes since last stable.

You should probably rm -rf /lib/initng before installing the new version too.

Tronic · Post by **Tronic** » Sat Jul 16, 2005 4:05 pm

neuron wrote:how would logging in ram help?

It won't disappear in really warm reboot (ngc -x)! Or so I'd believe, anyway... The shutdown process of the ngc -x reboot behaves exactly like the real thing, so the logs would still be perfectly valid.

What comes to the SVN version, I'd rather wait for the next release if there is no ebuild for it. Gentoo has spoiled me.

Tronic · Post by **Tronic** » Sun Jul 17, 2005 8:31 am

I just came back from a seminar discussing a next generation init system for Debian. The guy giving presentation described how the system should be made. Unfortunately he had not heard of initng before, but I found out that his list of requirements was pretty identical to what initng does. Of course, I did give him a hint and he was very anxious to check it out, saying that NIH (not invented here) syndrome is a very bad thing and that he simply wants the system working, but doesn't have time to actually implement it. I'll try to discuss it more with him later.

EDIT: this was at Debconf5

neuron · Post by **neuron** » Sun Jul 17, 2005 8:06 pm

0.1.4 released.

The changelog in NEWS isn't doing it justice though, a LOT has changed

(also a lot of rather big changes).

There is one known segfault, but we've been completely unable to reproduce it (and we've been trying), and in either case it's probably far more stable than our last version

Evil Dark Archon · Post by **Evil Dark Archon** » Mon Jul 18, 2005 11:34 pm

hopefully a new ebuild will hit portage soon, cause if its radically better than before than i think its worth a shot.

Tronic · Post by **Tronic** » Tue Jul 19, 2005 11:07 am

How does the readahead work? How does it find out what it needs to cache?

neuron · Post by **neuron** » Tue Jul 19, 2005 12:28 pm

Tronic wrote:How does the readahead work? How does it find out what it needs to cache?

there are two seperate readahead systems, system/readahead, which uses /usr/bin/readahead and reads /etc/readahead/readahead (on ubuntu), and the plugin which I wrote, which uses a list of files in the scripts/* dir to manually generate the readahead files.

It needs some cleaning up though.

Tronic · Post by **Tronic** » Thu Jul 21, 2005 1:22 pm

There seems to be a bug in system/initial. It sets up my system with udev, even though the kernel has already mounted devfs.

neuron · Post by **neuron** » Thu Jul 21, 2005 2:08 pm

Tronic wrote:There seems to be a bug in system/initial. It sets up my system with udev, even though the kernel has already mounted devfs.

we only support udev right now (because of the initscripts, and nobody uses devfsd, so we can't test it :p )

normen · Post by **normen** » Thu Jul 21, 2005 5:58 pm

I don't seem to be able to get fbsplash working, to save my life. I'm running genkernel 2.6.12_rc4 on my Sony Vaio laptop. It works perfectly when booting with init. With initng, I can see the initial splash image during the initialisation phase, and as soon as the kernel boots, messages are written on top of that image, eventually scrolling it out of the screen.
I also noticed that the splash and splah_update services are loaded pretty much towards the very end of the boot process.
I tried to load splash in the system level, but didn't change anything.
Any ideas anyone...?

When I was trying to install bootchart to figure out what is going on, but I had a strange problem with bootchart. Details are here http://forums.gentoo.org/viewtopic-p-25 ... ml#2590771

BTW, everything else is working so far. I almost cut my boot time in half, from 80 secs to about 45 secs from GRUB to a the startup sound of KDE.
GREAT JOB JIMMY !!!

neuron · Post by **neuron** » Thu Jul 21, 2005 6:27 pm

bootsplash isn't really working atm, it's not a priority

, we have a lot of things on our todo list.

normen · Post by **normen** » Thu Jul 21, 2005 6:41 pm

neuron wrote:bootsplash isn't really working atm, it's not a priority , we have a lot of things on our todo list.

Do you happen to remember the last version when fbsplash worked, if ever?

Cinder6 · Post by **Cinder6** » Thu Jul 21, 2005 9:54 pm

Hmm, I really want to try this thing, but I run into an odd error... :

Code: Select all

--------------------------- ACCESS VIOLATION SUMMARY ---------------------------
LOG FILE = "/var/log/sandbox/sandbox-sys-apps_-_initng-0.1.4-32672.log"

open_wr:   /etc/initng/system.runlevel
open_wr:   /etc/initng/system.runlevel
open_wr:   /etc/initng/default.runlevel
open_wr:   /etc/initng/up.i
open_wr:   /etc/initng/up.i
open_wr:   /etc/initng/up.i
--------------------------------------------------------------------------------

What is going on? Never seen this before...

neuron · Post by **neuron** » Thu Jul 21, 2005 10:21 pm

normen wrote:
neuron wrote:bootsplash isn't really working atm, it's not a priority , we have a lot of things on our todo list.
Do you happen to remember the last version when fbsplash worked, if ever?

ancient, and not worth using

Cinder6 wrote:Hmm, I really want to try this thing, but I run into an odd error... :

Code: Select all

--------------------------- ACCESS VIOLATION SUMMARY ---------------------------
LOG FILE = "/var/log/sandbox/sandbox-sys-apps_-_initng-0.1.4-32672.log"

open_wr:   /etc/initng/system.runlevel
open_wr:   /etc/initng/system.runlevel
open_wr:   /etc/initng/default.runlevel
open_wr:   /etc/initng/up.i
open_wr:   /etc/initng/up.i
open_wr:   /etc/initng/up.i
--------------------------------------------------------------------------------

What is going on? Never seen this before...

hm, permission problem trying to write to /etc/initng files

Cinder6 · Post by **Cinder6** » Fri Jul 22, 2005 2:09 am

but what would be causing the permission problems? I have tried emerging both with sudo and the actual root account (via su, but that shouldn't make any difference).

running amd64, gcc-3.4.3, if it makes any difference.

I am tempted to stick in a livecd and chroot to emerge that way...

neuron · Post by **neuron** » Sat Jul 23, 2005 1:23 pm

well do you have a /etc/initng?

if you do try removing it, if not mkdir, and if anything solves it let us know so we can fix it

kill · Post by **kill** » Sat Jul 23, 2005 4:26 pm

Provided my understanding of portage is correct. Everything is built and installed into an temporay sandbox directory. After this all files are then recored and moved into their proper place on the filesystem. What is happening with those errors is in the installation phase, before the files are merged, initng trys to write directly to /etc. By not writing into the temporary sandbox it causes an access violation. Since it's being built in a sandbox the applicaiton in question is not allowed to write outside of it. How the installation of files is handled seems to have changed from the 0.1.3 and 0.1.4 releasse has changed. Once the installation method is changed to account for installing in custom locations it should merge correctly.

Cinder6 · Post by **Cinder6** » Sat Jul 23, 2005 8:50 pm

I tried making the dir. Still no go

EDIT: Okay, I got it (0.1.4) working by emerging 0.1.3 then upgrading to .4. I am pretty impressed with the boot time (and it's so goddamn colorful!), but I have 2 gripes.

1. It doesn't shut down all the way (won't power off, regular init says my filesystems weren't properly unmounted).
2. X won't work, period. Won't initialize gdm on startup, nor will it run via startx.

The latter is a major issue for me, so until that is fixed, I will continue to use regular init. I noticed in the download dir on the site, there is a 0.1.5 version with no ebuild. I will try installing that when I get the chance.

EDIT 2: Oof, guess my system was hit harder than I thought. Those unclean unmounts gave 2 of my filesystems "fatal errors", and I had to spend an hour rebuilding them. Then my internet wouldn't work, so I had to do an emerge -e system. Everything is fine and dandy now, and I upgraded to 0.1.5, though I am (understandably) hesitant. Maybe I will wait till 0.1.6...hehe.

Still, I consider the fault to be mine, not yours (running reiser4 on an amd64 with an experimental, albeit excellent, kernel has a certain riskfactor involved...), and I look forward to when initng has matured enough for me to replace init with it!

Keep up the good work!

Pythagoras1 · Post by **Pythagoras1** » Wed Jul 27, 2005 2:13 pm

the prefetch discussion continues here: http://forum.initng.thinktux.net/viewtopic.php?t=166

hoschi · Post by **hoschi** » Wed Jul 27, 2005 6:24 pm

How i can Start Getty with TTY1?
And adding GDM to Default-Runlevel ist definitly not the answer...

Tronic · Post by **Tronic** » Fri Jul 29, 2005 4:14 pm

Shutdown doesn't seem to be working very well (in 0.1.4). It stops some services when I enter the reboot command, but then it hangs. When I re-enter the reboot command (in some shell that is still open), it actually reboots.

EDIT: fixed the version number