[NEWS] OpenRC and baselayout 2 will reach ~arch soon

Message

gringo · Post by **gringo** » Fri Apr 18, 2008 11:17 am

Or maybe it is just that you only notice some improvement if your boot time is >3 minutes or so.

yup, it actually made a difference here on a very slow box ( pentium 133 - 64 MB ram), it shaved about 12 seconds which is quite cool.
One of my laptops ( G4 alubook) also looks like it is booting faster, but that´s probably placebo effect as you said, i have no numbers. In case there´s a difference it surely would be something minimal.

Kudos to Roy !

cheers

entity · Post by **entity** » Fri Apr 18, 2008 2:51 pm

Whoa! Smooth, swift bootup I see - just like what Gentoo should be!

Kate Monster · Post by **Kate Monster** » Fri Apr 18, 2008 5:42 pm

Updated to OpenRC today, everything went smoothly - rewrote my /etc/conf.d/net and recreated /etc/init.d/net.eth0 symlink, and that was it. It seems like it's booting faster, but I won't be able to tell for sure until I try it on my slower machine. Well done, uberlord.

rogerx · Post by **rogerx** » Fri Apr 18, 2008 7:49 pm

NOTE 1: /etc/env.d/90xsession is provided by =x11-apps/xinit-1.0.8-r2. The article fails to mention this, but does make note xinit provides the /etc/env.d file. So, I would imagine the more proper method would be, unmasking =x11-apps/xinit-1.0.8-r2 might also be appropriate instead of just adding the /etc/env.d/90xsession file?

NOTE 2: I was still using the old "broadcast" instead of "brd" variable for the /etc/conf.d/net file. Don't know if the order of defining "brd" or "netmask" was an issue either.

--- /etc/conf.d/net.old
+++ /etc/conf.d/net
-config_eth0=( "192.168.1.3 broadcast 192.168.1.255 netmask 255.255.255.0" )
+config_eth0="192.168.1.3 netmask 255.255.255.0 brd 192.168.1.255"

NOTE 3: spamd init.d script requires "--name" arg. See bugs.gentoo.org for opened bug with patch/solution (- not committed or marked as fixed yet).

RESULTS: The migration process, as long as following the Gentoo OpenRC Migration guide, is pretty much flawless. The increase with boot timings is what I would expect when using C code!!

Mikkl · Post by **Mikkl** » Sat Apr 19, 2008 9:33 am

rogerx wrote:NOTE 2: I was still using the old "broadcast" instead of "brd" variable for the /etc/conf.d/net file. Don't know if the order of defining "brd" or "netmask" was an issue either.

When starting net.eth0 with brd defined before netmask in /etc/conf.d/net it always gave me this error:

Code: Select all

 * Caching service dependencies ...                                       [ ok ]
 * Bringing up interface eth0
 *   10.0.0.6 ...
Error: either "local" is duplicate, or "netmask" is a garbage.            [ !! ]

after having switched the two statements it now works again

Boot time increased noticeable, especially the gaps where the computer seemed to do nothing (but wasting time on executing the old shell-script?) between two init-scripts are now gone.

krinn · Post by **krinn** » Sat Apr 19, 2008 9:44 am

no speed increase, but no slowdown also, so it's cool

i might be the only one to notice it, so i put it here, but it also kill /etc/hosts file here, as i refer my hosts in it to be use with distcc, this has disturb distcc working (not much an issue, but still an issue)

rogerx · Post by **rogerx** » Sat Apr 19, 2008 8:23 pm

Mikkl wrote:
rogerx wrote:NOTE 2: I was still using the old "broadcast" instead of "brd" variable for the /etc/conf.d/net file. Don't know if the order of defining "brd" or "netmask" was an issue either.
...

after having switched the two statements it now works again

Boot time increased noticeable, especially the gaps where the computer seemed to do nothing (but wasting time on executing the old shell-script?) between two init-scripts are now gone.

Ah! Great. Then if this works with brd or broadcast, but order is needed, this is likely a bug requiring documentation. (If they haven't already.)

Yes, the boot time activity you describe was what I paraphrased by stating quicker boot timing. (Those hangs add up to minutes here, especially on slower cpus/drives.)

Ditto on the /etc/hosts file. Caught this too but forgot to make note.

NOTE 4: Fries /etc/hosts file!

(Also, if you have cryptsetup-luks, luks is now merged into =>cryptsetup-1.5 and also requires a reboot before you can use it successfully. Also, re-add it's init script to the boot level ... 1.0.6 init script is only for baselayout-2.0.0. Not really related, but threw me for a spin last night.)

brot · Post by **brot** » Sun Apr 20, 2008 8:48 am

Runs very nice here on my boxes (~amd64 and ~x86). Thanks for the newsentry and the nice work with baselayout and openrc!

Mikkl · Post by **Mikkl** » Sun Apr 20, 2008 9:39 am

ah, i had a thinking error. actually i meant this the other way.

mikkl wrote:Boot time increased noticeable, ...

boot time decreased of course.

rogerx wrote:Ah! Great. Then if this works with brd or broadcast, but order is needed, ...

yes, it seems if brd or broadcast doesn't matter, works both for me.

rogerx · Post by **rogerx** » Mon Apr 21, 2008 3:49 am

LOL. Ditto. Dyslexia. :-/

The Unknown · Post by **The Unknown** » Mon Apr 21, 2008 5:41 am

Well I have to say, very nice!
I run stable systems but after seeing all the hullabaloo I had to check it out for myself.

The migration was absolutely painless. Nothing of mine nuked, updating the config's was trivial.
My boot time was only slight decrease, I already boot under 30 sec's so I wasn't expecting much. Although I could really care less, I will take any boot time for a modern and robust system. It feels good, and I'm happy to see things evolving and moving forward.

Great job on the guide

Great job on the new rc-system and thanks to Roy Marples and all those involved.

natros · Post by **natros** » Mon Apr 21, 2008 12:31 pm

from grub to kdm in just 15s, it's amazing
thank you

dentharg · Post by **dentharg** » Tue Apr 22, 2008 7:58 am

Thanks a lot! OpenRC is faster - from my previous boot time (since grub to kdm) I am now getting there almost instantly

Shamus397 · Post by **Shamus397** » Tue Apr 22, 2008 2:47 pm

It's no placebo effect, the boot time is much faster on my slow machines here.

Thanks guys!

Eckos · Post by **Eckos** » Tue Apr 22, 2008 3:09 pm

It booted a little slower. Maybe it was because it was saving cache or w/e

rogerx · Post by **rogerx** » Thu Apr 24, 2008 6:51 am

Shamus397 wrote:It's no placebo effect, the boot time is much faster on my slow machines here.

Thanks guys!

This is what I would figure. The speed increase would be less noticeable on faster machines and more noticeable on slower machines. However, most people owning the faster machines, are not too concerned about milliseconds.

Not only this, people willing to spend the bucks on faster machines, aren't too concerned about compiling with optimizations or worried about migrating code from shell scripting to C, etc. <shrugs, politics with computers>

rogerx · Post by **rogerx** » Thu Apr 24, 2008 6:58 am

Eckos wrote:It booted a little slower. Maybe it was because it was saving cache or w/e

If it's slower, might be a really good indicator of a bad/incompatible compiler option(s)?

Somethings definitely not right as C is faster then scripting. (Less CPU processing as C is a lower level language then Shell/Bash scripting.)

warrens · Post by **warrens** » Thu Apr 24, 2008 10:15 am

Ok, here are some hard numbers showing the difference between baselayout-1 and baselayout-2+OpenRC:

Baselayout-1 41.88 seconds lilo to kdm
Baselayout-2+OpenRC 38.41 seconds lilo to kdm

A difference of 3.47 seconds give or take a few hundredths of a second due to using a stop watch to measure time, no placebo effect here. ~x86

dnadesign · Post by **dnadesign** » Thu Apr 24, 2008 11:54 am

Ok, got baselayout-2 and openrc installed.

It's seems slightly faster, but not too much. The automatic migration went smooth, I had to edit only one file to preserve my custom settings. Kernel modules, rc settings, everything migrated nicely.

Parallel startup works nice and finally any eventual errors are displayed correctly (as in to the correct module, that failed, which in previous baselayout with parallel startup was a problem

).
As for the boot logging (into /var/log/rc.log), it works good and contains all the information seen at boot time.

Conclusion: Nice work!

d2_racing · Post by **d2_racing** » Thu Apr 24, 2008 12:15 pm

Does anyone knows when it will hit stable ?

SlackwareInAZ · Post by **SlackwareInAZ** » Thu Apr 24, 2008 5:43 pm

Migrated yesterday to OpenRC/Baselayout 2. No real difficulties once I read and followed the migration guide.

Running 2.6.25 gentoo-sources with ext3 on single / partition.

Conventional baselayout: 23secs to boot to init3 logon

OpenRC Series boot: 12 secs
Parallel boot: 11 secs.

So a very noticeable improvement in startup time. Didn't time shutdown time, but it's much quicker as well.

Good job guys!

Gary

rogerx · Post by **rogerx** » Fri Apr 25, 2008 8:48 am

Ha!

Love the new openrc /etc/rc.conf file!
dmesg_level="1"
(Maybe the option was thar' before)

Also, consolefont now seems to use UTF-8 font without any issues now.

Dagger · Post by **Dagger** » Fri Apr 25, 2008 2:21 pm

I've been using Baselayout2 since early beta days, and OpenRC since uber decided to make it. Except few times when it screwed my pc (advantages of using GIT ebuild from dev's overlay) it's been working VERY stable and makes hell of the lot of difference. I absolutely love it and never had any problems with it. Also every bug I posted on Roy's bugzilla was answered/explained instantly and every question was answered with all the details.

btw, how many of you enabled rc_parallel="YES"?

d2_racing · Post by **d2_racing** » Fri Apr 25, 2008 2:23 pm

This option seems to split the boot time in half... impressive

Paapaa · Post by **Paapaa** » Fri Apr 25, 2008 2:27 pm

Stupid question, but:

Will my old baselayout get updated automatically when baselayout-2 reaches stable?

[NEWS] OpenRC and baselayout 2 will reach ~arch soon

Re: [NEWS] OpenRC and baselayout 2 will reach ~arch soon