Hello all,
I just had to find out that reboot does not work as expected when using sysvinit. Instead the shutdown procedure hangs after “localfs” down. Using systemd reboot works, but I do not want to use systemd at all. Who broke sysvinit procedure in favor of systemd ?
The problem arises on all HP servers we checked. We have to use a stock kernel.org kernel for the configs (in case that matters)…
On 2012-04-20 19:26, skraw wrote:
>
> Hello all,
>
> I just had to find out that reboot does not work as expected when using
> sysvinit. Instead the shutdown procedure hangs after “localfs” down.
> Using systemd reboot works, but I do not want to use systemd at all. Who
> broke sysvinit procedure in favor of systemd ?
> The problem arises on all HP servers we checked. We have to use a stock
> kernel.org kernel for the configs (in case that matters)…
As far as I know, it works for many people, so you must have something
peculiar.
–
Cheers / Saludos,
Carlos E. R.
(from 11.4 x86_64 “Celadon” at Telcontar)
So, I would suggest you try and stick with systemd because it will continue with all future versions and you will be trying to swim upstream trying to not use it.
Here is the announcement on the use of systemd: openSUSE News
I have a few utilts that can be helpful with systemd usage: systemd and using the after.local script in openSUSE 12.1 - Blogs - openSUSE Forums
And this might be helpful as well: DKMS, systemd & Virtual Box - How to get Dynamic Kernel Module Support to work in openSUSE 12.1 - Blogs - openSUSE Forums
Thank You,
Thank you, but I would suggest the rest of my fellow german SuSE people not to be supersmart about what people really do with the distro. It might be nice for users with desktops booting every other hour if the boot process looks short and fast. But for people interested in servers not (remote) booting over months it is a lot better to have a sequential boot process with significant ascii output about what is going on.
So systemd is and will be for desktop people, but not for me. And that is why sysvinit is highly preferred. And if it really is not seen as a bug that sysvinit is broken by something trivial like installing opensuse from scratch and then replacing the kernel with a stock kernel.org one, then probably they managed to disappoint the last server-interested opensuse users…
On 2012-04-21 14:46, skraw wrote:
> So systemd is and will be for desktop people, but not for me. And that
> is why sysvinit is highly preferred.
No matter what you think or prefer, systemd is here to stay, and systemv
will be removed, eventually. You can not do anything about it. This is not
a democracy.
And yes, I do prefer systemv.
–
Cheers / Saludos,
Carlos E. R.
(from 11.4 x86_64 “Celadon” at Telcontar)
So I will be the first to admit that not all change is embraced. When we get something working, we don’t like to try something new. It appears that systemd has a German origin in that its creator Lennart Poettering is German and also created PulseAudio as I understand it. And the heart of openSUSE is also German located in Nürnberg. I think that German interest are being served by openSUSE very well which does not mean we all must agree to every change being made. Further, SuSE has Linux versions such as SuSE Linux Enterprise Server (SLES) designed and written to work like you are working now. On the other hand, openSUSE is designed and written pointing to the future of Linux and embracing everything that is new for Linux and including all the latest “stuff”. In my mind, you must learn to use it, such as systemd, or point out and make bug reports for fixes in openSUSE 12.2 our next OS, or find solutions such as my after.local blog post. In addition, its possible to continue using older Linux versions when you know how to upgrade the kernel such as you are doing. I hope you can find it to continue with and help promote openSUSE and we can use all the help we can get as we continue going forward. If you should like some help in kernel installation, have a look at these two blogs on that subject:
S.A.K.C. - SUSE Automated Kernel Compiler - Version 2.62 - Blogs - openSUSE Forums
AND
S.G.T.B. - SuSE Git Kernel Tarball Creator - Version 1.80 - Blogs - openSUSE Forums
Please have a look at my blogs as I have spent a lot of effort in making this work as well as I can with openSUSE. Good luck in what ever you decide to do.
Thank You (for using openSUSE),
One idea that is fundamentally broken throughout all distros is the splitup between the “driving” standard (desktop-thought) release and the “server-release”. The simple fact of life is that the long-term idea behind the server releases of almost all distros (not only SuSE) is completely away from what is going on. Most servers are somehow linked to the LAMP idea. and since this is only a small subset of packages there should be no problem to incorporate them in the latest standard releases. Most static setups cannot be updated anyway for incompatibility reasons between older and newer stuff (php is the example per se).
On the other hand it looks obvious from your “there shall be light” wording that another fundamental principle has completely vanished from the distro decision makers: the word is “option”.
There is absolutely no good reason why systemd and sysvinit should not live together on one distro - besides the fact that someone with minor vision does not want that, whoever that may be.
There is nothing such as “optimal”, there is only “optional”.
On Sat, 21 Apr 2012 20:06:02 +0530, skraw
<skraw@no-mx.forums.opensuse.org> wrote:
> There is absolutely no good reason why systemd and sysvinit should not
> live together on one distro
there isn’t, except that there aren’t enough people to do both,
apparently. if someone came forward offering to maintain sysvinit in the
future, i don’t think there would be any objections.
–
phani.
On 2012-04-21 16:43, phanisvara das wrote:
> On Sat, 21 Apr 2012 20:06:02 +0530, skraw <skraw@no-mx.forums.opensuse.org>
> wrote:
>
>> There is absolutely no good reason why systemd and sysvinit should not
>> live together on one distro
>
> there isn’t, except that there aren’t enough people to do both, apparently.
> if someone came forward offering to maintain sysvinit in the future, i
> don’t think there would be any objections.
There are.
If you follow the factory mail list, you will see that they intend to
remove systemv as soon as they can, and not have to keep compatibility with
systemd. It was very difficult to convince them to keep both for 12.2. 
–
Cheers / Saludos,
Carlos E. R.
(from 11.4 x86_64 “Celadon” at Telcontar)
On 04/21/2012 01:23 PM, Carlos E. R. wrote:
> On 2012-04-21 16:43, phanisvara das wrote:
>> On Sat, 21 Apr 2012 20:06:02 +0530, skraw<skraw@no-mx.forums.opensuse.org>
>> wrote:
>>
>>> There is absolutely no good reason why systemd and sysvinit should not
>>> live together on one distro
>>
>> there isn’t, except that there aren’t enough people to do both, apparently.
>> if someone came forward offering to maintain sysvinit in the future, i
>> don’t think there would be any objections.
>
> There are.
>
> If you follow the factory mail list, you will see that they intend to
> remove systemv as soon as they can, and not have to keep compatibility with
> systemd. It was very difficult to convince them to keep both for 12.2.
Yes, but the main reason was a lack of developers to maintain SystemV, which is
the same reason for switching from GRUB to GRUB2.
If you have troubles with systemd, it is critical to document them and file the
appropriate bug reports.
On Sat, 21 Apr 2012 23:53:06 +0530, Carlos E. R.
<robin_listas@no-mx.forums.opensuse.org> wrote:
> On 2012-04-21 16:43, phanisvara das wrote:
>> On Sat, 21 Apr 2012 20:06:02 +0530, skraw
>> <skraw@no-mx.forums.opensuse.org>
>> wrote:
>>
>>> There is absolutely no good reason why systemd and sysvinit should not
>>> live together on one distro
>>
>> there isn’t, except that there aren’t enough people to do both,
>> apparently.
>> if someone came forward offering to maintain sysvinit in the future, i
>> don’t think there would be any objections.
>
> There are.
>
> If you follow the factory mail list, you will see that they intend to
> remove systemv as soon as they can, and not have to keep compatibility
> with
> systemd. It was very difficult to convince them to keep both for 12.2.
>
>
i did follow that discussion – admittedly not with my full attention –
and am still under the impression that the reason for giving up sysvinit
as soon as possible, i.e., once everything works, presumably, was or is
that there isn’t enough (wo)men power to maintain both indefinitely. i do
remember slightly heated exchanges re. the wisdom of doing it at all
(switching to systemd), and the time for which both have to be maintained
in parallel, but i also remember reading that people were welcome to
continue maintaining sysvinit if they could.
–
phani.
On 2012-04-21 20:53, Larry Finger wrote:
> On 04/21/2012 01:23 PM, Carlos E. R. wrote:
> Yes, but the main reason was a lack of developers to maintain SystemV,
> which is the same reason for switching from GRUB to GRUB2.
No, the same devs that are creating systemd, could instead maintain
systemv, with less effort probably. Usually developers prefer to create
their own thing instead of maintaining something that was created time ago
by somebody else. And I say so because I’m a developer when they pay me for it.
> If you have troubles with systemd, it is critical to document them and file
> the appropriate bug reports.
This is what I have been saying many times, and I have reported and tracked
several systemd bugs - even though I’m staying on 11.4, I have two test
installs to check on it.
–
Cheers / Saludos,
Carlos E. R.
(from 11.4 x86_64 “Celadon” at Telcontar)
For completeness of the thread I have tracked down the original problem. K04boot.localfs has problems.
First, around line 372 you have to add /media with continue as tmpfs does not like to be umounted -rv. Since all other tmpfs are ignored it seems valid to ignore /media, too.
Second, it looks like /boot (if you happen to do /boot partitions like me) seems to have troubles with umount -rv, too at this stage. Fact is the umount hangs, line 430. This looks more like a bug to me as /boot can be umounted during normal operation without problems.
If you solve those, reboot works again.
Now for systemd. Its true problem is its pure concept. It is completely bogus in setups that are booted once a year to try to gain a speedup by paralleling the boot process only to loose the advantage of having step-by-step debugging at hand. The RUN_PARALLEL hack with setpar inside sysvinit procedure btw should be eliminated, too.
Overall systemd violates one of the all-time true words from Big L: keep it simple.
It’s fast, it’s a hacker playground, it’s completely intransparent to an ordinary user, it’s probably everything an admin likes, but it’s really not simple and on servers its exactly zero useful.