On Sun, 2011-05-29 at 13:40 +0200, Bruno Friedmann wrote:

On 05/29/2011 09:55 AM, Tim wrote:

...

to my knowledge systemd supposed to be the default init daemon in 12.1. Doesn't it make sense to activate it now (or as early as possible) to get at least some feedback from the people running factory?

Tim I'm also in favor of that, but before becoming the default systemd has to support fully lvm and encrypted volume group

Systemd only supports encrypted LUKS volumes out-of-the-box. I personally never used LVM. The SUSE init scripts might need integration to work properly with systemd. I can't really tell what's needed here. I know Fedora needed some hacks to their init scripts/service files to make stacked LVM/crypto configs work.

(particulary in shutdown process)

That should work already. We should cleanup all DM devices.

Having fsck forced at each reboot is a pain.

That sounds weird. I've never seen that. Maybe it's some init script going crazy. If you use a recent systemd version and can provide some output, please open a bug.

Then make sure yast systemlevel can work with systemd if it is present.

Yeah, many things need integration still.

I think Kay can comment on what is the actual status of integration, and the specific help he needs

It's a lot of work ahead. We really need to dedicate resources to it, means people working full time on it, otherwise it would leave too many things in unknown state. Also people finally need to make hard decisions, I fear, otherwise we will run into a test matrix of alternative options to support, which we can not really handle. It works fine for me, but I use only a very tiny subset of the tings that need to work in the distro. LVM, nfs, iscsi, unlocking of encrypted SSL certificates, openvpn server, ..., all the 'non-laptop things' I never used with systemd. Some things are mentioned here, I need to update it the next days to reflect the current state: http://en.opensuse.org/openSUSE:Systemd_status Kay -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

On 2011/05/29 19:29 (GMT+0200) Markus Slopianka composed:

Bruno Friedmann composed:

...

Tim I'm also in favor of that, but before becoming the default systemd has to support fully ...

For a release: yes. But Factory isn't meant to be stable anyway, so a few hickups here and there are tolerable in Factory and would help not only exposing missing aspects but also would help to not put integration work on the back burner. ;-)

I agree with Kay's assessment. My opinion of the Novell staff cutbacks during the 11.3 cycle is that little was left but a maintenance skeleton unable to fully cope with such >minimal changes as KMS, Grub2 & systemd without giving up the non-bleeding-edge reliability most openSUSE users expect even after the death of SaX2 that shortly followed in its wake. I don't believe required resources are or will be made available to make a systemd-dependent release ready with the scheduled release of 12.1, especially if Grub2 also gets scheduled for full support. After seeing what Fedora has been going through, I don't think anything short of a systemd-only fork with freedom to take as long as necessary could ever result in an openSUSE release that won't make an unreasonable number of upgraders wish they hadn't, or jump ship to something that has retained the sysvinit they know works. I'd much rather a boot that takes up to 3 minutes than a quick-to-boot system incapable of doing all the many things that had been working fine for many years, and already as of 11.3 don't. Newer != better. Time spent dealing with or hampered by brokenness, and grasping & retaining an entirely new set of 12-20+ character long system control commands, far outweighs mere seconds gained at boot time. -- "The wise are known for their understanding, and pleasant words are persuasive." Proverbs 16:21 (New Living Translation) Team OS/2 ** Reg. Linux User #211409 ** a11y rocks! Felix Miata *** http://fm.no-ip.com/ -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On 2011-05-29 22:25, Roger Luedecke wrote:

On Sunday, May 29, 2011 11:54:50 AM Felix Miata wrote:

...

I'd much rather a boot that takes up to 3 minutes than a quick-to-boot system incapable of doing all the many things that had been working fine for many years, and already as of 11.3 don't.

Newer != better. Time spent dealing with or hampered by brokenness, and grasping & retaining an entirely new set of 12-20+ character long system control commands, far outweighs mere seconds gained at boot time.

I completely and absolutely agree. A slightly faster boot time is not worth problems with simply booting. And besides, there are other areas to focus development in.

+1 - -- Cheers / Saludos, Carlos E. R. (from 11.2 x86_64 "Emerald" at Telcontar) -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.12 (GNU/Linux) Comment: Using GnuPG with SUSE - http://enigmail.mozdev.org/ iEYEARECAAYFAk3i9OoACgkQtTMYHG2NR9UOaACeIynVIxEaXKcfOWfVlub+ptAP EbgAn3KDzK6KqsnxQrb167tySgLXESgt =DonD -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On 2011-05-30 04:02, Greg KH wrote:

On Mon, May 30, 2011 at 03:37:46AM +0200, Carlos E. R. wrote:

It's not _just_ a matter of booting faster at all. I suggest everyone take a look at the "systemd for administrators" series of posts that describe exactly why we want and need systemd on our systems.

I searched for a subject line of that name and did not find it. Also searched 7500 bodies, no hit. Is it on another list? Another subject line? Could you provide a link? - -- Cheers / Saludos, Carlos E. R. (from 11.2 x86_64 "Emerald" at Telcontar) -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.12 (GNU/Linux) Comment: Using GnuPG with SUSE - http://enigmail.mozdev.org/ iEYEARECAAYFAk3j5RwACgkQtTMYHG2NR9XdZgCeNZ7DuHMt0z8lzikN14l0gAA0 PjsAoIM5W+yHDNJBi06NsNi4p5J2ExIS =gK8Q -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On 2011-05-30 20:47, Dimstar / Dominique Leuenberger wrote:

On Mon, 2011-05-30 at 20:42 +0200, Carlos E. R. wrote:

...

On 2011-05-30 04:02, Greg KH wrote:

...

On Mon, May 30, 2011 at 03:37:46AM +0200, Carlos E. R. wrote:

...

It's not _just_ a matter of booting faster at all. I suggest everyone take a look at the "systemd for administrators" series of posts that describe exactly why we want and need systemd on our systems.

I searched for a subject line of that name and did not find it. Also searched 7500 bodies, no hit. Is it on another list? Another subject line? Could you provide a link?

You forgot to ask google:

http://goo.gl/LwSdY

It would have brought you to some nice blog posts about the topic The series starts at http://0pointer.de/blog/projects/systemd-for-admins-1.html

He said it was “a series of posts”, so I have to assume he posted it here. I never imagined he meant an outside source. How else were we to know that systemd is something else than speed? That blog is not Greg's, so I have to wonder if that is indeed what he wanted us to read? - -- Cheers / Saludos, Carlos E. R. (from 11.2 x86_64 "Emerald" at Telcontar) -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.12 (GNU/Linux) Comment: Using GnuPG with SUSE - http://enigmail.mozdev.org/ iEYEARECAAYFAk3kD1QACgkQtTMYHG2NR9UdqwCfaDbQCzPz6cl0dJQIgpOXkHcP +AYAn05Os/wPjlQ6pS3buGMKqTzYOTY/ =Xyp/ -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

Am Sun, 29 May 2011 13:25:49 -0700 schrieb Roger Luedecke :

I completely and absolutely agree. A slightly faster boot time is not worth problems with simply booting.

JFTR: systemd is much much more than only "faster booting". Faster booting (if it is even booting faster) is only a side effect of systemd. -- Stefan Seyfried "Dispatch war rocket Ajax to bring back his body!" -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

On Mon, May 30, 2011 at 09:05:22AM +0200, Tim wrote:

...

Am Sonntag 29 Mai 2011, 13:40:08 schrieb Bruno Friedmann: For a release: yes. But Factory isn't meant to be stable anyway, so a few hickups here and there are tolerable in Factory and would help not only exposing missing aspects but also would help to not put integration work on the back burner. ;-) Exactly, that's what I meant. It is better to make it default now and switch back if things doesn't work out than to activate it later on with less testing. I mean factory users are most likely not representative for the user

On 05/29/2011 07:29 PM, Markus Slopianka wrote: base but in case of testing more is always better I suppose. :)

Factory should be stable though, otherwise it is kind of useless. Ciao, Marcus -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

Am Montag, 30. Mai 2011 schrieb Stefan Seyfried:

Am Mon, 30 May 2011 11:51:52 +0200

schrieb Marcus Meissner :

...

Factory should be stable though, otherwise it is kind of useless.

You're confusing FACTORY with Tumbleweed?

I mean, how could we explain "commit half of GNOME3 to Factory, and let it stay broken for 3 weeks" with "Factory should be stable"?

Not sure your "jokes" help in any way to get this done quicker. The problem from what I see is not that only half of GNOME3 was commited, but that GNOME3 is only half[1] of GNOME2 and many things aren't ported yet. Greetings, Stephan [1] Where "half" is a strong exaggeration. -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

On Mon, May 30, 2011 at 09:20:50AM +0200, Jiri Slaby wrote:

On 05/29/2011 09:55 AM, Tim wrote:

...

Hi,

to my knowledge systemd supposed to be the default init daemon in 12.1. Doesn't it make sense to activate it now (or as early as possible) to get at least some feedback from the people running factory?

While we are at it: Greg, could we have the latest systemd (in factory there is 26) in TW? I'm using systemd in 11.4 as the default, but it is too buggy. The bugs were mostly resolved in later versions. (The most annoying for me is 671673.)

I'll be glad to add it, but shouldn't I also add other packages as well to reflect the changes needed by 26? udev at the least I imagine, anything else? thanks, greg k-h -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

On 05/30/2011 09:37 AM, Greg KH wrote:

On Mon, May 30, 2011 at 09:20:50AM +0200, Jiri Slaby wrote:

...

On 05/29/2011 09:55 AM, Tim wrote:

...

Hi,

to my knowledge systemd supposed to be the default init daemon in 12.1. Doesn't it make sense to activate it now (or as early as possible) to get at least some feedback from the people running factory?

While we are at it: Greg, could we have the latest systemd (in factory there is 26) in TW? I'm using systemd in 11.4 as the default, but it is too buggy. The bugs were mostly resolved in later versions. (The most annoying for me is 671673.)

I'll be glad to add it, but shouldn't I also add other packages as well to reflect the changes needed by 26? udev at the least I imagine, anything else?

Yeah, you are correct, let me figure out and test. thanks, -- js suse labs -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

On Mon, May 30, 2011 at 10:48:08AM +0200, Stephan Kulow wrote:

Am Montag, 30. Mai 2011 schrieb Jiri Slaby:

...

On 05/30/2011 09:37 AM, Greg KH wrote:

...

On Mon, May 30, 2011 at 09:20:50AM +0200, Jiri Slaby wrote:

...

On 05/29/2011 09:55 AM, Tim wrote:

...

Hi,

to my knowledge systemd supposed to be the default init daemon in 12.1. Doesn't it make sense to activate it now (or as early as possible) to get at least some feedback from the people running factory?

While we are at it: Greg, could we have the latest systemd (in factory there is 26) in TW? I'm using systemd in 11.4 as the default, but it is too buggy. The bugs were mostly resolved in later versions. (The most annoying for me is 671673.)

I'll be glad to add it, but shouldn't I also add other packages as well to reflect the changes needed by 26? udev at the least I imagine, anything else?

Yeah, you are correct, let me figure out and test.

systemd conflicts with 11.4's filesystem.rpm and mkinitrd.rpm for support for /run.

Ok, then we can just add those to Tumbleweed as well, upgrading should be fine with that, right? thanks, greg k-h -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

On Mon, 2011-05-30 at 17:29 +0800, Greg KH wrote:

On Mon, May 30, 2011 at 10:48:08AM +0200, Stephan Kulow wrote:

...

Am Montag, 30. Mai 2011 schrieb Jiri Slaby:

...

On 05/30/2011 09:37 AM, Greg KH wrote:

...

On Mon, May 30, 2011 at 09:20:50AM +0200, Jiri Slaby wrote:

...

On 05/29/2011 09:55 AM, Tim wrote:

...

Hi,

to my knowledge systemd supposed to be the default init daemon in 12.1. Doesn't it make sense to activate it now (or as early as possible) to get at least some feedback from the people running factory?

While we are at it: Greg, could we have the latest systemd (in factory there is 26) in TW? I'm using systemd in 11.4 as the default, but it is too buggy. The bugs were mostly resolved in later versions. (The most annoying for me is 671673.)

I'll be glad to add it, but shouldn't I also add other packages as well to reflect the changes needed by 26? udev at the least I imagine, anything else?

Yeah, you are correct, let me figure out and test.

systemd conflicts with 11.4's filesystem.rpm and mkinitrd.rpm for support for /run.

Ok, then we can just add those to Tumbleweed as well, upgrading should be fine with that, right?

It should be udev, mkinitrd, aaa_base, filesystem. Kay -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

On 05/30/2011 03:30 PM, Jiri Slaby wrote:

On 05/30/2011 12:54 PM, Kay Sievers wrote:

...

On Mon, 2011-05-30 at 17:29 +0800, Greg KH wrote:

...

On Mon, May 30, 2011 at 10:48:08AM +0200, Stephan Kulow wrote:

...

systemd conflicts with 11.4's filesystem.rpm and mkinitrd.rpm for support for /run.

Ok, then we can just add those to Tumbleweed as well, upgrading should be fine with that, right?

It should be udev, mkinitrd, aaa_base, filesystem.

Not really. I ended up with: https://build.opensuse.org/project/monitor?project=home%3Ajirislaby%3Atumble...

I'll test the build results later.

Tested a bit, I see no problems at all. So what should I do now :)? Link to TW testing, update the scripts at github and request a pull? thanks, -- js suse labs -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

On 05/31/2011 11:24 PM, Greg KH wrote:

But yes, a github request would be best,

Done.

I could get it building in the openSUSE:Tumbleweed:Testing repo first so I can test it when I return on Monday.

You can also test it by adding home:jirislaby:tumbleweed to your repos and dup. thanks, -- js suse labs -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

On 05/30/2011 10:29 AM, Frederic Crozat wrote:

Le dimanche 29 mai 2011 à 09:55 +0200, Tim a écrit :

...

Hi,

to my knowledge systemd supposed to be the default init daemon in 12.1. Doesn't it make sense to activate it now (or as early as possible) to get at least some feedback from the people running factory?

Damn, you're cutting all the surprise effect ;)

I was planning to announce I'll be working full-time for June month, starting next week, on systemd integration in Factory.

Of course, feedback from people running Factory and help from packagers will be extremely useful, to get the best possible experience in 12.1 for systemd.

More on this next week ;)

My actual result ( tested 15 minutes ago ) are quite bad. with systemd system-sysinit and networkmanager activated I just get no network kde refuse to start due to dbus error and a lot of others services. cat /var/log/messages | grep -i systemd May 30 10:38:11 c-3po kernel: [ 27.616858] systemd[1]: systemd 26 running in system mode. (+PAM +LIBWRAP +AUDIT +SELINUX +SYSVINIT +LIBCRYPTSETUP; suse) May 30 10:38:11 c-3po kernel: [ 27.659579] systemd[1]: Set hostname to . May 30 10:38:11 c-3po kernel: [ 28.413789] systemd[1]: [/etc/init.d/postfix:12] Failed to add LSB Provides name sendmail.service, ignoring: File exists May 30 10:38:11 c-3po kernel: [ 28.462158] systemd[1]: Found ordering cycle on basic.target/start May 30 10:38:11 c-3po kernel: [ 28.462164] systemd[1]: Walked on cycle path to sockets.target/start May 30 10:38:11 c-3po kernel: [ 28.462169] systemd[1]: Walked on cycle path to dbus.socket/start May 30 10:38:11 c-3po kernel: [ 28.462173] systemd[1]: Walked on cycle path to sysinit.target/start May 30 10:38:11 c-3po kernel: [ 28.462177] systemd[1]: Walked on cycle path to sysstat.service/start May 30 10:38:11 c-3po kernel: [ 28.462182] systemd[1]: Walked on cycle path to remote-fs.target/start May 30 10:38:11 c-3po kernel: [ 28.462186] systemd[1]: Walked on cycle path to network.target/start May 30 10:38:11 c-3po kernel: [ 28.462190] systemd[1]: Walked on cycle path to network.service/start May 30 10:38:11 c-3po kernel: [ 28.462194] systemd[1]: Walked on cycle path to SuSEfirewall2_init.service/start May 30 10:38:11 c-3po kernel: [ 28.462198] systemd[1]: Walked on cycle path to basic.target/start May 30 10:38:11 c-3po kernel: [ 28.462203] systemd[1]: Breaking ordering cycle by deleting job dbus.socket/start May 30 10:38:11 c-3po kernel: [ 28.462218] systemd[1]: Found ordering cycle on basic.target/start May 30 10:38:11 c-3po kernel: [ 28.462222] systemd[1]: Walked on cycle path to sysinit.target/start May 30 10:38:11 c-3po kernel: [ 28.462226] systemd[1]: Walked on cycle path to sysstat.service/start May 30 10:38:11 c-3po kernel: [ 28.462230] systemd[1]: Walked on cycle path to remote-fs.target/start May 30 10:38:11 c-3po kernel: [ 28.462234] systemd[1]: Walked on cycle path to network.target/start May 30 10:38:11 c-3po kernel: [ 28.462239] systemd[1]: Walked on cycle path to network.service/start May 30 10:38:11 c-3po kernel: [ 28.462243] systemd[1]: Walked on cycle path to SuSEfirewall2_init.service/start May 30 10:38:11 c-3po kernel: [ 28.462247] systemd[1]: Walked on cycle path to basic.target/start May 30 10:38:11 c-3po kernel: [ 28.462252] systemd[1]: Breaking ordering cycle by deleting job sysstat.service/start May 30 10:38:11 c-3po kernel: [ 28.462278] systemd[1]: Found ordering cycle on basic.target/start May 30 10:38:11 c-3po kernel: [ 28.462283] systemd[1]: Walked on cycle path to sysinit.target/start May 30 10:38:11 c-3po kernel: [ 28.462287] systemd[1]: Walked on cycle path to cpufreq.service/start May 30 10:38:11 c-3po kernel: [ 28.462291] systemd[1]: Walked on cycle path to basic.target/start May 30 10:38:11 c-3po kernel: [ 28.462295] systemd[1]: Breaking ordering cycle by deleting job cpufreq.service/start May 30 10:38:11 c-3po kernel: [ 29.572210] systemd-fsck[530]: Root directory is writable, skipping check. May 30 10:38:11 c-3po kernel: [ 31.054692] systemd-fsck[741]: srv : propre, 29329/6553600 fichiers, 8240453/26214400 blocs May 30 10:38:11 c-3po kernel: [ 31.055490] systemd-fsck[747]: home : propre, 738766/16777216 fichiers, 51330223/67108864 blocs May 30 10:38:11 c-3po kernel: [ 32.223395] systemd-cryptsetup[855]: Encountered unknown /etc/crypttab option 'none', ignoring. May 30 10:38:11 c-3po kernel: [ 32.224951] systemd-cryptsetup[855]: Volume cr_sda2 already active. May 30 10:38:11 c-3po kernel: [ 32.281183] systemd-fsck[864]: boot : propre, 63/66264 fichiers, 104813/264192 blocs May 30 10:38:11 c-3po kernel: [ 38.410953] systemd[1]: network.service: control process exited, code=exited status=2 May 30 10:38:11 c-3po kernel: [ 38.411091] systemd[1]: Unit network.service entered failed state. May 30 10:38:12 c-3po kernel: [ 39.604429] systemd[1]: nscd.service: control process exited, code=exited status=7 May 30 10:38:12 c-3po kernel: [ 39.604438] systemd[1]: Unit nscd.service entered failed state. May 30 10:38:55 c-3po kernel: [ 83.208881] systemd[1]: network.service: control process exited, code=exited status=2 May 30 10:38:55 c-3po systemd[1]: Unit network.service entered failed state. May 30 10:39:15 c-3po kernel: [ 102.540229] systemd[1]: postgresql.service: control process exited, code=exited status=1 May 30 10:39:15 c-3po kernel: [ 102.548159] systemd[1]: Unit postgresql.service entered failed state. May 30 10:41:41 c-3po systemd[1]: Reloading. May 30 10:41:42 c-3po systemd[1]: [/etc/init.d/postfix:12] Failed to add LSB Provides name sendmail.service, ignoring: File exists May 30 10:43:33 c-3po systemd[1]: dbus.service: main process exited, code=exited, status=1 May 30 10:43:33 c-3po kernel: [ 360.961086] systemd[1]: Unit dbus.service entered failed state. May 30 10:46:42 c-3po systemd[1]: apache2.service operation timed out. Terminating. May 30 10:46:42 c-3po kernel: [ 549.329395] systemd[1]: kbd.service operation timed out. Terminating. May 30 10:46:42 c-3po kernel: [ 549.348341] systemd[1]: Unit kbd.service entered failed state. May 30 10:46:42 c-3po kernel: [ 549.354371] systemd[1]: Unit apache2.service entered failed state. May 30 10:46:42 c-3po systemd[1]: Startup finished in 27s 675ms 424us (kernel) + 8min 41s 768ms 674us (userspace) = 9min 9s 444ms 98us. May 30 10:49:11 c-3po systemd-initctl[11103]: Got request for unknown runlevel u, ignoring. May 30 10:49:26 c-3po systemd-fsck[11135]: Root directory is writable, skipping check. May 30 10:49:30 c-3po systemd-initctl[11232]: Received environment initctl request. This is not implemented in systemd. May 30 10:49:30 c-3po systemd[1]: Found ordering cycle on sysinit.target/start May 30 10:49:30 c-3po systemd[1]: Walked on cycle path to sysstat.service/start May 30 10:49:30 c-3po systemd[1]: Walked on cycle path to remote-fs.target/stop May 30 10:49:30 c-3po systemd[1]: Walked on cycle path to network.target/stop May 30 10:49:30 c-3po systemd[1]: Walked on cycle path to network.service/stop May 30 10:49:30 c-3po systemd[1]: Walked on cycle path to SuSEfirewall2_init.service/stop May 30 10:49:30 c-3po systemd[1]: Walked on cycle path to basic.target/start May 30 10:49:30 c-3po systemd[1]: Walked on cycle path to sockets.target/start May 30 10:49:30 c-3po systemd[1]: Walked on cycle path to dbus.socket/start May 30 10:49:30 c-3po systemd[1]: Walked on cycle path to sysinit.target/start May 30 10:49:30 c-3po systemd[1]: Breaking ordering cycle by deleting job sysstat.service/start May 30 10:49:30 c-3po kernel: [ 717.356990] systemd[1]: Found ordering cycle on sysinit.target/start May 30 10:49:30 c-3po kernel: [ 717.357017] systemd[1]: Walked on cycle path to cpufreq.service/start May 30 10:49:30 c-3po kernel: [ 717.357025] systemd[1]: Walked on cycle path to basic.target/start May 30 10:49:30 c-3po kernel: [ 717.357032] systemd[1]: Walked on cycle path to sockets.target/start May 30 10:49:30 c-3po kernel: [ 717.357038] systemd[1]: Walked on cycle path to dbus.socket/start May 30 10:49:30 c-3po kernel: [ 717.357044] systemd[1]: Walked on cycle path to sysinit.target/start May 30 10:49:30 c-3po kernel: [ 717.357050] systemd[1]: Breaking ordering cycle by deleting job cpufreq.service/start May 30 10:49:30 c-3po kernel: [ 717.666527] systemd[1]: sshd.service: main process exited, code=exited, status=255 May 30 10:49:30 c-3po kernel: [ 717.819091] systemd[1]: Unit sshd.service entered failed state. mainly everything coming from that kind of ay 30 10:38:11 c-3po kernel: [ 38.465793] NetworkManager[1137]: <error> [1306744691.184409] [nm-dbus-manager.c:278] nm_dbus_manager_init_bus(): Could not get the system bus. Make sure the message bus daemon is running! Message: Failed to connect to socket /var/run/dbus/system_bus_socket: No such file or directory I know my factory is a constant rolling release started last year in september. Kay, Frederic should we open a king of meta-bug for systemd that we let open until we success it's integration ? -- Bruno Friedmann Ioda-Net Sàrl www.ioda-net.ch openSUSE Member & Ambassador GPG KEY : D5C9B751C4653227 irc: tigerfoot -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

On Mon, 2011-05-30 at 11:17 +0200, Bruno Friedmann wrote:

My actual result ( tested 15 minutes ago ) are quite bad. with systemd system-sysinit and networkmanager activated I just get no network kde refuse to start due to dbus error and a lot of others services.

cat /var/log/messages | grep -i systemd May 30 10:38:11 c-3po kernel: [ 27.616858] systemd[1]: systemd 26 running in system mode. (+PAM +LIBWRAP +AUDIT +SELINUX +SYSVINIT +LIBCRYPTSETUP; suse)

May 30 10:38:11 c-3po kernel: [ 28.462252] systemd[1]: Breaking ordering cycle by deleting job sysstat.service/start

Can you try to remove or disable sysstat from the bootup? It might have dependencies in its headers with can't be resolved by systemd.

May 30 10:38:11 c-3po kernel: [ 28.462295] systemd[1]: Breaking ordering cycle by deleting job cpufreq.service/start

Care to remove cpufreq. Nothing should fiddle around with CPU governors these days. The kernel's on-demand governor is the only sensible thing to do.

May 30 10:38:11 c-3po kernel: [ 32.223395] systemd-cryptsetup[855]: Encountered unknown /etc/crypttab option 'none', ignoring.

Care to check what 'none' means to express in that file, and if that can just be removed? Did you, or Yast added it?

May 30 10:38:11 c-3po kernel: [ 32.224951] systemd-cryptsetup[855]: Volume cr_sda2 already active. May 30 10:38:11 c-3po kernel: [ 32.281183] systemd-fsck[864]: boot : propre, 63/66264 fichiers, 104813/264192 blocs May 30 10:38:11 c-3po kernel: [ 38.410953] systemd[1]: network.service: control process exited, code=exited status=2 May 30 10:38:11 c-3po kernel: [ 38.411091] systemd[1]: Unit network.service entered failed state. May 30 10:38:12 c-3po kernel: [ 39.604429] systemd[1]: nscd.service: control process exited, code=exited status=7 May 30 10:38:12 c-3po kernel: [ 39.604438] systemd[1]: Unit nscd.service entered failed state. May 30 10:38:55 c-3po kernel: [ 83.208881] systemd[1]: network.service: control process exited, code=exited status=2 May 30 10:38:55 c-3po systemd[1]: Unit network.service entered failed state. May 30 10:39:15 c-3po kernel: [ 102.540229] systemd[1]: postgresql.service: control process exited, code=exited status=1 May 30 10:39:15 c-3po kernel: [ 102.548159] systemd[1]: Unit postgresql.service entered failed state. May 30 10:41:41 c-3po systemd[1]: Reloading. May 30 10:41:42 c-3po systemd[1]: [/etc/init.d/postfix:12] Failed to add LSB Provides name sendmail.service, ignoring: File exists May 30 10:43:33 c-3po systemd[1]: dbus.service: main process exited, code=exited, status=1 May 30 10:43:33 c-3po kernel: [ 360.961086] systemd[1]: Unit dbus.service entered failed state.

The unresolvable/cyclic dependencies cause unfortunately that D-Bus does not start properly, which will let NetworkManager and a lot of other stuff fail. Kay -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

On 05/30/2011 04:29 PM, Carlos E. R. wrote:

On 2011-05-30 15:26, Kay Sievers wrote:

...

...

May 30 10:38:11 c-3po kernel: [ 28.462295] systemd[1]: Breaking ordering cycle by deleting job cpufreq.service/start

...

Care to remove cpufreq. Nothing should fiddle around with CPU governors these days. The kernel's on-demand governor is the only sensible thing to do.

Bug 653540

...

...

May 30 10:38:11 c-3po kernel: [ 32.223395] systemd-cryptsetup[855]: Encountered unknown /etc/crypttab option 'none', ignoring.

...

Care to check what 'none' means to express in that file, and if that can just be removed? Did you, or Yast added it?

'none' is a valid option in that file.

#/dev/mapper/name device or file none options #----------------·······--------------··················-----···--------

cr_something /dev/disk/by-id/ata-XXXXX-partX none none

I've this content cr_sda2 /dev/disk/by-id/ata-ST95005620AS_5YX06VRQ-part2 none none which is mainly the partition where my volume group is setup and full encrypted (that works with systemd start) Kay thanks for the other suggestions, I will try to cleanup my factory to make it more systemd friendly :-) -- Bruno Friedmann Ioda-Net Sàrl www.ioda-net.ch openSUSE Member & Ambassador GPG KEY : D5C9B751C4653227 irc: tigerfoot -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

On 05/30/2011 05:20 PM, Kay Sievers wrote:

On Mon, 2011-05-30 at 16:29 +0200, Carlos E. R. wrote:

...

On 2011-05-30 15:26, Kay Sievers wrote:

...

...

May 30 10:38:11 c-3po kernel: [ 28.462295] systemd[1]: Breaking ordering cycle by deleting job cpufreq.service/start

Care to remove cpufreq. Nothing should fiddle around with CPU governors these days. The kernel's on-demand governor is the only sensible thing to do.

Bug 653540

Ah, I see. It's not the weird userspace daemon, it's only a crazy init script to to load a few modules which should be compiled into the kernel. :)

...

...

...

May 30 10:38:11 c-3po kernel: [ 32.223395] systemd-cryptsetup[855]: Encountered unknown /etc/crypttab option 'none', ignoring.

Care to check what 'none' means to express in that file, and if that can just be removed? Did you, or Yast added it?

'none' is a valid option in that file.

#/dev/mapper/name device or file none options #----------------·······--------------··················-----···--------

cr_something /dev/disk/by-id/ata-XXXXX-partX none none

Seems, there is no support for an options=="none" in systemd's native crypttab support, only for key=="none". It does not support a few other options of the Debian/SUSE crypttab parser too.

The 'none' for the options is not mentioned in the manpage of SUSE's crypttab though.

Did you add the last 'none' manually, or was it the installer/Yast which added it?

Kay

This one was write like this if you do an install by yast and use the lvm full encrypted choice. So you get the lvm volume group encrypted and on top of this one the different logical volume (in my case / /srv /home /var and swap) -- Bruno Friedmann Ioda-Net Sàrl www.ioda-net.ch openSUSE Member & Ambassador GPG KEY : D5C9B751C4653227 irc: tigerfoot -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

On Mon, 2011-05-30 at 20:33 +0200, Carlos E. R. wrote:

On 2011-05-30 17:20, Kay Sievers wrote:

...

On Mon, 2011-05-30 at 16:29 +0200, Carlos E. R. wrote:

...

On 2011-05-30 15:26, Kay Sievers wrote:

...

...

Bug 653540

Ah, I see. It's not the weird userspace daemon, it's only a crazy init script to to load a few modules which should be compiled into the kernel. :)

Whatever, but we were told to start that service to make gnome happy.

It's to make your kernel happy, that the CPU can safe power if it has nothing to do. It has nothing to do with the desktop stuff. Ideally we would just compile them into the kernel, as only the kernel knows which of the drivers should actually be used, and not depend on the loading order userspace provides. Changing the governors from on-demand makes almost no sense today. It's definitely nothing a distro needs to provide knobs for. But I enabled that cpufreq service here too now, and I don't get any cycles, so it must be something else in the list that creates the cycle systemd can't resolve.

...

...

'none' is a valid option in that file.

#/dev/mapper/name device or file none options #----------------·······--------------··················-----···--------

cr_something /dev/disk/by-id/ata-XXXXX-partX none none

Seems, there is no support for an options=="none" in systemd's native crypttab support, only for key=="none". It does not support a few other options of the Debian/SUSE crypttab parser too.

The 'none' for the options is not mentioned in the manpage of SUSE's crypttab though.

Did you add the last 'none' manually, or was it the installer/Yast which added it?

I don't remember. The first time it probably was yast, later ones, me.

I suggest you look into the system service script "/etc/init.d/boot.crypto", which lists the authors whom you may ask about that. All of them Novell/SUSE staffers.

But I don't understand why systemd cares about what is inside the scripts, anyway.

It doesn't look at any of these scripts. But it looks at /etc/crypttab and parses it. But the systemd crypttab parser does not support "none" like you have, if no options is given. I just wanted to know if we should expect more of these problems, like in case Yast added that line. Kay -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

On Tue, 2011-05-31 at 00:13 +0200, Carlos E. R. wrote:

On 2011-05-30 20:49, Kay Sievers wrote:

...

On Mon, 2011-05-30 at 20:33 +0200, Carlos E. R. wrote:

...

...

Whatever, but we were told to start that service to make gnome happy.

It's to make your kernel happy, that the CPU can safe power if it has nothing to do. It has nothing to do with the desktop stuff.

If that system service is not loaded, my laptop cpu is stuck at the highest speed. And I knew that because gnome told me so, not the kernel.

Ah, good that it tells. But the desktop does not need it.

...

Ideally we would just compile them into the kernel, as only the kernel knows which of the drivers should actually be used, and not depend on the loading order userspace provides.

Yes and no. The user can decide, for example, to force low speed to maximize battery. Or the contrary.

Low speed does in no way mean less energy on modern boxes. It's a legend, and proven wrong many times. To save energy the system needs to sleep, and when work is to do, it need to do the work as fast as possible, to be able to sleep again as much as possible. Only sleeping really saves energy not doing work slowly. All that is the kernel's job with the on-demand governor. The powersave governor usually does consume more energy, and is pretty useless today. Kay -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

On Mon, 2011-05-30 at 19:06 -0400, Jeff Mahoney wrote:

On 05/30/2011 06:29 PM, Kay Sievers wrote:

...

On Tue, 2011-05-31 at 00:13 +0200, Carlos E. R. wrote:

...

On 2011-05-30 20:49, Kay Sievers wrote:

...

On Mon, 2011-05-30 at 20:33 +0200, Carlos E. R. wrote:

...

...

Whatever, but we were told to start that service to make gnome happy.

It's to make your kernel happy, that the CPU can safe power if it has nothing to do. It has nothing to do with the desktop stuff.

If that system service is not loaded, my laptop cpu is stuck at the highest speed. And I knew that because gnome told me so, not the kernel.

Ah, good that it tells. But the desktop does not need it.

...

...

Ideally we would just compile them into the kernel, as only the kernel knows which of the drivers should actually be used, and not depend on the loading order userspace provides.

Yes and no. The user can decide, for example, to force low speed to maximize battery. Or the contrary.

Low speed does in no way mean less energy on modern boxes. It's a legend, and proven wrong many times.

To save energy the system needs to sleep, and when work is to do, it need to do the work as fast as possible, to be able to sleep again as much as possible. Only sleeping really saves energy not doing work slowly.

All that is the kernel's job with the on-demand governor. The powersave governor usually does consume more energy, and is pretty useless today.

FWIW, the cpufreq script loads powernow-k8 on my system. Regardless of your thoughts on power saving, it definitely steps down my processor (Phenom II X6) and as a result my fan is much, much quieter.

Oh, sure. We need the cpufreq _drivers_ loaded, otherwise stuff behaves terrible, no doubt about that. I mentioned the different _governors_. What I meant is that the 'powersave' vs. 'ondemand' governor usually does not save any power, it behaves worse. The kernel does its job right, and does not need userspace fiddling with governors. Kay -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

Am Tue, 31 May 2011 00:13:07 +0200 schrieb "Carlos E. R." :

Yes and no. The user can decide, for example, to force low speed to maximize battery. Or the contrary.

Low speed does not maximize battery on todays CPUs. But static high speed does minimize some types of latency for certain workloads, and does no longer necessarily waste significant power. (All on modern CPUs with C3 or deeper) -- Stefan Seyfried "Dispatch war rocket Ajax to bring back his body!" -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

On Tue, May 31, 2011 at 09:57:12AM +0200, Stefan Seyfried wrote:

Am Mon, 30 May 2011 20:49:33 +0200 schrieb Kay Sievers :

...

Changing the governors from on-demand makes almost no sense today. It's definitely nothing a distro needs to provide knobs for.

Explain to my customers who encounter >30% performance drop for their application with ondemand defaults.

I never expected to see such pathological workload in the wild, but it exists ;)

Have you reported this bug to the kernel developers? -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

On Wed, 2011-06-01 at 11:34 +0200, Stefan Seyfried wrote:

Am Wed, 1 Jun 2011 05:25:13 +0800 schrieb Greg KH :

...

On Tue, May 31, 2011 at 09:57:12AM +0200, Stefan Seyfried wrote:

...

I never expected to see such pathological workload in the wild, but it exists ;)

Have you reported this bug to the kernel developers?

This is no bug. There are pathological workloads.

Pseudo code:

for(;;) { receive(&data); /* blocks until data is received */ compute_with_high_cpuload_but finish_quickly(&data); send(&data); }

If the computation is finished in << 100ms, the cpufreq will not be switched up. Sleep time between data blocks is large (> 100ms)

Now imagine three such processes on three servers, handing data around in turns.

Switching to "performance" governor did increase the end-user performance by 30%.

I've seen this with ping-pong loads to localhost. On-demand doesn't (or at least didn't) always kick the cpu up a gear when we're using two cores with shared cache. The thing seems to get confused by the idle time.. doesn't realize that high frequency idle/busy transitions mean high speed communication. -Mike -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

On Mon, 2011-05-30 at 16:30 -0400, Jeff Mahoney wrote:

On 05/30/2011 09:26 AM, Kay Sievers wrote:

...

The unresolvable/cyclic dependencies cause unfortunately that D-Bus does not start properly, which will let NetworkManager and a lot of other stuff fail.

Unfortunately, they tend to fail mostly silently and don't allow the administrator to log in to fix it. I've been using systemd on several systems for a while now and this is consistently the biggest hurdle, in my mind, to making it the default.

Well, it's broken init scripts. They need to be fixed. We can't really work around that with systemd. It's a generic graph resolver, it can't tell what to do if the system specifies loops, it just tries to kick out things until the cycle gets broken. In this case it's sysstat which causes the cyclic deps.

I shouldn't need to work up a special target to drop into a shell in order to debug an issue in some random dependency -- especially when, sometimes, it can just be a simple package installation that can cause a complete boot failure. Without an *easy* workaround mode, I'm afraid it won't get off the ground as the default in openSUSE.

Add 'emergency' or 'single' to the kernel commandline? Or what are you looking for? Kay -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

On Mon, 2011-05-30 at 15:26 +0200, Kay Sievers wrote:

On Mon, 2011-05-30 at 11:17 +0200, Bruno Friedmann wrote:

...

My actual result ( tested 15 minutes ago ) are quite bad. with systemd system-sysinit and networkmanager activated I just get no network kde refuse to start due to dbus error and a lot of others services.

cat /var/log/messages | grep -i systemd May 30 10:38:11 c-3po kernel: [ 27.616858] systemd[1]: systemd 26 running in system mode. (+PAM +LIBWRAP +AUDIT +SELINUX +SYSVINIT +LIBCRYPTSETUP; suse)

...

May 30 10:38:11 c-3po kernel: [ 28.462252] systemd[1]: Breaking ordering cycle by deleting job sysstat.service/start

Can you try to remove or disable sysstat from the bootup? It might have dependencies in its headers with can't be resolved by systemd.

It is systat, I can reproduce it here. The init script seems to cause the issues. Please remove or disable sysstat until someone is able to fix the sysstat LSB headers. Thanks, Kay -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

On Tue, May 31, 2011 at 12:11:07AM +0200, Kay Sievers wrote:

On Mon, 2011-05-30 at 15:26 +0200, Kay Sievers wrote:

...

On Mon, 2011-05-30 at 11:17 +0200, Bruno Friedmann wrote:

...

My actual result ( tested 15 minutes ago ) are quite bad. with systemd system-sysinit and networkmanager activated I just get no network kde refuse to start due to dbus error and a lot of others services.

cat /var/log/messages | grep -i systemd May 30 10:38:11 c-3po kernel: [ 27.616858] systemd[1]: systemd 26 running in system mode. (+PAM +LIBWRAP +AUDIT +SELINUX +SYSVINIT +LIBCRYPTSETUP; suse)

...

May 30 10:38:11 c-3po kernel: [ 28.462252] systemd[1]: Breaking ordering cycle by deleting job sysstat.service/start

Can you try to remove or disable sysstat from the bootup? It might have dependencies in its headers with can't be resolved by systemd.

It is systat, I can reproduce it here.

The init script seems to cause the issues. Please remove or disable sysstat until someone is able to fix the sysstat LSB headers.

I will look into it. It would help if somebody creates a bugreport and assign it to me. Petr -- Petr Uzel IRC: ptr_uzl @ freenode