Systemd, which overlay to choose? How well is it performing?

Message

benneque · Post by **benneque** » Sat Dec 18, 2010 10:57 am

Again I get those messages, too:

systemd[1]: Failed to initialize automounter: No such file or directory
systemd[1]: Unit sys-kernel-security.automount entered failed state.
systemd[1]: Failed to initialize automounter: No such file or directory
systemd[1]: Unit sys-kernel-debug.automount entered failed state.

And everything gets mounted!

Your problem should be this:

Code: Select all

systemd[1]: lvm.service: main process exited, code=exited, status=1
systemd[1]: Unit lvm.service entered failed state.
/etc/init.d/lvm[1419]: WARNING: lvm has already been started

Seems like sysv or sth. already started lvm and now systemd can't take control.

Have you tried this, 'cause it should be a fix for the same error message you get:
http://en.gentoo-wiki.com/wiki/Systemd# ... C_services

When you compile with "-sysv" you don't have to emerge -C openrc. By using init=/bin/systemd the kernel won't use openrc! So you can have both installed and switch between them using grub

soya · Post by **soya** » Sat Dec 18, 2010 8:56 pm

Yes, i did it and that's what i get:

Code: Select all

systemctl enable openrc-init.service
Unit files contain no applicable installation information. Ignoring.

costel78 · Post by **costel78** » Mon Jan 10, 2011 12:31 pm

Can somebody post an samba init script for systemd, please ?
It's the only service which still missing...
Thank you!

benneque · Post by **benneque** » Mon Jan 10, 2011 6:40 pm

look at this: https://bugzilla.redhat.com/show_bug.cgi?id=650589
somewhere exists "smb.service", but it doesn't seem to work right now...

costel78 · Post by **costel78** » Mon Jan 10, 2011 7:46 pm

Yep, it's not working. At least on Gentoo. But I'll try to fix it, maybe I have more luck.
Thank you!

benneque · Post by **benneque** » Mon Jan 10, 2011 10:40 pm

Would be nice

And post your results (maybe to gentoo-wiki)

codestation · Post by **codestation** » Tue Jan 11, 2011 12:48 am

costel78 wrote:Can somebody post an samba init script for systemd, please ?
It's the only service which still missing...
Thank you!

http://git.exherbo.org/?p=net.git;a=tre ... es/systemd

costel78 · Post by **costel78** » Thu Jan 13, 2011 7:48 am

codestation wrote: http://git.exherbo.org/?p=net.git;a=tre ... es/systemd

Thank you!

benneque · Post by **benneque** » Thu Jan 13, 2011 12:35 pm

Thanks for SMB service =)

I have now 2 other issues!

First: cups. In all systemd reviews etc. there's ALWAYS an example with cups. But I can't find any cups.service

Is it somewhere hidden or is it simply non-existant ?

Second: Something like /etc/init.d/local (or /etc/init.d/local.start) is missing for me. I need to undervolt my notebook at every start (saves some battery and more important: It won't overheat after one hour of compiling).
Does something like that exist? Or have I to write my own service ?

trippels · Post by **trippels** » Fri Jan 14, 2011 10:45 am

benneque wrote:Thanks for SMB service =)

I have now 2 other issues!

First: cups. In all systemd reviews etc. there's ALWAYS an example with cups. But I can't find any cups.service Is it somewhere hidden or is it simply non-existant ?

Second: Something like /etc/init.d/local (or /etc/init.d/local.start) is missing for me. I need to undervolt my notebook at every start (saves some battery and more important: It won't overheat after one hour of compiling).
Does something like that exist? Or have I to write my own service ?

A simple cups.service can be found on the wiki: http://en.gentoo-wiki.com/wiki/Systemd

I use something like this for my local startup scripts:

/etc/systemd/system/local.service:

Code: Select all


[Unit]
Description=local_misc
DefaultDependencies=false

[Service]
Type=oneshot
RemainAfterExit=yes
ExecStart=/etc/systemd/scripts/local

[Install]
WantedBy=sysinit.target

/etc/systemd/scripts/local:

Code: Select all

#!/bin/sh

echo "1" > /proc/sys/kernel/sysrq
echo "auto" > /sys/class/drm/card0/device/power_profile

chmod 0777 /sys/fs/cgroup/cpu
echo "/sbin/rm_cgroup" > /sys/fs/cgroup/cpu/release_agent

/usr/sbin/edac-ctl --register-labels
echo "1" > /sys/module/edac_core/parameters/edac_mc_panic_on_ue

exit 0

benneque · Post by **benneque** » Fri Jan 14, 2011 7:16 pm

Thanks!

Does this have to do anything with systemd ?

Code: Select all

chmod 0777 /sys/fs/cgroup/cpu 
echo "/sbin/rm_cgroup" > /sys/fs/cgroup/cpu/release_agent

trippels · Post by **trippels** » Fri Jan 14, 2011 9:40 pm

benneque wrote:Thanks!

Does this have to do anything with systemd ?
Code: Select all
chmod 0777 /sys/fs/cgroup/cpu 
echo "/sbin/rm_cgroup" > /sys/fs/cgroup/cpu/release_agent 

No, it's part of Lennart's alternative auto cpu cgroup approach (which is faster
than the in-kernel CONFIG_SCHED_AUTOGROUP according to my measurements
some time ago):
https://lkml.org/lkml/2010/11/16/392
http://en.gentoo-wiki.com/wiki/Improve_ ... th_cgroups

benneque · Post by **benneque** » Sat Jan 22, 2011 10:45 pm

systemd is making progress

seem's like they tagged v17!

I wonder if systemd will still get faster or (from now on) it still gets fatter and slower?
Or (the question other way round): Are there still features missing that could/will speed up the boot process?

zoolook · Post by **zoolook** » Sat Feb 05, 2011 10:39 pm

trippels wrote:
benneque wrote:Thanks!

Does this have to do anything with systemd ?
Code: Select all
chmod 0777 /sys/fs/cgroup/cpu 
echo "/sbin/rm_cgroup" > /sys/fs/cgroup/cpu/release_agent 
No, it's part of Lennart's alternative auto cpu cgroup approach

Is that all that's needed from the file /usr/local/sbin/cgroup_start from the wiki article you linked when using systemd, together with /usr/local/sbin/cgroup_clean? Is the ~/.bashrc part obsolete with systemd?

trippels wrote: (which is faster
than the in-kernel CONFIG_SCHED_AUTOGROUP according to my measurements
some time ago):

Can the two methods coexist smoothly, or do they bite each other?

Regards,
Lucian

trippels · Post by **trippels** » Sun Feb 06, 2011 12:07 pm

zoolook wrote:
trippels wrote:
benneque wrote:Thanks!

Does this have to do anything with systemd ?
Code: Select all
chmod 0777 /sys/fs/cgroup/cpu 
echo "/sbin/rm_cgroup" > /sys/fs/cgroup/cpu/release_agent 
No, it's part of Lennart's alternative auto cpu cgroup approach
Is that all that's needed from the file /usr/local/sbin/cgroup_start from the wiki article you linked when using systemd, together with /usr/local/sbin/cgroup_clean? Is the ~/.bashrc part obsolete with systemd?
trippels wrote: (which is faster
than the in-kernel CONFIG_SCHED_AUTOGROUP according to my measurements
some time ago):
Can the two methods coexist smoothly, or do they bite each other?

You still need the .bashrc part. And yes, the two methods bite each other.
So you either use CONFIG_SCHED_AUTOGROUP or the .bashrc thing.

heirecka · Post by **heirecka** » Fri Feb 18, 2011 1:57 pm

systemd 18 is out and works nicely for me, only NetworkManager keeps crashing during shutdown. I wonder if this happens for others, too?

benneque · Post by **benneque** » Sun Feb 20, 2011 10:24 am

yeah, I have this NM problem, too (since v15 or maybe earlier)

benneque · Post by **benneque** » Sun Feb 27, 2011 5:55 pm

Hey guys!

I've got new things to do with systemd: splashscreen!

There are several questions:
1. Which system should one prefer for splashscreen experience (last time I set it up war 3-4 years ago) ?
2. Is there a built-in system in systemd or sth. planned?

Thanks

benneque · Post by **benneque** » Tue Mar 08, 2011 3:53 pm

Does anyone know how to run systemd with plymouth?
I want to install systemd to my girlfriends laptop, but you know: Female computers NEED a fancy boot experience

What I did so far:
plymouth with openrc is running fine! (just a 2 sec prompt before X is there, but that's ok)
I've tried the same setup with systemd on my notebook. There's something wrong.
"systemctl --all | grep ply" gives me some service that failed...

I don't know what the service does, where exactly it belongs and why it fails. How can I retrieve an error message?

viralex · Post by **viralex** » Mon Apr 18, 2011 7:10 pm

Hi,
systemd is fantastic but I 've got a problem after gdm-3.0.0 starts.

With openrc/baselayout I see my username and normally logins entering my password.

When I try systemd there's no username in list.
I choose "other" then type username and password but gdm-session does not start.
I can't cancel running operation, so I have to kill gdm.

I'll try to enable syslog-ng and see what happends while I try to login.

Are there any pam problems with systemd or other?
Does anyone have this problem??

amade · Post by **amade** » Mon Apr 18, 2011 8:09 pm

You probably want ebuilds for services which are used with gnome3
https://bugs.gentoo.org/show_bug.cgi?id=318365 comments #154 to #157
I've added them to my local overlay with needed patches from normal gnome3 and systemd ebuilds and it worked

However there seems to be added gnome3 branch http://git.overlays.gentoo.org/gitweb/? ... ads/gnome3 so it is probably better if you use it, there is also a comment #158 which suggests problems without accounts-daemon.service

I have no problems running gnome 3 using only systemd services, except occasionally on shutdown, there seems to be unmounted device, and it timeouts before halting

viralex · Post by **viralex** » Mon Apr 18, 2011 8:28 pm

thanks!
I'll try these in my overlay

EDIT:

ok I have selected gnome3 branch on systemd overlay

"git checkout gnome3"

Uzytkownik · Post by **Uzytkownik** » Thu May 19, 2011 9:03 pm

I had 2 problems:

The system did not boot as /bin/systemd used /usr/lib/libdbus.so.<ABI> and I have separated /usr
The system fail to fsck.ext2 of root filesystem which is on LVM as it does not fine device.

The first one was fixed by quick'n'dirty copy to /lib. I don't know how to fix the second one - I found advice to update to newest LVM but I have ~amd64 so I should already have it. I have specified volumes by UUID and it correctly fould the /dev path to it.

keet · Post by **keet** » Thu May 26, 2011 3:24 am

I'm using systemd-27-r1 and systemd-units-9999 from the systemd overlay. My computer boots with systemd, loads kdm, then OpenBox (or presumably whichever W.M./D.E. one chooses). I have a few problems, though:

1. I have no network connection. I can't ping www.google.com, load websites, et cetera.
2. I can start programs in OpenBox, but if I try to use su and type the password correctly, it never goes to the new prompt; it just stays blank until I press ctrl+c. Likewise, if I press ctrl+alt+f1 (or f2, et cetera), it lets me tpe the username and password, and tells the last login time, but never runs bash or loads the prompt.
3. I can't turn off the computer properly. I'd provide a log of this, but I don't know where to find it. I'd like to say that I see no significant errors, but I'm not sure. I know that this is vague, but maybe it will help.

Here is my dmesg:

Code: Select all

[    4.723001][    6.585562] <29>systemd[1]: hald.service: main process exited, code=exited, status=203
[    6.604420] r8169 0000:0c:00.0: eth0: link down
[    6.604491] r8169 0000:0c:00.0: eth0: link down
[    6.605294] ADDRCONF(NETDEV_UP): eth0: link is not ready
[    6.605474] <29>systemd[1]: Unit hald.service entered failed state.
[    6.811353] <29>systemd[1]: metalog.service: control process exited, code=exited status=203
[    6.824768] <29>systemd[1]: Unit metalog.service entered failed state.
[    6.851133] <29>systemd[1]: metalog.service: control process exited, code=exited status=203
[    6.871637] <29>systemd[1]: Unit metalog.service entered failed state.
[    6.907632] <29>systemd[1]: metalog.service: control process exited, code=exited status=203
[    6.932461] <29>systemd[1]: Unit metalog.service entered failed state.
[    6.949812] <29>systemd[1]: metalog.service: control process exited, code=exited status=203
[    6.971350] <29>systemd[1]: Unit metalog.service entered failed state.
[    6.997732] <29>systemd[1]: metalog.service: control process exited, code=exited status=203
[    7.010242] <29>systemd[1]: Unit metalog.service entered failed state.
[    8.233065] r8169 0000:0c:00.0: eth0: link up
[    8.233803] ADDRCONF(NETDEV_CHANGE): eth0: link becomes ready
[    8.984710] <29>systemd[1]: metalog.service: control process exited, code=exited status=203
[    9.001543] <29>systemd[1]: Unit metalog.service entered failed state.
[    9.022751] <28>systemd[1]: metalog.service start request repeated too quickly, refusing to start.

I might just go through the whole Gentoo SystemD guide. Fortunately, I can boot using openrc.

Edit: It is mostly fixed:

1. I corrected the kernel path and root partition in /etc/systemd/scripts/kexec_load.
2. I am not sure whether this is correct, but I added this line to /etc/systemd/system/lvm.service:

Code: Select all

ExecStart=/usr/local/sbin/cgroup_start

3. I changed network.service, making these two lines instead of the previous ones:

Code: Select all

ExecStart=/sbin/ifconfig eth0 up
ExecStart=/sbin/dhcpcd eth0

Now the computer seems to boot fine, though it's a bit slower than I was hoping (OpenBox loads completely in 25 seconds, counting from the GRUB screen). I haven't tried shutting it down yet, though.

Dr.Willy · Post by **Dr.Willy** » Thu May 26, 2011 8:46 am

keet wrote:3. I changed network.service, making these two lines instead of the previous ones:
Code: Select all
ExecStart=/sbin/ifconfig eth0 up
ExecStart=/sbin/dhcpcd eth0

Revert that and do 'systemctl enable dhcpcd.service'