systemd: Yet Another Init Replacement

dl.zerocool · 2011-04-30 22:14:27

loonyphoenix wrote:

I've got a wierd problem. When I start with systemd, I've got an empty tray icon which is called org.kde.StatusNotifierItem-1702-1/StatusNotifierItem, and system notification don't work. (I'm using KDE.)
Also, #systemctl enable kdm.service doesn't make it start at startup. However, making graphical.target the default does. It's weird.
Also, I have these services which have the error status on loading:
auditd.service error inactive dead auditd.service
dbus-org....Avahi.service error inactive dead dbus-org.freedesktop.Avahi.service
plymouth...t-wait.service error inactive dead plymouth-quit-wait.service
I don't seem to have any of the .service files in /lib/systemd/system

Just do the following :

systemctl enable  avahi-daemon.service
systemctl status  avahi-daemon.service

If it's not started (but it should be started with the enable)

systemctl start  avahi-daemon.service

MartinZ · 2011-05-01 03:15:06

falconindy wrote:

I suggest switching to rsyslog, until syslog-ng can get its act together (and implement proper socket activation and takeover). This should fix a large number of problems that people have been experiencing recently wrt hanging at shutdown (and probably bootup as well).
community/rsyslog includes its own service file, so its simply a matter of HUP'ing systemd and enabling the new logger.

Thanks a lot, this fixed everything. By the way, I didn't know rsyslog and after a short reading, I think I'll stick with it.

Kane · 2011-05-01 11:09:28

I have a question about 'Early Start':

From the wiki:

Early start
One central feature of systemd is dbus and socket activation, this causes services to be started when they are first accessed, and is generally a good thing. However, if you know that a service (like console-kit) will always be started during boot, then the overall boot time might be reduced by starting it as early as possible. This can be acheived (if the service file is set up for it, which in most cases it is) by issuing:
# systemctl enable console-kit-daemon.service
This will cause systemd to start console-kit as soon as possible, without causing races with the socket or dbus activation.

Now i assumed

systemctl enable blah.service

just enabled a service like gdm for example to start up at boot. However the above seems to suggest that doing 'enable' sets it to start as early as possible.

Assuming it does start it as early as possible, how does it know what other services, e.g openntpd, to start normally (as there is nothing like the daemons array in rc.conf ?) or does it just go through every service file and check to see if the program is installed?
Or am i just misunderstanding what this early start thing is about.

Cheers guys,
Kane

eworm · 2011-05-01 12:52:50

frankieboy wrote:

I've tried systemd from the community repo again, what I've found that both on my laptop and on my home computer, the mount of the boot partition fails with a time-out, and if I gave the root password, I can manually mount the boot partition, why is that?
In my fstab, there is nothing specific:
UUID=something /boot ext4 defaults,noatime,nodiratime 1 2
On the home machine, after the timeout, it fails to mount /home and another partition mounted under home, it says dependency error.
On my laptop, networkmanager with systemd starts randomly, booting once everything is ok, on the other time, it fails.

Make sure you have up to date udev and mkinitcpio, then update your initrd.
I think this is because old initrd does not handle /run correctly (or at all).

falconindy · 2011-05-01 14:26:07

Kane wrote:

I have a question about 'Early Start':
From the wiki:
Early start
One central feature of systemd is dbus and socket activation, this causes services to be started when they are first accessed, and is generally a good thing. However, if you know that a service (like console-kit) will always be started during boot, then the overall boot time might be reduced by starting it as early as possible. This can be acheived (if the service file is set up for it, which in most cases it is) by issuing:
# systemctl enable console-kit-daemon.service
This will cause systemd to start console-kit as soon as possible, without causing races with the socket or dbus activation.
Now i assumed
systemctl enable blah.service
just enabled a service like gdm for example to start up at boot. However the above seems to suggest that doing 'enable' sets it to start as early as possible.
Assuming it does start it as early as possible, how does it know what other services, e.g openntpd, to start normally (as there is nothing like the daemons array in rc.conf ?) or does it just go through every service file and check to see if the program is installed?
Or am i just misunderstanding what this early start thing is about.
Cheers guys,
Kane

Each unit file declares its dependencies. Read up on systemd.target(5) and peruse /lib/systemd/system. "Early" in this context has nothing to do with the "early" that gets referenced by, say, KMS. The wiki is a bit a misleading.

nordmoen · 2011-05-01 14:31:28

First of all I want to thank Falconindy, it works very good, but I seem to have the syslog-ng problem. But I don't understand how to activate rsyslog, could someone please explain that quickly

radiomist · 2011-05-01 14:36:35

i have a weird trouble, i enable networkmanager

systemctl enable netoworkmanager

but, it doesn't start, when kde is fully loaded and type into konsole: systemctl start networkmanager. Systemctl hangs
but when i simply type NetWorkmanager, the daemons works

Last edited by radiomist (2011-05-01 14:37:26)

MartinZ · 2011-05-01 14:52:40

nordmoen wrote:

First of all I want to thank Falconindy, it works very good, but I seem to have the syslog-ng problem. But I don't understand how to activate rsyslog, could someone please explain that quickly

https://wiki.archlinux.org/index.php/Rsyslog

Cdh · 2011-05-01 15:28:30

When can we expect zsh completion for systemctl, especially with "systemctl enable [tab] *list of .service files*"?

falconindy · 2011-05-01 17:11:27

Cdh wrote:

When can we expect zsh completion for systemctl, especially with "systemctl enable [tab] *list of .service files*"?

This is no place to be asking such a question.

dl.zerocool · 2011-05-01 18:46:16

To everyone having troubles with KDE networkmanager.

I don't yet figured out how to exactly fully solve the problem, sometimes it works sometimes I've to start the networkmanager manually ...

So here is how I manage it :

pacman -S rsyslog
systemctl disable networkmanager.service
(EDIT : /etc/rc.conf and on DAEMONS keep networkmanager )

It works for me, well kind of...
I had to remove, clean kde cache and socket fold under ~/.kde and ~/.kde4

sometimes after a reboot I've to restart networkmanager
/etc/rc.d/networkmanager restart

By the way networkmanager just has been updated on main depository maybe it will help.

[EDIT]
Problem not related to systemd, but to kde
https://bbs.archlinux.org/viewtopic.php?pid=927415

Last edited by dl.zerocool (2011-05-03 09:18:00)

falconindy · 2011-05-01 19:17:06

Do not use the daemons array. networkmanager has a native service file which will ensure correct dependency ordering.

dl.zerocool · 2011-05-01 19:27:16

falconindy wrote:

Do not use the daemons array. networkmanager has a native service file which will ensure correct dependency ordering.

I know this, it just don't work by enable the networkmanager.service.

Don't know why, but using systemctl enable networkmanager.service instead of the DAEMONS array, does indeed start the networkmanager.
But I'm unable to connect to any wireless network from the kde networkmanager. Which I really can't explain...

lowie82ph · 2011-05-01 22:08:47

I just started using systemd and I think it's faster than Arch's default init system.

I have a question, is root fsck'ed during boot?
I only saw /boot and /home being fsck'ed.

Thanks

loonyphoenix · 2011-05-01 22:44:45

Hello. I understand that this is a bit premature, but has systemd been discussed as the possible replacement as the default init system for Arch? Is it likely to happen in the future?

ngoonee · 2011-05-01 23:04:12

loonyphoenix wrote:

Hello. I understand that this is a bit premature, but has systemd been discussed as the possible replacement as the default init system for Arch? Is it likely to happen in the future?

Yes it has. Search [arch-general] and see it yourself. Please note that almost everyone in that discussion wasn't a dev (some TUs though as I recall).

loonyphoenix · 2011-05-01 23:17:20

ngoonee wrote:

loonyphoenix wrote:
Hello. I understand that this is a bit premature, but has systemd been discussed as the possible replacement as the default init system for Arch? Is it likely to happen in the future?
Yes it has. Search [arch-general] and see it yourself. Please note that almost everyone in that discussion wasn't a dev (some TUs though as I recall).

Hmm, that was way back in January, and systemd has matured a bit since then. I feel systemd is pretty stable for me at least. I hope Arch does switch to systemd.

On the other hand, switching to systemd by default, imho, would require to gradually move away from rc.conf in favour of some other config files which are native for systemd. I predict quite a bit of inertia on that point, because Arch without /etc/rc.conf is difficult to imagine...

Ledti · 2011-05-01 23:37:25

loonyphoenix wrote:

Hello. I understand that this is a bit premature, but has systemd been discussed as the possible replacement as the default init system for Arch? Is it likely to happen in the future?

My magic eight ball says that it's "likely"...

I'm guessing any serious discussion about systemd integration won't happen 'til late next year, after package signing becomes a thing of the past and the community goes into withdraw for drama.

Place your bets!

loonyphoenix · 2011-05-01 23:42:53

Ledti wrote:

I'm guessing any serious discussion about systemd integration won't happen 'til late next year

So let me get this straight -- you're giving this a whole year and a half?

falconindy · 2011-05-01 23:52:13

lowie82ph wrote:

I just started using systemd and I think it's faster than Arch's default init system.
I have a question, is root fsck'ed during boot?
I only saw /boot and /home being fsck'ed.
Thanks

mounts will always be fsck'd prior to being mounted. root is no exception.

$ systemctl status fsck-root.service
fsck-root.service - File System Check on Root Device
      Loaded: loaded (/lib/systemd/system/fsck-root.service)
      Active: inactive (dead) since Fri, 29 Apr 2011 23:42:32 -0400; 1 day and 20h ago
    Main PID: 121 (code=exited, status=0/SUCCESS)
      CGroup: name=systemd:/system/fsck-root.service

Ledti · 2011-05-02 00:22:11

loonyphoenix wrote:

Ledti wrote:
I'm guessing any serious discussion about systemd integration won't happen 'til late next year
So let me get this straight -- you're giving this a whole year and a half?

I only make safe bets, "late next year" and "serious discussion" are ambiguous phrases... I always win!

Anyway, sorry for the semi-OT posts.

loonyphoenix · 2011-05-02 00:32:26

I've found something strange -- there doesn't seem to be a pulseaudio unit file. Why is that? I would have thought that would be one of the first daemons to get one, seeing as Lennart is the main systemd developer.

Ledti · 2011-05-02 00:45:15

loonyphoenix wrote:

I've found something strange -- there doesn't seem to be a pulseaudio unit file. Why is that? I would have thought that would be one of the first daemons to get one, seeing as Lennart is the main systemd developer.

Gnome and KDE start Pulse automatically. If you're not using a login manager, you can start it with start-pulseaudio-x11 as a user. As far as I know PulseAudio doesn't run with root privileges.

ngoonee · 2011-05-02 01:52:12

loonyphoenix wrote:

I've found something strange -- there doesn't seem to be a pulseaudio unit file. Why is that? I would have thought that would be one of the first daemons to get one, seeing as Lennart is the main systemd developer.

Pulseaudio does not need to be started at init, it is started as your user by your DE.

frankieboy · 2011-05-02 05:12:48

eworm wrote:

frankieboy wrote:
I've tried systemd from the community repo again, what I've found that both on my laptop and on my home computer, the mount of the boot partition fails with a time-out, and if I gave the root password, I can manually mount the boot partition, why is that?
In my fstab, there is nothing specific:
UUID=something /boot ext4 defaults,noatime,nodiratime 1 2
On the home machine, after the timeout, it fails to mount /home and another partition mounted under home, it says dependency error.
On my laptop, networkmanager with systemd starts randomly, booting once everything is ok, on the other time, it fails.
Make sure you have up to date udev and mkinitcpio, then update your initrd.
I think this is because old initrd does not handle /run correctly (or at all).

That didn't help, I've an up to date system. If you read back in the forum a little bit, you may find, that I've a special setup: a ocz ssd with hardware raid, containing /boot, /, /home, and a softraid raid0 with two hdd-s, containing /var, and a partition mounted under home. With earlier version, systemd didn't mount anything at all, now /, and /var are mounted, (/run and /var/run also), but I have to mount manually /boot, /home and the third partition. My question is do I need the mdadm hook, or /var can be mounted during boot without the mdadm hook?

Any help appreciated

Arch Linux

#476 2011-04-30 22:14:27

Re: systemd: Yet Another Init Replacement

#477 2011-05-01 03:15:06

Re: systemd: Yet Another Init Replacement

#478 2011-05-01 11:09:28

Re: systemd: Yet Another Init Replacement

#479 2011-05-01 12:52:50

Re: systemd: Yet Another Init Replacement

#480 2011-05-01 14:26:07

Re: systemd: Yet Another Init Replacement

#481 2011-05-01 14:31:28

Re: systemd: Yet Another Init Replacement

#482 2011-05-01 14:36:35

Re: systemd: Yet Another Init Replacement

#483 2011-05-01 14:52:40

Re: systemd: Yet Another Init Replacement

#484 2011-05-01 15:28:30

Re: systemd: Yet Another Init Replacement

#485 2011-05-01 17:11:27

Re: systemd: Yet Another Init Replacement

#486 2011-05-01 18:46:16

Re: systemd: Yet Another Init Replacement

#487 2011-05-01 19:17:06

Re: systemd: Yet Another Init Replacement

#488 2011-05-01 19:27:16

Re: systemd: Yet Another Init Replacement

#489 2011-05-01 22:08:47

Re: systemd: Yet Another Init Replacement

#490 2011-05-01 22:44:45

Re: systemd: Yet Another Init Replacement

#491 2011-05-01 23:04:12

Re: systemd: Yet Another Init Replacement

#492 2011-05-01 23:17:20

Re: systemd: Yet Another Init Replacement

#493 2011-05-01 23:37:25

Re: systemd: Yet Another Init Replacement

#494 2011-05-01 23:42:53

Re: systemd: Yet Another Init Replacement

#495 2011-05-01 23:52:13

Re: systemd: Yet Another Init Replacement

#496 2011-05-02 00:22:11

Re: systemd: Yet Another Init Replacement

#497 2011-05-02 00:32:26

Re: systemd: Yet Another Init Replacement

#498 2011-05-02 00:45:15

Re: systemd: Yet Another Init Replacement

#499 2011-05-02 01:52:12

Re: systemd: Yet Another Init Replacement

#500 2011-05-02 05:12:48

Re: systemd: Yet Another Init Replacement

Board footer