[Test-Announce] Systemd Test Day on Tuesday 2010/09/07

Michał Piotrowski mkkp4x4 at gmail.com
Wed Sep 8 19:48:00 UTC 2010


2010/9/7 Adam Williamson <awilliam at redhat.com>:
> On Tue, 2010-09-07 at 02:41 +0200, Michał Piotrowski wrote:
>> Hi,
>>
>> I want to compare systemd and upstart boot speed (I don't expect much,
>> so I won't be disappointed :))
>>
>> systemd
>> http://i56.tinypic.com/ilk4fq.jpg
>>
>> upstart
>> http://i53.tinypic.com/dnhrvm.png
>>
>> both starts system in 31 seconds, but systemd starts much more services.
>>
>> It looks like ntsysv doesn't have a systemd support. How can I disable
>> service in systemd? I suppose I need to use systemctl disable command,
>> but I don't know when to look for job names - list-jobs fails here
>>
>> [michal at dio ~]$ systemctl list-jobs
>> Failed to issue method call: Launch helper exited with unknown return code 1
>> [michal at dio ~]$ sudo systemctl list-jobs
>> Failed to get D-Bus connection: Failed to connect to socket
>> /org/freedesktop/systemd1/private: Connection refused
>> [michal at dio ~]$ sudo systemctl list-units
>> Failed to get D-Bus connection: Failed to connect to socket
>> /org/freedesktop/systemd1/private: Connection refused
>>
>> Regards,
>> Michal
>
> That sounds like you have some kind of dbus bug on your system.

My mistake. I tried to use systemctl after running the system with upstart.

> To
> answer your main question, though, if the service is still a SysV
> service, you can still disable it just as you would have done before -
> 'chkconfig ntsysv off'

I've never used the chkconfig. I always used the ntsysv - it's an old
tool that I know since RH6. Now it doesn't work as expected - I can't
disable services with it - maybe it's time to start using chkconfig?

>. It's only systemd-native services that need the
> use of systemctl.

I disabled some services and I noticed something interesting - it
almost doesn't speed up the boot process.

Before
http://i56.tinypic.com/ilk4fq.jpg

After
http://i56.tinypic.com/mh5347.png

The difference is 1 second. Interesting. An old method of speeding up
boot process doesn't work - is it a regression?

Okay, joking :)

Regards,
Michal


More information about the test mailing list