[Freeipa-devel] [PATCH] 0097 Avoid --list option of chkconfig

[Martin Kosek]

On Wed, 2011-05-04 at 17:16 -0400, Simo Sorce wrote:
> On Wed, 2011-05-04 at 16:57 -0400, Simo Sorce wrote:
> > On Wed, 2011-05-04 at 17:12 +0200, Martin Kosek wrote:
> > > On Wed, 2011-05-04 at 11:01 -0400, Simo Sorce wrote:
> > > > On Wed, 2011-05-04 at 16:56 +0200, Martin Kosek wrote:
> > > > > On Wed, 2011-05-04 at 10:28 -0400, Simo Sorce wrote:
> > > > > > The chkconfig --list option is not reliable as it works differently on a
> > > > > > F15 system with systemd enabled services, avoid using it as we have
> > > > > > alternative better ways.
> > > > > > 
> > > > > > Simo.
> > > > > > 
> > > > > 
> > > > > NACK. This breaks the installation:
> > > > > 
> > > > > # ipa-server-install
> > > > > ...
> > > > > Configuring ntpd
> > > > >   [1/4]: stopping ntpd
> > > > >   [2/4]: writing configuration
> > > > >   [3/4]: configuring ntpd to start on boot
> > > > > Unexpected error - see ipaserver-install.log for details:
> > > > >  Command '/sbin/chkconfig --level 3 ntpd' returned non-zero exit status 1
> > > > > 
> > > > > I think you want to make use of raiseonerr=False parameter in
> > > > > ipautil.run().
> > > > > 
> > > > > Plus, I would prefer integers to be used in runlevels array instead of a
> > > > > string containing an integer. Makes more sense to do it this way.
> > > > 
> > > > New patch attached.
> > > > 
> > > > Simo.
> > > > 
> > > 
> > > ACK. Worked for me on both F-14 and F-15.
> > 
> > Hold on pushing this one, it seem that chkconfig service and chkconfig
> > service --list were not mututally exchangable in this context as we
> > thought ...
> 
> After the conversation here [1] and after looking more closely at how
> the functions that used --list were actually used in the code I came to
> the conclusion that the use of chkconfig service --list was not really
> needed in either cases.
> 
> In the client we assume all services we need to be present through
> packaging dependencies anyway, and even if we fail we properly catch any
> exception so it is never a fatal issue.
> 
> In the server install we used to check if the service was enabled at all
> run levels before considering it really enabled, but upon careful
> consideration I believe we do not really care. In each place we called
> is_enabled() what we care about is really just if the service is enabled
> in the current runlevel, not if it is normally enabled at other run
> levels, and actually checking for other runlevels could cause issue
> because a service not enabled at runlevel 4 but currently enabled at the
> one the install is running with would result disabled, and that's not
> what callers are really expecting from that test.
> 
> So long story short I have a patch the basically radically strips away
> the previous code and radically simplifies it to one simple call in the
> server code and none at all in the client code.
> 
> Simo.
> 

The concept looks usable, but I am having some problems with chkconfig
$SERVICE for systemd native package (our favorite ntpd):

# systemctl disable ntpd.service
# chkconfig ntpd; echo $?
Note: Forwarding request to 'systemctl is-enabled ntpd.service'.
0
# systemctl enable ntpd.service
ln -s '/lib/systemd/system/ntpd.service'
'/etc/systemd/system/multi-user.target.wants/ntpd.service'
# chkconfig ntpd; echo $?
Note: Forwarding request to 'systemctl is-enabled ntpd.service'.
0

Does this work for you? Looks like a bug in `systemctl is-enabled`.

I am running systemd-26-1.fc15.x86_64.

Martin