[Users] host deploy and after reboot not responsive
Alon Bar-Lev
alonbl at redhat.com
Wed Jan 23 10:16:09 UTC 2013
----- Original Message -----
> From: "Dan Kenigsberg" <danken at redhat.com>
> To: "Gianluca Cecchi" <gianluca.cecchi at gmail.com>
> Cc: "Alon Bar-Lev" <alonbl at redhat.com>, "Adam Litke" <agl at us.ibm.com>, "users" <users at ovirt.org>
> Sent: Wednesday, January 23, 2013 12:09:59 PM
> Subject: Re: [Users] host deploy and after reboot not responsive
>
> On Wed, Jan 23, 2013 at 10:57:03AM +0100, Gianluca Cecchi wrote:
> > On Wed, Jan 23, 2013 at 9:40 AM, Dan Kenigsberg wrote:
> > > Gianluca, how about softening our ntpd requirement with something
> > > like
> > > http://gerrit.ovirt.org/11291 ?
> > >
> > > Could you verify that it's working on your system?
> >
> > Yes, I can test the change the service modification for vdsmd.
> > I presume you want me to test it with default fedora 18, so chronyd
> > enabled and ntpd disabled, correct?
>
> Yes. though my patch may generate excessive noise, with its attempt
> to
> start two conflicting service. I suppose that we can/should "want"
> only
> chrony. But please provide your input.
We should use either, and default.
On none systemd other distributions there is the concept of 'provides'.
This means that a service like vdsmd can depend on timesync and a service like chronyd *AND* ntpd provide timesync.
This is kind of logical name for service.
Also other non systemd distributions provides the ability to script the dependencies.
I could not find either of he above methods in systemd documentation.
>
> >
> > BTW: in case of indipendent services both enabled in a target
> > level,
> > how systemd processes them, in parallel?
> > In old style init there were
> > S25
> > S35
> > that ruled the start order, together with some rules inside scripts
> > themselves such as
> >
> > # Proveides: ....
> > # Required-Start: ....
>
> I'm no systemd expert, but I believe that it build a dependency tree,
> and attempts to start independet services in parallel, to save time.
>
> Dan.
>
More information about the Users
mailing list