Quoth Christine Tran on Wed, Mar 15, 2006 at 02:33:23PM -0500:
> What's the *functional* difference between milestone "multi-user-server"
> and  "all"?  I found that these services are disabled in
> "multi-user-server" where they are online in "all".
...

At the multi-user-server milestone, services which
milestone/multi-user-server doesn't depend on (directly or indirectly)
will be disabled.  For milestone all, no services are automatically
disabled.

> One could argue that shell, ftp, telnet etc. should be enabled in a
> milestone called "multi-user-server", since that's analogous to
> run-level 3 and the old rc script was rc2.d/S72inetsvc.

Right, all services which were started from rc2.d should declare
milestone/multi-user as a dependent, so they are run in the multi-user
milestone.  Similarly, services from rc3.d should declare
multi-user-server as a dependent.  I don't know why shell, ftp, and
telnet were left out of the milestones, but it would be legitimate to
file a bug about it.

>                                                          In fact, why
> even have a milestone called "all" at all, why couldn't we have started
> everything up at multi-user-server?

Because then everything would have to have multi-user-server as
a dependent.  See, for example, system/zones, which depends on
multi-user-server.

> I use svcprop -p options_ovr/milestone system/svc/restarter:default to
> find my current milestone.

A more accurate method would be to query options_ovr/milestone, and if
it doesn't exist, use options/milestone (and if that doesn't exist,
assume "all").  I'm pretty sure this is not stable.

>                             If the box has just gone straight from
> booting to "all", running this command gives me an error message.  If I
> were to bring the box down to another milestone, then back to "all",
> running this command gives me "all", the correct milestone.  Please see
> below.
...
> Why does options_ovr/milestone not exist when the box first boot up?
> The only other ovr thing that comes to mind is when I use the -t option
> to temporarily enable/disable a service, then I have general_ovr/enabled
> in the instance.

options_ovr is a nonpersistent property group, and is used to enact
temporary changes.  general_ovr is the analogous group for temporary
enables and disables.  options_ovr doesn't exist after boot because it
is nonpersistent, which is the reason for its existance.

> Customer thinks it would be useful for SMF to incorporate a time
> component.  Seeing that services are now controlled using svcadm, he
> thinks its superfluous to have to atjob some service that he wants to
> turn on a 10Pm and stop at 2AM, say.  I agree with him; also, cron is OK
> but is there any reason why we can't/shouldn't make SMF handle cron-like
> service management?

Well I believe there was talk, once upon a time, of a cron restarter.
I don't think it was meant to start and stop other services at certain
times, though.  That sounds like a reasonable RFE to me.


David

Reply via email to