On Mon, Jan 28, 2008 at 10:52:51AM -0800, David Bustos wrote:
> Quoth Alan Maguire on Mon, Jan 28, 2008 at 08:55:56AM +0000:
> > one consequence of this would be that network/physical:default
> > would have to be changed from a transient service to a
> > delegated restarter, but that makes sense too i think, since
> > it will also control the datalink and IP instances.
>
> Right. After thinking about this some more, I've come to a similar
> conclusion.
>
> ...
>
> If we can agree that this is the right long-term plan, then we can
> discuss whether we should make changes to SMF to make this easier, and
> whether we should opt for cheaper designs or implementations in the
> short-term.
>
> David
I think you've characterized it properly. But I'll make a follow-on point:
I think the root issue is that the networking stuff and other things like
it need to have the notion that we want to let startd run a method script,
but also poke some external managing process, if running, to tell it what
is going on (i.e. pre-stop or post-start) -- right now the current librestart
model doesn't support that well. You either have to do all of the work
or none of it, and as you say this idea of who is the restarter isn't
designed to automatically vary based on other service configuration.
I think two things are needed:
(a) The notion of this "proxy restarter" (come up with some better name) where
startd does method script but hooks are provided for interaction with
another service that is involved in managing or configuring the resource
(b) The idea that who the proxy is varies based on another service config.
e.g. network-physical:old, or network-physical:nwam, or
network-physical:cluster being enabled determines who's in charge,
but they all are sharing method scripts and a common SMF service
model for the underlying datalink and interface abstractions.
-Mike
--
Mike Shapiro, Solaris Kernel Development. blogs.sun.com/mws/
