Edward Pilatowicz wrote:
wrt smf_disable.lst, it seems like this file should be empty to me.
looking at the current services in this file, it seems like they are
services which should never exist within a zone, not just be disabled
within a zone.  so if these services shouldn't exist within a zone, then
the smf manifest that defines them shouldn't be installed in a zone, and
if the manifests for these services are being installed within zones
then that seems like a packaging bug.  does what i'm saying make sense?

Ed,

You are correct that these services shouldn't exist in a
zone but since we're doing a p2v, the services could exist
in the original physical system archive.  Most SMF services
are disabled based on their pkg metadata.  That is, SMF
services whose opensolaris.zone property is 'global'.  However
there are some services which should be tagged this way but
aren't.  Since we know those services won't work in a zone,
we disable them using the smf_disable.lst to avoid users
having problems with those services.

You are correct that this is a packaging bug for these svcs but
thats the way they currently exist.  We should fix the pkg
metadata but that doesn't fix the problem for people who are
taking pre-existing systems which have incorrect pkg metadata
and putting them into a zone.  That's really the whole purpose
of this conf file, to handle this well known brokeness.  If
all pkg metadata had always been correct and will always be
correct, we wouldn't need this file.  This file gives us a
mechanism to handle the cases we know about which are broken.

As an alternative we could remove this mechanism and simply
acknowledge that there are broken pkgs which the user will have
to manually fix up after they p2v.  This is a mechanism which
I carried over from the SVr4 p2v code so its possible we don't
want it with the IPS p2v.

Thanks,
Jerry

_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss

Reply via email to