SUMMARY: multiple instances of a pkgadd - same version?

Thanks to Alan Pae and Reggie Beavers for responding.

Summary:

Seems that regardless of any overriding setting, whether instance=unique
or MAXINST=2, if the package is the same architecture and version, it is
only going to install one instance - period.  This is unfortunate for
installs such as a chroot environment, but somewhat understandable.

I recieved no answer as to whether this has/will cause problems
(instance=unique and then installing a package multiple times).  We may
try to back those out, re-install one, and then proceed with one of these:

The obvious answer is to build two packages [eg OURlibz and
OUTlibz-chroot], which we were really trying to avoid.  As I said, we're
already having to build variations of 32 and 64 bit packages, solaris 8
and 9 packages (when need be), etc.  We didn't want to have to throw
chroot/non-chroot into the mix.  It's a quick change, but the time adds up
with all the packages and variations and etc.

The not so obvious answer is to install the package VIA chroot and build
it's own /CHROOT/var/sadm library for installs inside the chroot.  Not
sure we want those extra files floating around the chroot, not to
mention a copy of pkgadd and pkgrm, but maybe...

thanks,
ivo


On Tue, 19 Aug 2003, Ian Veach wrote:

> We are planning on upgrading our servers with a few of our custom packages
> - asap because these are security fixes.  Some of our production servers
> run a chroot, and we thought we had installed the packages as
> instance=unique, once for the main OS and once inside the chroot [because
> some we need in both places, some we need in one or the other].  But we've
> recently found that a unique instance is only installed if the version or
> arch.  Because of this, we've got two questions/issues:
>
> (1) There is only once instance of a package instead of two.  It looks
> like the files are in both / and chroot though.  Theoretically I can just
> run pkgadd twice again [like we did originally] to upgrade it, but does
> this cause any problems that perhaps haven't manifested themselves yet?
> Has anyone done something similar and encountered problems?  That's one of
> our options - just to pkgadd twice but live with just one pkginfo record -
> presuming there aren't problems doing this.
>
> (2) Does anyone know of a way to force unique instances [same ver/arch] to
> avoid potential problems?  When building the packages, we are already
> building sol8/sol9/64bit/32bit versions, so we really don't want to have
> to build yet another combonation by changing version numbers or
> architecture settings, etc. unless we have to.
>
> thanks,
> ivo
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers