> In my experiences,
> > LiveUpgrade appears to only upgrade the metacluster
> and any other
> > packages installed on the machine. It doesn't
> appear to always
> > properly add packages that might have been added to
> the metacluster
> > in certain circumstances.
> >
>
> If you have the data to indicate this is happening,
> please file a bug, as it shouldn't be.
Since it was a one-shot deal, I didn't bother keeping any of the data. It
shouldn't be too hard to replicate though.
> > This means it isn't out of the realm of impossibility to create a
> > non-bootable machine in some instances. For example, I had a S9 box
> > installed with SUNWCreq+other packages that I attempted to LU to S10.
> > It had some significant issues because for some reason the zone
> > packages didn't get installed as part of the LU.
> >
>
> The zones packages are not included in the SUNWCreq
> metacluster.
Correct. Which means that anything that calls zonename breaks:
/sbin/zonename f none 0555 root bin 10144 3935 1146256529 SUNWcsr
/usr/bin/zonename=../../sbin/zonename s none SUNWzoneu
sunos/sparc tsunami$ ldd /sbin/zonename
libzonecfg.so.1 => /usr/lib/libzonecfg.so.1
....
sunos/sparc tsunami$ grep libzonecfg /var/sadm/install/contents
/usr/lib/libzonecfg.so.1 f none 0755 root bin 127248 32768 1145661297 SUNWzoneu
/usr/lib/sparcv9/libzonecfg.so.1 f none 0755 root bin 145968 20676 1145661297
SUNWzoneu
I leave it as an exercise to the reader to find how many calls to the zonename
command are sprinkled throughout the OS and how many of those calls aren't
properly error checked. A fun place to start is /lib/svc/method.
[... and, yes, I know I should use pkgchk to find this info, but I can never
remember the screwy syntax.]
This message posted from opensolaris.org
