> First of all, thanks to all of you for contributing to this discussion.
> From what I hear, it seems to me that the fatal state can be relaxed in
> this case and we can resync the files after boot. Only until we pass
> this test will Sun be able to certify some oracle products, so this is
> *very* urgent. What I would like to ask the alias, to focus a bit more
> the discussion is:
>
> 1) what is the easiest way to resync the files in single user mode so
> that the systems can boot again? (I'd hate to reinstall the OS)
The really safe thing to do, which is only required if non interfaced
kernel modules such as say genunix and say strplumb conflict, is to
reboot to the fail-safe and have it re-sync the boot archive. If
> 2) if I move back to S10GA would I still see this problem?
You won't.
> in other words, was this introduced in update 1?
It was. However it's not a simple regression, but a side effect of a
major architectural change.
> 3) what is the fastest/easiest way to relax this fatal state in a
> supported way? a patch?
Likely, however given what's going on with u3 (possibly no patches
for bugfixes until u4), it might take a while to get a patch out. I
presume binary relief with the promise of a patch is not sufficient
for this certification?
> changes to smf configuration parameters in S10U1?
You could eliminate the boot-archive check service by disabling it:
svcadm disable boot-archive
However that would leave you open to blowing through miss-matched
kernel modules. I'd really want to discuss this carefully with you
(and perhaps another brain or two in the room) to make sure that that
would be an acceptable workaround for your situation. If you tightly
control updates to the OS, it may well be acceptable.
-jan
