Testbed Support for GridPP member institutes
> [mailto:[log in to unmask]] On Behalf Of Fisher, SM (Steve)
said:
> I don't think that is practical. Once a patch has gone out it can only
> be fixed by a new patch with higher RPM numbers in order to fix those
> sites that have installed the buggy patch
In general it should be possible for sites to back up to the previous
version, although in specific cases, e.g. with database schema changes,
it may not be easy. The problem is that at the moment it's all or
nothing, if you roll back (or don't install in the first place) you lose
everything in the update and you can't safely apply any further updates.
Unless we break the testing process, which has dangers in itself, it
takes at least 3 weeks or so to get a new patch released, and in some
cases you may well want to back out a broken update immediately. Anyway,
the procedure has to allow for patches to be rejected in the PPS, so
extending that to the PS should be possible.
> If broken patches are getting through then its best to take
> this up with SA3
It would obviously be better to catch things in the PPS, but I suspect
that some things will always get through.
> rather than installing arbitrary combinations of RPMs.
That's really the point, only by preserving the patch information can
you make sure that the rpms are coherent - each patch should be
self-contained (although it may have dependencies on other patches).
Stephen
|