When upgrade fails during our pre/post steps (usually rake tasks), there is no way to recover other than manually investigating what happened. This is a suggestion to store list of successful steps so upgrade can be either reexecuted (if the rest is idempotent) or at least providing an option to skip executed steps. Easy implementation would be touching some timestamp files in /var/lib and informing user about skipped steps (and how to re-execute them when needed). This is a RFE but I think this should be handled with priority for the next major upgrade.
We are already doing the touching thing. How does it differ form https://bugzilla.redhat.com/show_bug.cgi?id=1264597 ? I though the upgrades are idempotent. -- Ivan
Sure.