Hi, On Sat, Apr 08, 2017 at 10:56:24PM -0700, Ryan Tandy wrote: > Fixing this for upgrades from jessie to stretch requires a jessie > update as well, to dump out the databases upon remove. > > The patches I intend to submit are attached. I would appreciate it a > lot if anyone reading this could take the time to review and try to > point out any edge cases I missed. > > The changes are intentionally minimal. Some refactoring will > probably be in order, but that can wait for buster. > > The intended paths are: > > - standard upgrades from jessie to stretch should not dump/reload > the databases needlessly > - remove-upgrade-install from jessie to stretch with the ppolicy > schema installed should succeed: the config should be dumped before > remove so that it can be updated and reloaded in postinst > - upgrades from current jessie to stretch should still work > - if preinst fails after dumping in prerm (for example because the > ppolicy check decides a manual update is required), and the upgrade > is retried, the second prerm should dump again and the updated data > should be used for the upgrade
What happens on remove if the slapcat fails? There are 2 options: - don't ignore errors This causes errors on removal when the ldap config is broken. There should be a way to remove a broken installation. - ignore errors This could cause data loss if the dump can't happen anymore after the upgrade. I'm not sure which option is best. Cheers, Ivo
