Confluence backup and restore: diagnosing recovery path after failed upgrade

Goal

Diagnose whether to restore pre-upgrade Confluence backups or attempt forward repair after an upgrade fails to complete.

Constraints

Downgrading Confluence after database schema migration is generally unsupported. Data Center rolling upgrades can leave cluster nodes on different versions, while Server upgrades affect one node. Restoring requires database and home-directory backups from the same point in time; mismatched copies risk corruption. Residual upgrade markers or lock files may also block startup.

No specific failure is assumed. The choice depends on backup completeness, cluster topology, and whether migration finished.

Should pre-upgrade backup validation determine the restore-versus-repair decision? How does a split cluster version affect restore scope? What indicates that upgrade markers or lock files must be cleared before restart?