Upgrade without losing your work
- Stop the server cleanly and back up the entire
plugins/DungeonForge/directory and relevant worlds. - Keep one DungeonForge JAR in
plugins/, then start a staging copy with the preview. - Check the startup log, edit and save an existing dungeon, then test a full run, rewards, death, disconnect, and restart.
- Verify your optional integrations before upgrading the live server.
Builder inventory recovery
Snapshots are stored at builder-hotbars/<uuid>.yml. Recovery handles ordinary exits, disconnects, online shutdown, rejoin, and respawn, including a player who disconnects while dead. Restored player data is saved before the snapshot is deleted. Displaced items are returned to inventory; overflow drops at the player.
If a snapshot is unreadable, it is retained and an error is logged. Preserve it and the server log for investigation. This is not a fully atomic crash journal for every possible inventory/server failure.
Imported dungeons
Imports now retain serialized gameplay settings, including start room, timers, rewards, procedural/code generation options, and marker defaults. The imported copy gets a new identity and author, starts unpublished, and resets statistics and builder-world linkage. Inspect rooms and templates before publishing. Whole-pack provisioning and rollback still need further work.
Temporary instance cleanup
Deletion requires a world created and owned by the current process, successful player evacuation, successful unloading, and a safe direct-child path within the world container. Name collisions are refused. Worlds that cannot be cleaned safely are retained for operator review; do not assume an old instance folder is disposable based on its name alone.
At shutdown, cleanup avoids scheduling fresh world/reset work. Persistent ownership reconciliation after crashes is a future task.