Skip to content

Migration from Dendrite

Zendrite is a fork of Dendrite. Follow these steps to move an existing Dendrite installation to Zendrite.

The binary is now zendrite instead of dendrite. Update systemd unit files, scripts, and process managers that use the old name.

The default config file is now zendrite.yaml instead of dendrite.yaml. If Zendrite can’t find zendrite.yaml, it falls back to dendrite.yaml and logs a warning. You can also pass the path explicitly:

zendrite --config dendrite.yaml

Rename the file to zendrite.yaml to silence the warning.

The sample config file is now zendrite-sample.yaml.

All admin API endpoints moved from the /_dendrite/ prefix to /_zendrite/. Update scripts, monitoring, and tools that call them.

For example:

  • /_dendrite/admin/evacuateRoom/{roomID} → /_zendrite/admin/evacuateRoom/{roomID}
  • /_dendrite/admin/registrationTokens → /_zendrite/admin/registrationTokens

The default JetStream topic_prefix is now Zendrite instead of Dendrite. With the built-in NATS server (the default), Zendrite ignores NATS data stored under the old prefix.

You have two options:

  1. Fresh start (recommended). Delete the old JetStream data directory and let Zendrite recreate the streams. You lose pending events in the old streams, but Zendrite recreates the streams from the database on startup.
  2. Keep the old prefix. Set topic_prefix: Dendrite in your config file to keep using the existing streams.

The Docker image changed its user and paths:

  • Container user: dendrite → zendrite
  • Config volume: /etc/dendrite → /etc/zendrite
  • Binary path: /usr/bin/dendrite → /usr/bin/zendrite

Update Docker Compose files and orchestration configs that mount volumes or use these paths.

DENDRITE_TRACE_HTTP is now ZENDRITE_TRACE_HTTP. Zendrite still accepts the old name for backwards compatibility.

Put the database user, password, and database name you used for Dendrite into zendrite.yaml. Zendrite migrates the database schema on first startup. It renames the internal dendrite_version column in the db_migrations table to zendrite_version, with no manual steps.

Zendrite purges a room once no local user has a non-forgotten membership row. The room_server.auto_purge_empty_rooms setting controls this and defaults to on_all_forgotten.

On the first restart after the upgrade, Zendrite purges every room where each local user has left and forgotten it, or never had a membership row.

  • To review these rooms before the restart, call GET /_zendrite/admin/emptyRooms.
  • To opt out, set room_server.auto_purge_empty_rooms: never (or the legacy false) in your config.

See Purge empty rooms automatically for details, including the more aggressive on_empty mode.