Storage

Enroute keeps Git data in an object store and transactional metadata in Postgres. Serving processes are stateless; they do not keep repository disks.

Object store

Git objects are stored in pack images. Reads use an index to locate a byte range in the object store. Active pushes use separate staging storage until ingestion completes.

Postgres

Postgres stores repository ownership, refs, object-ID-to-sequence mappings, and the index catalog. Enroute applies database migrations at startup.

Each Git object type has a separate sequence-number space. A sequence number is meaningful only with its object type.

Indexes

The commit graph and object index are segmented. Small segments are stored in Postgres; larger segments are stored in the object store. Writes append a segment, reads compose the needed segments, and maintenance merges segments.

The commit graph supports ancestry, merge-base, and history queries. The object index locates objects and records tree entries. See Architecture.

Ingestion and retention

The server keeps Git connections open, calls hooks, and updates refs. Ingestion stores uploaded objects and builds indexes. It can run locally or in Lambda. See Storage backends.

Enroute does not garbage-collect objects made unreachable by branch deletion. Delete the repository to remove its data. Configure encryption at rest in Postgres and the object store; Enroute does not provide it. Fetches that include have lines receive a complete pack rather than a thin pack.