State Store

pgsense-rs keeps replication progress (LSN checkpoints, table-sync state) in a store. The choice between in-memory and PostgreSQL-backed storage is the single setting store under the [pipeline] section:

[pipeline]
store = "memory"     # default — state is lost on restart
# or
store = "postgres"   # state persists in the source DB under the `etl` schema

Memory

In-memory state is the default and requires no setup.

Warning

On every restart, the pipeline starts from PostgreSQL’s current LSN and may miss events written during the restart window. Use memory only for local development and ephemeral environments.

PostgreSQL

When store = "postgres", state is stored in the source database itself under a hardcoded etl schema, using the same credentials as the replication connection. The scanner installs the schema and bootstrap helper functions on startup if they don’t exist.

This means:

• Restarts resume from the last persisted LSN — no missed events.
• State and source data are backed up together, so they stay consistent.
• The replication user must be able to CREATE SCHEMA (or the etl schema must already exist with appropriate ownership).
• First-time migration application requires superuser because one of the upstream migrations creates an event trigger (supabase_etl_ddl_message_trigger). After migrations apply, the role can be downgraded. See PostgreSQL Setup → etl schema bootstrap for the full permission matrix.

Note

There is no separate state-database option — state always lives in the source database. This is a deliberate trade-off; see the project README for the reasoning.

Switching stores

Caution

The two stores are not interchangeable at runtime. Switching from memory to postgres (or back) on an existing deployment effectively restarts the pipeline from scratch — there is no migration path.