Resetting a Source

Resetting a source drops the replication slot (and optionally the publication), clears all position tracking, and forces a full re-baseline on the next startup cycle. This is more disruptive than a reload — it destroys all replication state.

When to reset

Scenario	Why reset is needed
Slot invalidation	WAL data is gone; the slot cannot resume
Slot corruption	PostgreSQL reports the slot is in an inconsistent state
Changing publication tables	Adding or removing tables from the publication (with create = true, Laredo handles this automatically; reset is for manual publication management)
Decommissioning a source	Clean up the slot and publication from PostgreSQL
Switching from ephemeral to stateful mode	Old slot configuration is incompatible
Changing the slot name	Old slot must be dropped manually or via reset

Reset via CLI

Basic reset (slot only)

laredo reset-source pg_main

This prompts for confirmation:

This will reset source "pg_main" (drop and recreate replication slot).
Continue? [y/N] y

The command drops the replication slot named in the source configuration. The next engine startup cycle creates a new slot and runs a full baseline.

Reset with publication drop

laredo reset-source pg_main --drop-publication

This additionally drops the PostgreSQL publication. Use this when decommissioning a source entirely, or when you need to rebuild the publication from scratch.

Reset via gRPC

The OAM gRPC service exposes the ResetSource RPC:

rpc ResetSource(ResetSourceRequest) returns (ResetSourceResponse);

{
  "source_id": "pg_main",
  "confirm": true,
  "drop_publication": false
}

The confirm field must be true — the server rejects requests without confirmation as a safety measure.

What happens during a reset

1. Slot drop: Laredo executes DROP REPLICATION SLOT on the PostgreSQL replication connection. If the slot does not exist (already dropped), this step is skipped.

2. Publication drop (if --drop-publication): Laredo executes DROP PUBLICATION IF EXISTS for the publication associated with the source.

3. Position cleared: The engine clears its internal position tracking (slotLSN and slotExisted flags). The source no longer has a resume point.

4. Re-baseline on next cycle: When the engine restarts the source (either because it was restarted or via the internal restart loop), it creates a new replication slot and performs a full baseline for all tables on that source.

All targets bound to the reset source receive OnTruncate followed by the full baseline row set.

Verifying the reset

After the reset completes, check that the source recovered:

# Source should show a new slot and be streaming
laredo source pg_main

# Pipelines should transition through BASELINE → STREAMING
laredo pipelines

Verify in PostgreSQL that the old slot is gone and the new one exists:

SELECT slot_name, active, restart_lsn, confirmed_flush_lsn
FROM pg_replication_slots
WHERE slot_name LIKE 'laredo%';

Safety considerations

• Reset is destructive. All position tracking is lost. The next baseline reads every row from every table on the source. For large tables, this can take significant time and put load on PostgreSQL.
• Downstream targets are cleared. All targets receive a truncate event before the new baseline rows arrive. During this window, queries against in-memory targets return empty results.
• The CLI requires confirmation. The y/N prompt prevents accidental resets. The gRPC API requires confirm = true.
• Only resettable sources support this. The source must implement the Resettable interface. The PostgreSQL source (source/pg) supports it. Other sources may not.

Decommissioning a source

To fully remove a source from PostgreSQL:

# Drop the slot and publication
laredo reset-source pg_main --drop-publication

# Then remove the source from the laredo-server config and restart

If Laredo is not running, you can drop the slot and publication directly in PostgreSQL:

SELECT pg_drop_replication_slot('laredo_01');
DROP PUBLICATION IF EXISTS laredo_01_pub;