Create a signals table in the monitored database with columns: id (VARCHAR PK), type (VARCHAR), data (VARCHAR/JSON) and grant INSERT to the Debezium user
Set signal.enabled.channels=source and signal.data.collection=schema.signals_table in the connector configuration to enable the signaling channel
Insert a signal row to trigger an incremental snapshot: INSERT INTO signals_table VALUES ('snap-1', 'execute-snapshot', '{"data-collections":["schema.table_name"],"type":"incremental"}')
Monitor the Debezium connector logs for messages indicating snapshot start and the watermark events that bracket the snapshot window in the change stream
Use snapshot.isolation.mode=repeatable_read (for MySQL) or exported (for PostgreSQL) in the connector config to control the consistency level of incremental snapshots
Known gotchas
Incremental snapshots use low/high watermarks inserted into the signaling table to bracket the snapshot window; if the signaling table is not in the list of captured tables, the watermark events will not appear in the change stream and the snapshot will hang
For PostgreSQL, the replication slot must be active and the Debezium connector must be streaming when the snapshot signal is inserted; signals delivered while the connector is stopped are not replayed on restart
Large tables undergoing an incremental snapshot generate a high volume of READ events on the Kafka topic before the streaming events resume; size the Kafka topic retention and consumer lag monitoring accordingly
Give your agent this knowledge — and 200+ more routes
One MCP install gives any agent live access to the full route map, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp