Skip to main content
With change data capture (CDC), Corvic keeps a copy of each Postgres table you name. It copies each table once, then follows every committed insert, update, and delete through your database’s logical replication. Changes reach Corvic within seconds of each commit and land in the table each time the pipeline runs. Nothing is installed in your database. It needs logical replication switched on, a publication naming the tables, and a user with the replication attribute that Corvic can reach, directly, from Dedicated IP addresses, or over Private Networking.

​
What your database needs

  • PostgreSQL 10 or later, on a primary. A read replica can’t hold a logical replication slot before Postgres 16.
  • wal_level = logical, which takes a restart:
  • Room for one more replication slot and WAL sender (max_replication_slots, max_wal_senders; both default to 10).
  • A replication login. Corvic supplies setup.sql, which creates a corvic_cdc user with the REPLICATION attribute (on RDS and Aurora, rds_replication), grants it SELECT on the schema, and creates the publication. Run it with psql as a superuser, or as the tables’ owner holding CREATEROLE.

​
Connect it

1

Add a PostgreSQL connector

In Settings → Connector Vault, add a PostgreSQL database connector. Turn on Replication login and enter the corvic_cdc username and password. The same connector can also hold a reader login for ordinary SQL queries.
2

Ask for the tables

In a room, ask Corvic to replicate the tables you want, for example “keep a copy of public.orders and public.customers”. Corvic sets up the watch and a pipeline that keeps them current.
Each row is keyed by the table’s replica identity: its primary key, a unique index named with REPLICA IDENTITY USING INDEX, or every column under REPLICA IDENTITY FULL.
Corvic holds a logical replication slot on the database. While the slot exists, the database keeps the WAL it hasn’t read. If you stop using CDC, ask Corvic to stop the watch so it drops the slot.