AWS RDS
RDS and Aurora Postgres or MySQL, connected over a network path you control.
Postgres on RDS
- Set
rds.logical_replication = 1in the parameter group and reboot the instance. - Grant the connecting role
rds_replication— preflight checks this ascan replicate. - The schema-tracking event trigger needs superuser; on RDS that is
rds_superuser. Without it, usereconcileafter migrations. - The roles dump (
pg_dumpall --roles-only) is not permitted on RDS and is skipped automatically.
MySQL on RDS
- Set
gtid-mode = ON,enforce_gtid_consistency = ON, andbinlog_format = ROWin the parameter group, then reboot. - Enable automated backups so binary logging is on.
- Grant the user
REPLICATION SLAVEplus read access.
Network
RDS instances are usually not publicly reachable. Run the anybranch server where it can reach the instance (same VPC, or via a bastion / tunnel), and point preflight/clone at the instance endpoint. Preflight is read-only, so it is safe to run first to confirm the path and settings.
Changing a parameter group requires a reboot, which is disruptive on a production instance. Do it in a maintenance window, or connect a read-restricted replica the same way.