The purpose of this tool is to simplify and automate database migration or upgrade process without downtime.

./db_relocate run

1. Run pre-flight checks.
2. Start background health check process.
3. Create publication and replication slot.
4. Take a snapshot of the source database.
5. Upgrade the snapshot to the desired engine version (apply encryption if none).
6. Restore a new database instance from the snapshot.
7. Create a subscription on the destination database.
8. Advance replication to the correct LSN at the moment when the snapshot was taken.
9. Sync the data that was changed after the snapshot was taken.
10. Verify that all the heartbeat records have been synced.
11. Update sequences by leaving a small gap to avoid any conflicts.

After all the above steps have been processed, you will have two RDS PostgreSQL databases fully synced without any data loss. It is also your responsibility to double-check that all the data has been synced after the snapshot has been taken. The health check process will help you to be more confident.

TODO: It will be automated soon as well. As a final step, you need to deploy a load balancer (e.g., HAProxy), point your application to it, and then switch the traffic straight away. Then, you need to do a cleanup and point your app directly to the new database.

The entire procedure takes about an hour to run.

The db_relocate accepts a YAML configuration file, the location of which can be specified by the -config flag.

Example configuration file:

---
aws:
 profile: "profile-name"
 region: "eu-west-1"
src:
 user: "user"
 password: "password"
 name: "mydatabase"
 host: "test-db.someid.eu-west-1.rds.amazonaws.com"
 instance_id: "test-db"
upgrade:
 subnet_group: "test-db-subnet"
 engine_version: "13.7" # target db version
 parameter_group: "test-dn"
 password: "upgradeUser" # required for replication