Updated November, 23 2023

Schema migration

After initialization is done, we can start schema migration. It’s split into two stages, with names schema and constraints. The schema stage must run before init and constraints after data. The second stage is needed to create indexes, constraints and triggers. It’s run before the data stage to make the migration process faster and avoid triggers to break something in the database.

Schema migration can be parallelized. For this, just run whatever number of threads are needed, and the migration services will evenly schedule the jobs between the nodes.

It’s also possible to skip some objects’ translation, by just updating dbt.migration_objects table’s stage field to S. All the objects dependent on this one won’t be translated too.

We can also adjust original DDLs to remove or change anything not supported by Liberatii. For this, just update the DDL1 and DDL2 columns of the corresponding dbt.migration_objects rows before running. DDL1 is executed on the schema stage, and DDL2 on the constraints stage.

The schema migration can be stopped and restarted at any time. The migration process won’t start from the beginning. But you may need to reset the stage column of dbt.migration_objects to I.