Hi,

If you using declarative pipelines then databricks respects the definition that you gave on the table declaration. If you subsequently do an alter table statement in the same pipeline it won't preserve it.

What you should do is either set the primary key in the declarative pipeline table definition or put the alter table as an additional task in the job. Although Databricks doesn't enforce primary key constraints it should still be aware of them.

CREATE OR REFRESH STREAMING TABLE orders_valid(
  CONSTRAINT pk_customer_id PRIMARY KEY (customer_id)
)
AS SELECT * FROM STREAM read_files("/databricks-datasets/retail-org/sales_orders");

Thanks,

Emma