Some years ago I had this weird encounter with hibernate where ids were getting mixed up among 2 different tables, due to last insert id and lazy commit.
That was another year, another time and another hibernate version.
This time I got a new learning with the ID generation behavior.
To the point :
Make sure database sequence table has the same increment by as the allocationSize in @SequenceGenerator.
SELECT * FROM table_id_seq;
ALTER SEQUENCE table_id_seq INCREMENT BY 20; // 20 is the allocationSize in hibernate
What happens if they are different ?
-
allocationSize(10) <incrementValue(20)Hibernate gets the next value = 100.
Generates Ids till 110.
Requests for nextValue. Gets 120.
Works fine, but holes are created.
-
allocationSize(10) >incrementValue(1)Hibernate gets the next value = 100.
Generates Ids till 110.
Requests for nextValue. Gets 101.
Fails with
ValueExistException.
Why is it a big deal? Should be caught during testing.
Well, what makes it a bit obscure is, when worked with previous version, some properties (I assume hibernate.id.new_generator_mapping) were set in a way that even if there was mismatch, it relied with the increment by value of database.
With the 2.0 version of Spring Boot, it comes forward.
What about identity type generation ?
Identity type generation is certainly more cleaner, may be not much performant. However, it also requires changing the table definition.