HHH000444 warning, follow on locking (appears after migration to Spring Boot 3)

Hibernate ORM
1

After migrating from Spring Boot 2.7 to Spring Boot 3, this warning appears in the logs when executing a query with lock:

"HHH000444: Encountered request for locking however dialect reports that database prefers locking be done in a separate select (follow-on locking); results will be locked after initial query executes"

I checked the Hibernate version, which also changed during the migration

before migration: org.hibernate:hibernate-core:5.6.15.Final.

after migration: org.hibernate.orm:hibernate-core:6.2.13.Final.

Here are sample queries I have, there are several and they all look quite similar to these:

    @Transactional(propagation = Propagation.MANDATORY)
    @Lock(LockModType.PESSIMISTIC_WRITE)
    @Query("SELECT n FROM #{#name} n WHERE n.id = ?1")
    T findById(UUID id);

    @Transactional(propagation = Propagation.MANDATORY)
    @Lock(LockModeType.PESSIMISTIC_READ)
    boolean IsExists(K id);

I also tried to force this warning to be disabled with query hints by adding such an annotation to methods:
@QueryHints(value = { @QueryHint(name = "hibernate.query.followOnLocking", value = "false") }, forCounting = false)

But that didn’t help either, and the warning still appears. (I am using a PostgreSQL database)

I read an article about this error: HHH000444 warning, follow on locking

And they write that in version 6.2.3.Final there was a bug that caused follow-on locking and it has been fixed, however this kind of warning appears and I don’t know what the bug really means and why it appeared.

Thank you in advance for help.

2

It is possible that you’re seeing this warning if you’re using an entity-graph for the query that causes a left join to be created in the SQL. Since for update locking doesn’t work in PostgreSQL when using left joins, Hibernate ORM will use follow-on locking and hence emit this warning.
Another possible reason for left joins to be generated is that you’re fetching an entity that has an association which is a @OneToOne(mappedBy = "...") or a @XToOne with @NotFound.

Show the HQL queries that Spring generates and passes to Hibernate ORM as well as the SQL queries that are generated by Hibernate ORM.

3

Thanks for the reply,

Since for update locking doesn’t work in PostgreSQL when using left joins, Hibernate ORM will use follow-on locking and hence emit this warning.

Referring to this, has this always been the case in Hibernate, or was it only introduced in version 6? Well, since I didn’t get any warning in version 5 (so what could have really changed in the new version (?))

4

Hmm, that doesn’t really make sense. Can you please try to create a reproducer with our test case template (https://github.com/hibernate/hibernate-test-case-templates/blob/master/orm/hibernate-orm-6/src/test/java/org/hibernate/bugs/JPAUnitTestCase.java) and if you are able to reproduce the issue, create a bug ticket in our issue tracker(https://hibernate.atlassian.net) and attach that reproducer.

5

Hi @beikov we are using hibernate-core:6.4.1.Final and postgresql DB, we have an entity A which have @ManyToOne association with entity B (without @NotFound). We still are facing this warning and the sql generated is left join. Now what is happening that we are facing lost update concurrency problem. The transactions are using jakarta.persistence.EntityManager#find(Class<T> entityClass, Object primaryKey, LockModeType lockMode) with PESSIMISTIC_WRITE lock.

Below is the sql trace -
==== txn 1 =====

2024-02-06 01:19:37.554 [ThreadPoolTaskExecutor-2] [] WARN  - org.hibernate.sql.results.jdbc.internal.DeferredResultSetAccess HHH000444: Encountered request for locking however dialect reports that database prefers locking be done in a separate select (follow-on locking); results will be locked after initial query executes 
Hibernate: 
    select
        a1_0.a_id,
        a1_0.b_id,
        b1_0.b_id,
        b1_0.name,
        b1_0.status,
        a1_0.dummy_col 
    from
        entity_a a1_0 
    left join
        entity_b b1_0 
            on b1_0.b_id=a1_0.b_id 
    where
        a1_0.a_id=?
Hibernate: 
    select
        b_id 
    from
        entity_b 
    where
        b_id=? for update
Hibernate: 
    select
        a_id 
    from
        entity_a 
    where
        a_id=? for update

==== txn 2 ====

Hibernate: 
    select
        a1_0.a_id,
        a1_0.b_id,
        b1_0.b_id,
        b1_0.name,
        b1_0.status,
        a1_0.dummy_col 
    from
        entity_a a1_0 
    left join
        entity_b b1_0 
            on b1_0.b_id=a1_0.b_id 
    where
        a1_0.a_id=?
Hibernate: 
    select
        b_id 
    from
        entity_b 
    where
        b_id=? for update

Both txns are using same method, however txn2 hangs (after step over through EntityManger.find()) with these 2 sql queries when txn1 has already taken lock on the row. After txn1 commits, txn2 fires

Hibernate: 
    select
        a_id 
    from
        entity_a 
    where
        a_id=? for update

Lost updates weren’t there in hibernate 5.6.15.Final which we were using ealier, neither this warning.
Many thanks for any help

6

Please create a reproducer based on our test case template and share that.