The repository now aggregates two independent modules. migration-behavior/ is the
original project, moved unchanged; it stays on Spring Boot 4.0.6 / JDK 21 because
that is what the four published migration articles were verified against, and
upgrading it would silently invalidate output they quote. The article-tagged trees
are untouched, so links into a tag are unaffected.
transactions/ Companion code for "@Transactional in Spring: Propagation, Isolation,
and the Six Ways It Silently Does Nothing". Spring Boot 4.1.1 / JDK 25.
Every row of the propagation matrix is produced by calling the method and asking the
transaction manager what it did. The transaction NAME is the exhibit: a scope that
joined reports its caller's name, a scope that started its own reports its own.
Three things the transcripts settle:
- Propagation.NESTED cannot be used with JpaTransactionManager. It fails twice,
with two different messages, the second of which blames your JPA provider. The
savepoint manager comes from the object the JpaDialect returns when it begins the
transaction, and Hibernate's does not implement one. It works on
DataSourceTransactionManager, because a savepoint is a JDBC concept -- shown
working there rather than only failing here.
- Catching a REQUIRED inner failure does not save the transaction. The inner scope
has already marked it rollback-only, so the commit throws
UnexpectedRollbackException from a place with no connection to the cause.
- A checked exception commits, and so does a swallowed one. Those two do not merely
fail to start a transaction; they commit work the code was abandoning.
19 contract tests, six captured transcripts, all regenerated by scripts/run-all.sh.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Gip4srpzMwjgoba6uEfbr5
3.4 KiB
← Rollback · Index · Six silent failures →
4. NESTED, and why it does not work with JPA
Transcript: 03-nested.txt.
NESTED is described everywhere as "uses savepoints, so the inner scope can roll back without
taking the outer transaction with it". That description is accurate and, on a stock Spring Boot
JPA application, unreachable.
Attempt 1 — a normal Spring Boot JPA application
NestedTransactionNotSupportedException:
Transaction manager does not allow nested transactions by default -
specify 'nestedTransactionAllowed' property with value 'true'
A clear message naming the fix. So:
Attempt 2 — do what the message says
JpaTransactionManager manager = new JpaTransactionManager(factory);
manager.setNestedTransactionAllowed(true);
NestedTransactionNotSupportedException:
JpaDialect does not support savepoints - check your JPA provider's capabilities
A different message, from a second check, pointing at your JPA provider rather than at your configuration.
It is not the provider. JpaTransactionManager obtains the savepoint manager from the object
the JpaDialect returns when it begins the transaction, and Hibernate's does not implement one.
Verified by disassembling JpaTransactionManager$JpaTransactionObject, where the second check
is getEntityManagerHolder().getSavepointManager() == null. No configuration clears it.
There is a trap inside the trap: constructing JpaTransactionManager by hand also discards the
JpaDialect Spring Boot would have supplied from the Hibernate vendor adapter, leaving the
no-op DefaultJpaDialect. That produces the same second message for a different reason, and
sends you looking at your database instead of at your @Bean.
Attempt 3 — the same propagation on a JDBC transaction manager
{
"rowsVisibleInsideNestedScope": 2,
"nestedScopeThrew": "nested scope fails",
"rowsAfterNestedRollback": 1,
"rowsAfterOuterCommit": 1,
"surviving": ["outer-row"],
"transactionManager": "DataSourceTransactionManager (not JpaTransactionManager)"
}
Two rows visible inside the nested scope; one after it rolls back; one after the outer transaction commits — and the survivor is the outer row. That is savepoint semantics working exactly as advertised.
A savepoint is a JDBC concept. DataSourceTransactionManager holds the JDBC connection and can
issue one. JpaTransactionManager holds an EntityManager and cannot.
The reference documentation does say NESTED works with JDBC resource transactions. What it
does not say is that the JPA path fails, twice, with two different messages, the second of
which blames your database.
What to do instead
Use REQUIRES_NEW. It solves most of what people reach for NESTED to solve — "let this
part fail without losing everything" — at the cost of a second connection and independent
commit semantics.
The genuine difference: REQUIRES_NEW commits the inner work even if the outer transaction
later fails, while NESTED would have discarded it. If you need "roll back this part, keep the
rest, and still lose everything if the outer transaction fails", you need savepoints and
therefore JDBC.
Mixing two transaction managers over one DataSource, as
JdbcNestedService does, is a
demonstration and not a recommendation.