DRAFT Panache Support multiple persistence unit in Hibernate Reactive #50084

lucamolteni · 2025-09-16T13:11:54Z

This is mostly a back port from the blocking version.

This reactive version shows a few problems and I wanted to discuss it with the team before going forward

Please share your thoughts about this, especially with the changes of the public API, as the synchronous isPersistent returns now a Uni

@FroMage @yrodiere

This fixes this reproducer provided by a user

FroMage

Answered inline

...rc/main/java/io/quarkus/hibernate/reactive/panache/common/runtime/AbstractJpaOperations.java

FroMage · 2025-09-16T13:41:05Z

...me/src/main/java/io/quarkus/hibernate/reactive/panache/common/runtime/SessionOperations.java

                context.removeLocal(SESSION_ON_DEMAND_OPENED_KEY);
-                return closeSession();
+                // TODO Luca this is just wrong
+                return closeSession(DEFAULT_PERSISTENCE_UNIT_NAME);


I suppose this should close all currently opened sessions, no?

This is used in WithSessionOnDemandInterceptor I think we could assume that the interceptor works only for the default persistence unit.

Or else we could add the Persistence Unit name to the @WithSessionOnDemand annotation so that the the interceptor knows about it. WDYT?

It's debatable.

My opinion is that this says "open a session automatically when asked", and then later (this is lazy), someone opens a session, at which point we know which session and which PU.

If we only allow a single session open at all times, then this should close that session that we opened, no matter its PU.

If we allow any number of sessions for different PUs open within a @WithSessionOnDemand then they should all be closed.

So wait, actually in both of the cases I mention, it would close all the currently opened sessions :)

...te-reactive-panache/runtime/src/main/java/io/quarkus/hibernate/reactive/panache/Panache.java

...e-panache/runtime/src/main/java/io/quarkus/hibernate/reactive/panache/PanacheEntityBase.java

FroMage · 2025-09-16T13:47:25Z

...nache/runtime/src/main/java/io/quarkus/hibernate/reactive/panache/PanacheRepositoryBase.java

    public default Uni<Void> flush() {
-        return INSTANCE.flush();
+        // TODO Luca not sure about this should it flush all the sessions?
+        // TODO Luca this will crash


Good question :-/
Yeah, flush all the sessions.

If needed, add an overload for flushing a single PU?

~~I'm not sure why I wrote this comment as this method might flush just the specific session/pu the entity is attached and I think it makes more sense WDYT? ~~

EDIT I remembered now, this in repository is a repository not an entity, so it can't work

In the new version I flush all the PUs, please take a look

Actually… a repository targets a single entity, so it should flush that entity's PU by default, no?

@FroMage yes but nowhere we save in the repository the type of the entity if I'm not mistaken and it gets erased at runtime. If you think we can find the type of ENTITY somewhere can you please me point me how?

...nache/runtime/src/main/java/io/quarkus/hibernate/reactive/panache/PanacheRepositoryBase.java

...-reactive-panache/src/test/java/io/quarkus/it/panache/reactive/PanacheFunctionalityTest.java

...rc/main/java/io/quarkus/hibernate/reactive/panache/common/runtime/AbstractJpaOperations.java

* Integration test for multiple reactive persistence units and Panache Backported from ORM the handling of different persistence units in entity in Panache * `@WithSessionOnDemand` works only with the default persistence unit * withSession overload to take the PU name * Execute update without entity runs on default session * Repository flush() flushes all PUs

lucamolteni · 2025-09-30T09:46:09Z

I've tackled most of the comments and the current tests are passing

I still need to add a few more test for the named persistence unit overload

github-actions · 2025-09-30T10:22:00Z

🎊 PR Preview de98066 has been successfully built and deployed to https://quarkus-pr-main-50084-preview.surge.sh/version/main/guides/

Images of blog posts older than 3 months are not available.
Newsletters older than 3 months are not available.

FroMage · 2025-09-30T12:57:28Z

...rc/main/java/io/quarkus/hibernate/reactive/panache/common/runtime/AbstractJpaOperations.java

+     * This method flushes *all* persistence units.
+     * Ideally, it should flush only the session associated with the repository’s entity,
+     * but we don’t currently track the entity inside the repository.
+     *


I'm not sure why this is the case, is it the case for the ORM version?

FroMage · 2025-09-30T13:00:01Z

...me/src/main/java/io/quarkus/hibernate/reactive/panache/common/runtime/SessionOperations.java

                context.removeLocal(SESSION_ON_DEMAND_OPENED_KEY);
-                return closeSession();
+                // TODO Luca this is just wrong
+                return closeSession(DEFAULT_PERSISTENCE_UNIT_NAME);


So wait, actually in both of the cases I mention, it would close all the currently opened sessions :)

FroMage · 2025-09-30T13:07:50Z

...nache/runtime/src/main/java/io/quarkus/hibernate/reactive/panache/PanacheRepositoryBase.java

     * @return the current {@link Mutiny.Session}
     */
    public default Uni<Mutiny.Session> getSession() {
        return INSTANCE.getSession();


OK, so this is the missing piece of the puzzle!

In the ORM version, we have:

@GenerateBridge default Session getSession() { throw implementationInjectionMissing(); }

Which is how we can pin a repo to its entity's PU.

Now, if you do the same here, you can obtain the proper session for the repo, and implement flush() exactly as ORM does:

default void flush() { getSession().flush(); }

:)

Exactly! It's what I've written in this comment

* Used by Panache repositories. * * This method flushes *all* persistence units. * Ideally, it should flush only the session associated with the repository’s entity, * but we don’t currently track the entity inside the repository. * * In Panache blocking, Quarkus injects the current EntityManager. * In reactive mode, however, there is not yet an injectable Mutiny.Session * (see https://github.com/quarkusio/quarkus/issues/47462).

So when #47462 is done, it should be easier to do that

No, that's not related to stateless sessions. This should be a change done in this PR. The same change was done for the ORM support of multiple PUs.

lucamolteni requested review from FroMage and yrodiere September 16, 2025 13:12

quarkus-bot bot added the area/panache label Sep 16, 2025

lucamolteni changed the title ~~[DRAFT] Panache Support multiple persistence unit in Hibernate Reactive~~ DRAFT Panache Support multiple persistence unit in Hibernate Reactive Sep 16, 2025

FroMage requested changes Sep 16, 2025

View reviewed changes

yrodiere reviewed Sep 19, 2025

View reviewed changes

...rc/main/java/io/quarkus/hibernate/reactive/panache/common/runtime/AbstractJpaOperations.java Outdated Show resolved Hide resolved

lucamolteni force-pushed the reproducer-multiple branch from 08d41a4 to 577434e Compare September 30, 2025 09:14

lucamolteni force-pushed the reproducer-multiple branch from 577434e to 500587e Compare September 30, 2025 09:44

lucamolteni requested review from FroMage and yrodiere September 30, 2025 09:45

FroMage requested changes Sep 30, 2025

View reviewed changes

DRAFT Panache Support multiple persistence unit in Hibernate Reactive #50084

Are you sure you want to change the base?

DRAFT Panache Support multiple persistence unit in Hibernate Reactive #50084

Uh oh!

Conversation

lucamolteni commented Sep 16, 2025

Uh oh!

FroMage left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lucamolteni Sep 30, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

lucamolteni commented Sep 30, 2025

Uh oh!

github-actions bot commented Sep 30, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

lucamolteni Sep 30, 2025 •

edited

Loading