The ReactiveConnectionPoolProvider implementation is leaking threads

[Sanne]

When I forget to start the database and then run all integration tests, I see errors like these being logged repeatedly:

Exception: java.lang.OutOfMemoryError thrown from the UncaughtExceptionHandler in thread "vertx-blocked-thread-checker"

I haven't verified my theory but in my experience this makes me suspect that the testsuite is leaking threads. If it's "just" the testsuite in these conditions that might not be critical, but I wonder if it could affect regular runtime.

[gavinking]

Probably related to the new schema export stuff. In which case it won't affect the runtime.

[Sanne]

Turns out I have the same problem even with databases turned on - it just takes longer to appear.

I'll try to get a profiler connected

[Sanne]

Probably related to the new schema export stuff. In which case it won't affect the runtime.

you guessed right, threads are stuck on:

at java.util.concurrent.locks.LockSupport.park(LockSupport.java:175)
at java.util.concurrent.CompletableFuture$Signaller.block(CompletableFuture.java:1707)
at java.util.concurrent.ForkJoinPool.managedBlock(ForkJoinPool.java:3323)
at java.util.concurrent.CompletableFuture.waitingGet(CompletableFuture.java:1742)
at java.util.concurrent.CompletableFuture.join(CompletableFuture.java:1947)
at org.hibernate.reactive.service.ReactiveGenerationTarget.accept(ReactiveGenerationTarget.java:29)
at org.hibernate.tool.schema.internal.SchemaDropperImpl$DelayedDropActionImpl.perform(SchemaDropperImpl.java:542)
at org.hibernate.internal.SessionFactoryImpl.close(SessionFactoryImpl.java:792)
at org.hibernate.reactive.BaseReactiveTest.after(BaseReactiveTest.java:75)
at org.hibernate.reactive.CompositeIdTest.lambda$after$0(CompositeIdTest.java:39)
at org.hibernate.reactive.CompositeIdTest$$Lambda$163/959646593.accept(Unknown Source)
at java.util.concurrent.CompletableFuture.uniWhenComplete(CompletableFuture.java:774)
at java.util.concurrent.CompletableFuture$UniWhenComplete.tryFire(CompletableFuture.java:750)
at java.util.concurrent.CompletableFuture.postComplete(CompletableFuture.java:488)
at java.util.concurrent.CompletableFuture.complete(CompletableFuture.java:1975)
at org.hibernate.reactive.impl.PoolConnection.lambda$toCompletionStage$5(PoolConnection.java:103)

Why wouldn't this affect runtime? but also, how am I supposed to launch the tests?

[gavinking]

Why wouldn't this affect runtime?

You don't run hundreds of schema exports at runtime.

[Sanne]

Why wouldn't this affect runtime?

You don't run hundreds of schema exports at runtime.

Even if it's just one, we can't have a thread stuck forever?

Anyway, my OOM seems caused by running it with JDK8: with JDK11 it works.

[gavinking]

Even if it's just one, we can't have a thread stuck forever?

It's actually only the schema dropper that has a problem, and that's probably because of how the session factory gets closed within the test suite.

If I set:

configuration.setProperty( AvailableSettings.HBM2DDL_AUTO, "create" );

in BaseReactiveTest, the problem seems to disappear.

[gavinking]

Pretty sure this is fixed now.

[Sanne]

Thanks. I still have OOMs when building with JDK8 though - I can't investigate now, I'll see if I can diagnose it an evening.

[Sanne]

Ok, I was able to figure it out - I'll send a PR later.

[Sanne]

Changed the title as the problem was in the actual runtime code