Discussion on "Java's reactive connection pooling performance caveat"

Piotr · 2023-07-02T20:06:57.767Z

R2DBC problem introduction You are using a framework like micronaut-data with r2dbc and pooling enabled. Everything is set as per documentation. But you might see that sometimes queries are becoming very slow, as if resources were exhausted. But upon...

Thanks for the post. I am trying to understand the situation. My first observation is that in 2025, the current Netty does not accept negative numbers, see github.com/reactor/reactor-netty/blob/main/reactor-netty-core/src/main/java/reactor/netty/resources/LoopResources.java#L136

Ahh, sorry, it is possible to pass -1, wrong line, the correct one is below github.com/reactor/reactor-netty/blame/main/reactor-netty-core/src/main/java/reactor/netty/resources/LoopResources.java#L172

Hi.

I tried raising the reactor-pool version to 1.0.1 and running tests.

If I do the runWithoutColocation test first and then the runWithColocation test, the runWithoutColocation test measures slower.

So it seems to be an effect of the order of the tests and not an effect of LoopResources.

Hey, maybe I have not mentioned it but the tests were meant to be launched separately. I did not take much effort to properly isolate them. I guess when you launch them separately the results are reproducible right?

Discussion

Java's reactive connection pooling performance caveat

Responses(2)

Recent in Forum

Search Hashnode

Java's reactive connection pooling performance caveat

Responses(2)

Recent in Forum