AxonThreadFactory should create virtual threads

[rsobies]

...instead of platform threads.
What do you think about this change? Will it cause any bugs?

[hjohn]

I wouldn't expect any problems when running on JDK25+ but we don't do this out of the box.

Depending on what scenario's you have in mind, you can do this yourself already.

For commands, queries and subscription events, the thread used is often restricted to the thread that triggered the action (as transaction management is often still thread bound) unless configured otherwise. If using Spring, that will be a Spring provided thread in many cases, and so you could configure Spring to use virtual threads.

For PooledStreamingEventProcessor, you can replace the executors that are used for workers and coordinators with executors that use virtual threads.

We've not done an extensive investigation whether or not virtual thread use will cause any bugs (this is also not completely up to the framework, but would also depend on other components of your software stack that are called via Axon Framework). There are of course known issues with JDK21 and use of synchronized blocks around code that does I/O, although it should work if you ensure you don't have any components that do this. For JDK25 we're not aware of any reasons it shouldn't work, but we've not extensively tested this.

[smcvb]

I'm with @hjohn here. Likely doesn't hurt, but we do not have backing investigation out of the box to proof it does.
Given that AF baselines on JDK21 with the aforementioned known-issues, I'd be hard pressed to enforce the AxonThreadFactory to virtual threads.
Once the baselines is passed that, perhaps we can.
So, with that said, I am leaning towards this being a user's configuration choice rather than a framework default, @rsobies.

Be sure to let us know what you think. Or, if you have the time to investigate whether the switch is merited; that input would be helpful too!

[rsobies]

thank you for clarification

[nils-christian]

Hello everyone,

Sorry for jumping in, but I happened to stumble across this issue.

We experimented with virtual threads in our platform a while ago, but we ran into problems specifically with the AxonThreadFactory. At the time, some deadlocks occurred. Unfortunately I don't remember the details and I haven't had the chance to look into it more closely since then. So I can't say for sure whether the issues were actually caused by the Axon framework itself or by some other components.

I just wanted to recommend to test this thoroughly before switching everything over to virtual threads.

Best regards

Nils