Threading model of BeanPersistListener

[gdoenlen]

Hi, I was planning on adopting BeanPersistListener to execute side effects that only happen after a transaction is committed. The docs say that BeanPersistListener runs in a background thread, but doesn't provide many more details.

• Is is a single thread for everything?
• Is the threadpool configurable?
• Is is a platform thread or a virtual?
• How does it handle ThreadLocals or ScopedValues? I imagine at some point I'll have to do queries in this thread, but a lot of that in my system depends on the current user or current tenant that are stored in threadlocals. It looks like I can supply a BackgroundExecutorWrapper to copy that stuff over? Is it automatically picked up like the other listeners?

[rbygrave]

Is is a single thread for everything?

There is a thread pool

Is the threadpool configurable?

Not very, no.

Is is a platform thread or a virtual?

Platform.

How does it handle ThreadLocals or ScopedValues?

Have a look at MdcBackgroundExecutorWrapper. That is by default used, copies the slf4j MDC context onto the background thread. It is composable. We can change it via DatabaseBuilder.setBackgroundExecutorWrapper(...)

[gdoenlen]

Thanks Rob. I found the background wrapper. I ended up moving everything into my own thread pool from the listeners so I wouldn’t block Ebean’s pool. I’m not particularly latency sensitive so any overhead of thread context switching wasn’t a big deal.

https://github.com/SFXD/trust-bot/blob/master/src/main/java/com/github/sfxd/trust/events/Pipeline.java#L24