Wait for complete disposal of Reactor Netty resources in ReactorResourceFactory

[bclozel]

While chatting with @violetagg about #23594, we noticed that ReactorResourceFactory, when disposing of HTTP resources, is using the "void returning" variants of the APIs (for example, see ConnectionProvider.dispose() in reactor-netty).

Those variants are returning a void type but aren't synchronous and are not waiting for the operation to complete. Under the hood, they're calling the reactive version and subscribing to it, even sometimes in an elastic scheduler.

Spring Framework should in this case use disposeLater().block(); to ensure that resources have been cleaned up before the method returns. This might explain part of the problem in #23594 if the resources aren't cleaned completely before the server shuts down.

[quaff]

after commit ca2b2f5 , webflux-http-kqueue-* is destroyed after context closed, but there is a thread named globalEventExecutor-1-1 not destroyed immediately, should wait for 1 second.

[violetagg]

@quaff This thread is used to notify that termination was completed.
https://github.com/netty/netty/blob/4.1/common/src/main/java/io/netty/util/concurrent/SingleThreadEventExecutor.java#L106
The termination of the event loop group is configured with 2s quiet time and 15s timeout.

[quaff]

@violetagg the quiet time should be 1s ? https://github.com/netty/netty/blob/c83904a12a603fe340477447d086674439ca8ac7/common/src/main/java/io/netty/util/concurrent/GlobalEventExecutor.java#L42

[violetagg]

@quaff the quiet period that I mentioned was for stopping the event loop. While this thread - yep it will be stopped when there is no task pending in the task queue for 1s
https://github.com/netty/netty/blob/c83904a12a603fe340477447d086674439ca8ac7/common/src/main/java/io/netty/util/concurrent/GlobalEventExecutor.java#L34-L36