Provides a throttling strategy based on the number of pending blocking tasks.

If all threads of a BlockingTaskExecutor are busy, a new incoming request is enqueued and will wait until its turn comes. It will cause GC pressure and a response will be sent when the request is timed out. Even though the request was timed out, the blocking task executor will execute regardless of the state of the request.

It would be useful to throttle incoming requests based on the queue size of BlockingTaskExecutor

class BlockingTaskLimitingThrottlingStrategy<T extends Request> extends ThrottlingStrategy<T> {
        @Override
    public CompletionStage<Boolean> accept(ServiceRequestContext ctx, T request) {

        var theadPoolExecutor = getTheadPoolExecutor(ctx.blockingTaskExecutor());
        if (theadPoolExecutor.getQueue.size() > MAX_QUEUE_SIZE) {
            CompletableFuture.completedFuture(false);
        } else {
            CompletableFuture.completedFuture(true);
        }
    }
}

[ikhoon]

What about serving throttling strategy with the mutable limiting size?

Sounds good to me. I think we can take the IntSupplier style we did in https://github.com/line/armeria/pull/3985.

[ikhoon]

Thank you, can I try it?

I’m looking forward to your PR. 😆