At the time it was on purpose but now I think about it more I agree it is
better to use some timeout.
Not sure when I will be able to get to it. Would you mind raising a Jira
for this so we won't forget about it?
Remko
On Thu, Jun 30, 2016 at 4:10 AM, David Leonhartsberger <leozi...@gmail.com>
wrote:
> I was just looking at the AsyncLoggerDistruptor#stop() and figured out that
> the shutdown time is not configurable and it even could be blocking forever
> waiting that the Disruptor backlog has been processed.
>
>
> // Calling Disruptor.shutdown() will wait until all enqueued events are
> fully processed,
> // but this waiting happens in a busy-spin. To avoid (postpone) wasting
> CPU,
> // we sleep in short chunks, up to 10 seconds, waiting for the ringbuffer
> to drain.
> for (int i = 0; hasBacklog(temp) && i < MAX_DRAIN_ATTEMPTS_BEFORE_SHUTDOWN;
> i++) {
> try {
> Thread.sleep(SLEEP_MILLIS_BETWEEN_DRAIN_ATTEMPTS); // give up the CPU for a
> while
> } catch (final InterruptedException e) { // ignored
> }
> }
> temp.shutdown(); // busy-spins until all events currently in the disruptor
> have been processed
>
>
> I was wondering why the "temp.shutdown()" call does not use a
> (configurable) timeout?
> Was this a design decision or was it just overlooked?
>
> Br,
> DavidL
>
