Salut,
second try. I've added some logging when the issue arise, and when I try
to recover from the failure.
Thanks!
-- Jeanfrancois
Jussi Kuosa wrote:
>
>> Let me know if that fix the issue.
>
> Hi Jean-Francois,
> and thank you for your incredibly fast response!
>
> I put your fix to test this morning and it ran almost 4 hours before going
> into 100% for one core. The symptoms are the same: selector thread for port
> 8282 is hot sitting in epollWait. As always, eden space mem graph looks like
> a heart-attack after the thrashing due to the amount of temporary objects
> that are created just to notice that there's nothing to do...
>
>>> One workaround for you is to set the
>>> CometContext.setExpirationDelay(-1). This will not enabled that extra
>>> Selector.
>
> I was supposed to do this in the morning, but I'll put that to test now.
> I'll let you know ASAP.
>
> -Jussi