On 24-2-2014 12:46, Mark Rotteveel wrote:

> I have thought this over during the weekend, and I will be investigating
> the bug. If I can identify the cause and the fix doesn't have too much
> impact, I will fix it in Jaybird 2.2.5 or 2.2.6.
>
> However my advice remains: switch to a third party connection pool (or use
> the one built-in in your application server).

I am investigating the problem and their are definitely some
inconsistent and insufficient synchronization issues and potential races
that could cause it. Unfortunately adding more synchronization similar
to the existing sync's (on the queue itself and on a 'takeMutex') would
probably cause some nasty deadlocks or starvation.

Fixing it isn't going to be as straight-forward as I'd like.

Does your code by any chance 'restart' or 'shutdown' and then 'start'
the pool occasionally? It looks like that has a higher chance of
breaking it.

Mark
--
Mark Rotteveel