On Mon, 2009-03-23 at 16:25 +0100, Jeanfrancois Arcand wrote:
> > It is because, most of the applications typically try to persist data.
> > However, the steps, flag posted should help the investigation.
> >> In v3 those threads will be killed by Grizzly
> >> is the wait/block for too long to avoid thread starvation.
> > The flag introduced as stated in
> > https://glassfish.dev.java.net/issues/show_bug.cgi?id=6624
> > should make sure that the connection creation will be done in separate threads.
>
> Great. But Grizzly will still interrupt thread that take forever to
> execute to avoid situation like we are observing quite often with v2
> where GF becomes almost unresponsive and Grizzly starts queuing requests
> because it can't get a WorkerThread to execute. This is specially
> important with Grizzly as the number of threads doesn't need to be as
> high as when blocking I/O is used (like Tomcat).
Sure ! that will be helpful and probably we need a flag to disable that
for certain cases :-)
Thanks,
-Jagadish
>
> Thanks
>
> -- Jeanfrancois
>