Jason Dillon wrote:
>
> > That shouldn't happen. Just looked into the code I
> > wrote, and it looks like you have found a bug in it:
> > In org.jboss.tm.TxManager line 314, the new timeout
> > in milliseconds is calculated with the expression:
> > timeout = 1000 * seconds;
> > I guess it should have been
> > timeout = 1000L * seconds;
> > to enforce widening _before_ the multiplication.
> >
> > Maybe Integer.MAX_VALUE/1000 is long enough for you
> > until i get around to committing a fix?
>
> I notice that Integer.MAX_VALUE was producing a -1 when looking at the JMX
> console, so I played around a little and ended up using 2000000 (which is
> about 24 days).
>
> I re-ran my tests with no errors. I am rerunning them now with a 24 hour
> timeout, to see what happens when these long transactions complete. My
> guess is nothing, but just to be sure. I can try Integer.MAX_VALUE/1000 and
> see if that works too. I just need something that is insainly high, since I
> don't want _any_ tx to timeout on the node where this stuff is running.
I just committed the fix to the main and 2.4 branches.
Please let me know it it fixes your problem.
You should be avare that even though the TM timeout is
set extremely high, there may be other timeouts in the
transactional resources that could still time out.
If a transactional resource times out before the
transaction you will end up with a heuristic.
So be careful about which resources you use, and if
possible raise their timeouts so they will not start
making heuristic decisions too soon.
Best Regards,
Ole Husgaard.
_______________________________________________
Jboss-development mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/jboss-development
