Thank you for having replied Sacha. I'll not provide the exception stack
trace as we've found what was causing the problem.
I'll post our finding here to end this thread with the explanation for our
problem (well, at least is what we suspect it was the problem. I'm sorry if
what I'm saying is not 100% correct, but it is our conclusion so far)
It was all about the design we were using to do the recovery:
Supose we had an EJB1 and a EJB2. Both ejb's have a recovery method
"recover()" to fix the transient parts that were lost during failover. That
recover method is called whenever we access an transient variable by an
accessor we have in the class:
class ejb1 {
transient xxx transientVariable;
getTransientVariable()
if (transientVariable== null) revover();
return transientVariable;
}
recover () { // Get the transient variables in a valid state again. }
method1()
// ...
getTransientVariable().doSomething();
}
method2() {
// ... access ejb2 and invoque someMethod();
}
}
ejb2 has a similar approach, but it had a difference: The recover method had
to access ejb1 to get a value.
our client (servlet) would, for instance, invoque ejb1.method2();
this would make a call to ejb2, so we have ejb1 invoquing ejb2.
ejb2.method2() would have the need to access a transient variable so it
would try to recover. In particular that transient variable was provided by
calling an ejb1 method. It has an reference to ejb1 that was kept safe
during failover, so it can invoque it something like this:
(ejb2) recover() {
// need to ask ejb1 for the correct value of some transient variable
that I have:
ejb1.giveMeSomeVariable();
}
so ejb1 is calling ejb2 that, when recovering, needs to call ejb1 again..
and... well "concurrent call on statefull beans"...
well, I actually thing that in some occasions it should be possible to use
this... In particular, if there existed a "const method()" in java (there is
no "const" is java, is there?) I think it should be possible. This would say
that the method is harmless to the state of the bean... But this is another
story...
Hope maybe our bad (old) design will help someone to avoid this mistakes :-)
Joao Clemente
----- Original Message -----
From: "Joao Clemente" <[EMAIL PROTECTED]>
Sent: Monday, July 28, 2003 4:36 PM
> We're changing our app so that we can recover the transient parts of our
> objects so that we can eventually support failover.
> One of our recovery methods fails with
> "Application error: no concurrent calls on stateful beans"
> and we're being unable to trace what is responsible for this behaviour. We
> are using a single client so there should be only one call, not two.
> One thing I am thinking is that maybe the server is activating the ejb and
> at the same time it invokes the recovery method that we wrote.
>
> Maybe one of you has already faced this situation and has suggestions.
> We're using jboss 3.0.8
-------------------------------------------------------
This SF.Net email sponsored by: Free pre-built ASP.NET sites including
Data Reports, E-commerce, Portals, and Forums are available now.
Download today and enter to win an XBOX or Visual Studio .NET.
http://aspnet.click-url.com/go/psa00100003ave/direct;at.aspnet_072303_01/01
_______________________________________________
JBoss-user mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/jboss-user
