Hi!
> ok so 300 threads all want to access a bean
>
> 300 threads go through TxInterceptorCMT and 300 threads start a new Tx....
>
> 1 thread goes through at a time and 299 sleep waiting... (we loop in each
> thread).
>
> Eventually the Txs timeout (with a big load it does) and the threads are
> sleeping.
>
> When the thread wakes up it goes on with a "DEAD" Tx and on commit we get
a
> "NOTX"
>
> bad bad bad....
>
> ideas?
In the InstanceInterceptor we could add a check if the calling thread not
only has a tx, but a live tx. If the tx is dead then throw an exception so
that the rollback functionality is used.
This should be right after the do { I think: have to be done on each loop,
and as soon as possible.
/Rickard
