LPS is also an important feature to have for connecting to CICS using cicseci.rar,
which is not XA-capable. Thus, at least one CICS transaction may participate within a
global transaction.
I'm not sure whether Websphere (which does support LPS) is capable of the proper
transaction logging/recov
Exactly ! It's not perfect, but does the work in most cases. For example combined JMS
and database transaction management with some 'more or less' popular Non-XA RDBMS (
e.g. 'MySQL' :-) would greatly benefit from this little extension !
View the original post :
http://www.jboss.org/index.html?
btw LRCO strikes me as a small benefit, almost a hack.
So in the case ONE resource doesn't support XA the XA protocol still works if that
resource is last. (woo hoo!)
but heck if we can implement it for cheap why not
View the original post :
http://www.jboss.org/index.html?module=bb&op=viewto
Patrick,
High end JTA with all the bells and whistles is one of these things that is used by 20
people in the world and they are eager to pay for it since they got the money.
Open Source satisfies the masses. Proprietary fills those expensive niches.
Arjuna's business and sole focus is that hi
I do think that it is of serious worth, because it's better that nothing :-) Or am I
wrong ?
Annotation
Further information on XA and a strategie for including a Non-XA-Resource in a XA
transaction ( look for "Doing XA w/out an XA driver: the Last-Resource Gambit) can
also be found in a good
Mark is correct, the default tm that comes with jboss 3.x does not support LRCO
It has been implemented for jboss4, I need to do some more testing on it
before backporting to 3.2.x
But since we don't do any logging/recovery either,
it is open to debate whether the optimization is of any serious w
What I described as LRCO is what you've quoted from IBM and Sun. Just some people call
them Resources, whereas others call them Participants :-)
The principle is the same: the participant/resource that can't do 2PC has to be the
*last* one in the first phase to go off and *only* if all other res
I'm not sure if LRCO solves the problem. What we need is to have the non-XA (in an XA
wrapper) to perform its commit (prepare and commit for the wrapper) at the end of
phase 1. This means that it is the last resource for phase 1 but the first resource
for phase 2. This way, the resource can b
I assume what you're after is the Last Resource Commit Optimization, but there may be
a couple of things mixed in here.
Firstly, you don't have a choice as to what kind of resources are registered with
JBoss TM - they have to be implementations of the XAResource interface since you're
coming in