On 1/3/07, David Jencks <[EMAIL PROTECTED]> wrote:
On Jan 3, 2007, at 10:56 AM, Mike Perham wrote: > I'm using the new 1.2-beta with the mysql-xa connector from wasce > 1.1.0.1. My application does some db work when it starts up and we > don't see this error on 1.1.0.1. I'm using Spring 1.2.8's > JTATransactionManager like this: > > <bean id="inboxTransactionManager" > class="org.springframework.transaction.jta.JtaTransactionManager"> > <property name="transactionManagerName"><value>java:/ > TransactionManager</value></property> > </bean> > > Is this the correct JNDI location? I would have expected it to be java:TransactionManager but it's possible both values work. I suspect it's working because apparently there is a transaction started and I assume it was started by Spring.
I assumed that value because it is JBoss's location and Geronimo seems to keep many of JBoss's choices (a good idea in order to make jboss porting easy, imo). All of our database access goes through Spring's transacation layer so your assumption is correct.
I haven't personally tried the tranql mysql xa adapter (is that the one from wasce?) but I wonder.... have you gotten it to work in other circumstances?
The application gets past this point in wasce 1.1.0.1 so yes.
One possibility is that there's a bug in the connector. When a connection is not enrolled in a transaction it's supposed have autocommit behavior. If this isn't working you probably wouldn't see a problem using a local-tx connector since there isn't much to the start of a transaction -- just turning off autocommit. If it was already off you might not see any errors. Can you put a breakpoint in a likely spot in perhaps the WrapperNamedXAResource and see if commit is actually being called?
Not easily. I don't have a geronimo dev/debug environment nor do I have the time to invest. I was given a week or so to port our app from jboss to wasce and I've shot way over that. The app does seem to work in wasce except for that hivemind classloader problem I emailed you about yesterday. That is the only thing preventing us from just running on 1.1.0.1. I'm trying 1.2-beta on your suggestion to see if this later code has the same problem but it appears I can't even get that far due to this issue.
Do you expect there to be any work done outside a transaction?
Never.
