Atomikos Forum

Hello,

I'm evaluating Atomikos for use at my company.

I have a simple demo spring application that starts an XA transaction, updates a database, and sends a JMS message.  My datasource is non XA, so I'm using AtomikosNonXaDataSource.

I put a breakpoint in the Atomikos source code in CoordinatorStateHandler.commit(boolean, boolean) method after the "coordinator_.setState ( TxState.COMMITTING );" line of code, and kill the application, to simulate a failure after Atomikos has decided to commit the transaction but before informing participants.

When I restart the application, I get repeated warnings in the logs, about 5 per second: "Local heuristic termination of coordinator 25.14.3.26.tm0000200001 with state HEUR_HAZARD"

This makes sense to me: it is trying to commit the transaction on the non xa datasource that has already rolled back, and so this is a heuristic hazard.

My issue is why does Atomikos effectively 'spam' the logs with 5 messages per second, can I slow this down?  Also, how I 'resolve' the heuristic hazard condition in a production environment, the only way I've found so far is to delete the transaction logs.

Thanks
David

Hi,

The non-xa datasource is viable only for at most one resource, not two like yours.

If you really want to do this then you could set the readOnly property on the datasource, this should eliminate the warnings but it has the same risks of data loss.

HTH

readOnly seems to clear up the warnings - thanks.  Actually I've managed to get my database's XA driver (ingres) working in any case.

I had only one non-xa resource: the database.  Why do you say I have two?  In fact my other XA resource I am enlisting manually (it is the weblogic transaction manager actually - which is needed to use weblogic JMS in a remote transaction manager like Atomikos).

I have another related issue - but I'll start a new topic for this.

Cheers
David