I have two entities User and UserInfo where User has N UserInfos, for instance. User resides in one DB and UserInfo in another, hence two different datasources.
I have got this going with no hitches but noticed a couple of things when configuring it.
1) I am specifying the at the bean level. If I do not specify a in my jbosscmp-jdbc.xml descriptor that specifies that the default datasource to be that of UserInfo (child), then the CMR lookup from User -> UserInfo will fail. If I specify the defaults to be that for the UserInfo, then all is well. I found this by looking at the stack and seeing that org.hsql classes were being used in the CMR lookup? Is this a bug?
2) Will I get proper TX semantics, or will I need to use an XA TM?