--------------Original Comment History----------------------------
Comment from = ScottKurz on 2016-05-25 20:45:26 +0000
The reasons for wanting to possibly avoid a global tran around open and close are distinct from each other, but it might be desirable to consider adding any new configuration in some similar, parallel way.
For close, just because we fail closing one of the reader/writer, do we really want to rollback any updates made to the other? What are we trying to coordinate here? Some overlap with Bug 5760.
For open, the desire to avoid a global tran on open may extend to wanting to avoid a global tran during each chunk as well. Not sure if we'll get to considering that. Another more narrow case would be simply to bypass the normal transaction timeout which would otherwise hit.
Originally opened as bug 7684 by ScottKurz
--------------Original Comment History---------------------------- Comment from = ScottKurz on 2016-05-25 20:45:26 +0000
The reasons for wanting to possibly avoid a global tran around open and close are distinct from each other, but it might be desirable to consider adding any new configuration in some similar, parallel way.
For close, just because we fail closing one of the reader/writer, do we really want to rollback any updates made to the other? What are we trying to coordinate here? Some overlap with Bug 5760.
For open, the desire to avoid a global tran on open may extend to wanting to avoid a global tran during each chunk as well. Not sure if we'll get to considering that. Another more narrow case would be simply to bypass the normal transaction timeout which would otherwise hit.
(Note bug 5760 is now https://github.com/eclipse-ee4j/batch-api/issues/52)