Problem:
Oracle Connector Stage commits data even when a IBM DataStage job aborts and record count is zero.
Symptom:
Oracle Connector Stage commits data even when a IBM Data Stage job aborts and record count is zero.
Cause:
This can happen when multiple nodes are used to run DataStage jobs.
Resolving the problem:
This is expected behavior if you are using 2 or more nodes to run the DataStage jobs.
Transaction support within the Oracle Connector stage itself simply ensures that records within a given transaction boundary (e.g. Record Count) are either committed or rolled back together. Earlier transactions have already been committed (including transactions from other player nodes). When one node reports error (throws an exception), the framework signals to other node processes to terminate – but the state of transactions on those nodes relative to the node that reported the error is generally unknown.
To resolve this issue, you can run the job using a single node setup – then you shouldn’t see any commits if the job fails (and record count is set to 0).
