ASA Programming Guide
Three-Tier Computing and Distributed Transactions
Using distributed transactions
If the database server faults while uncommitted operations are pending, it must either rollback or commit those operations on startup to preserve the atomic nature of the transaction.
If uncommitted operations from a distributed transaction are found during recovery, the database server attempts to connect to DTC and requests that it be re-enlisted in the pending or in-doubt transactions. Once the re-enlistment is complete, DTC instructs the database server to roll back or commit the outstanding operations.
If the reenlistment process fails, Adaptive Server Anywhere has no way of knowing whether the in-doubt operations should be committed or rolled back, and recovery fails. If you want the database in such a state to recover, regardless of the uncertain state of the data, you can force recovery using the following database server options:
-tmf If DTC cannot be located, the outstanding operations are rolled back and recovery continues.
For more information, see -tmf server option.
-tmt If re-enlistment is not achieved before the specified time, the outstanding operations are rolled back and recovery continues.
For more information, see -tmt server option.