|Oracle® Database Administrator's Guide
11g Release 2 (11.2)
Part Number E17120-11
|PDF · Mobi · ePub|
When you issue a SQL statement, the database attempts to lock the resources needed to successfully execute the statement. If the requested data is currently held by statements of other uncommitted transactions, however, and remains locked for a long time, a timeout occurs.
Consider the following scenarios involving data access failure:
A DML statement that requires locks on a remote database can be blocked if another transaction own locks on the requested data. If these locks continue to block the requesting SQL statement, then the following sequence of events occurs:
A timeout occurs.
The database rolls back the statement.
ORA-02049: time-out: distributed transaction waiting for lock
Because the transaction did not modify data, no actions are necessary as a result of the timeout. Applications should proceed as if a deadlock has been encountered. The user who executed the statement can try to reexecute the statement later. If the lock persists, then the user should contact an administrator to report the problem.
A query or DML statement that requires locks on a local database can be blocked indefinitely due to the locked resources of an in-doubt distributed transaction. In this case, the database issues the following error message:
ORA-01591: lock held by in-doubt distributed transaction identifier
In this case, the database rolls back the SQL statement immediately. The user who executed the statement can try to reexecute the statement later. If the lock persists, the user should contact an administrator to report the problem, including the ID of the in-doubt distributed transaction.
The chances of these situations occurring are rare considering the low probability of failures during the critical portions of the two-phase commit. Even if such a failure occurs, and assuming quick recovery from a network or system failure, problems are automatically resolved without manual intervention. Thus, problems usually resolve before they can be detected by users or database administrators.