During normal B-tree traversal, leaf and data pages are logically locked until the end of the transaction. B-tree index pages are only temporarily locked during query execution. The index page lock is released after the page has been searched.

When searching the B-tree index, ladder locking is used: a lock is taken on the first index page, which points to another index page. The next index page is locked and, once it is locked, the first lock is dropped, and so on down the index to the leaf level.

The locking system always locks the leaf and data pages when accessing B-tree tables. Because of this, locking in a B-tree table requires twice as many locks as locking an ISAM or hash table. It is wise to set the maxlocks escalation factor higher than the default when using the B-tree storage structure. For details, see the SET LOCKMODE statement in the SQL Reference Guide.