This occurs when two Sql Server processes are accessing the same resources, but in a different order. Therefore they end up both waiting for the other process, which is a deadlock.
There are a number of ways to prevent it, including:
- Avoid taking unneccessary locks. Review the transaction isolation level required for the query, use
with (nolock)locking hint for queries where appropriate. - Make sure that when taking locks you take locks on objects in the same order in each query.
E.g. if Proc1 locks table1 and then table2, but Proc2 locks table2 and then table1, the problem can arise. You can rewrite either proc to take locks in the same order to avoid this problem.