In a distributed transaction -- one in which "DATABASE A" calls "DATABASE B", only "DATABASE A" may commit. The reason is "database B" has no mechanism for co-ordinating with "database a" on the commit. We need to do a 2 phase commit to ensure that when the transaction commits any work performed on DATABASE A is committed *as well as* any work on database b.

Because the problem is more complex in general when database A calls both database B and database C. Here, if database B (which has NO IDEA that database C is even involved) attempted to commit -- it would not be able to coordinate with database A and having no knowledge of the existence of database C in the transaction -- could not coordinate it either.

A common workaround is an optional parameter to the existing routine -- a commit flag (make it be something like:

p_do_not_commit boolean in default FALSE );

You will then seek out all commit and rollback statements in the affected routine and rewrite them as:

if ( NOT p_do_not_commit ) then
commit;
end if;

so the routine can be called locally or remotely, and the parameter set accordingly.