我如何提高此oracle sql删除查询的性能？

Question

I got to delete some unwanted rows from a table based on the result of a select query from another table

DELETE /*+ parallels(fe) */  FROM fact_x fe
WHERE fe.key NOT IN(
  SELECT DISTINCT di.key
  FROM dim_x di
  JOIN fact_y fa
  ON fa.code         = di.code
  WHERE fa.code_type = 'ABC'
 );

The inner select query returns 77 rows and executes in few milliseconds. but the outer delete query runs forever(for more than 8 hrs). I tried to count how many rows got to be deleted by converting the delete to select count(1) and its around 66.4 million fact_x rows out of total 66.8 million rows. I am not trying to truncate though. I need to retain remaining rows.

Is there any other way to achieve this? will deleting this by running a pl/sql cursor will work better?

Answer 1

Would it not make more sense just to insert the rows you want to keep into another table, then drop the existing table? Even if there are FKs to disable/recreate/etc. it almost certain to be faster.

Answer 2

Could you add a "toBeDeleted" column? The query to set that wouldn't need that "NOT IN" construction. Deleting the marked rows should also be "simple".

Then again, deleting 99,4% of the 67 million rows will take some time.

Answer 3

Try /*+ parallel(fe) */. No "S".