Data update for 76 million rows

Question

Hello All,
we have added a new column into one of our tables and we need to populate the column.Problem is it has 76 million rows.If i issue the update coomand to populate this ,approximately it will take 120 hrs to complete.Please let me know the best way of doing this. Can i run it in batches applying commit in between??
Thanks.

Answer 1

It´d be something like this:
DECLARE
  V_QRY          VARCHAR2(10000);
  V_COMMIT_RANGE INTEGER;
BEGIN
  V_COMMIT_RANGE := 10000; -- Change this according with your environment
  * Prevents exceeding ROLL BACK segments.
  LOOP
    V_QRY := 'UPDATE transaction_fact a '
          || '   SET a.start_time = (SELECT TO_DATE ( TO_CHAR (:v_year        , ''fm0000'') '
          || '                                     || TO_CHAR (:v_month       , ''fm00'') '
          || '                                     || TO_CHAR (:v_day_of_month, ''fm00''), '
          || '                                     ''YYYYMMDD'' '
          || '                                    ) '
          || '                       FROM TIME '
          || '                      WHERE TIME.time_id = a.time_id) '
          || ' WHERE a.start_time IS NULL '
          || '   AND ROWNUM <= ' || V_COMMIT_RANGE;
    EXECUTE IMMEDIATE V_QRY
                USING YEAR
                    , MONTH
                    , day_of_month;
    EXIT WHEN SQL%ROWCOUNT = 0;
    COMMIT;
  END LOOP;
EXCEPTION
  WHEN NO_DATA_FOUND THEN
    DBMS_OUTPUT.PUT_LINE('no content');
  WHEN OTHERS THEN
    DBMS_OUTPUT.PUT_LINE(SQLERRM);
END;Assumptions made:
a) YEAR, MONTH and day_of_month are all variables;
b) a.start_time has null values for all rows (or all you wish to update);
Although this will do the job, it might not be as performatic as you need it to be. So, if your Oracle version allows you to rename tables, you should consider looking into Walter´s post (if you havent already).
All the best..