Bulk Update - Oracle Sequence - oracle

Question

I have a table with 150 million records
I want to update the sequence column from a sequence.
What is the fastest way to update?
I have used parallel, but it is taking hours and not ending
UPDATE /+ parallel (c, 50)/ rpm_future_retail_tmp c
SET future_retail_id = rpm_future_retail_seq.NEXTVAL;
what is the faster way?

Answer 1

Mostly creating a new table is faster than applying a DML, especially faster than an UPDATE statement. You can use as an alternative :
CREATE TABLE rpm_future_retail_tmp_ parallel 8 nologging AS
SELECT rpm_future_retail_seq.NEXTVAL AS future_retail_id,
<and the comma-separated columns other than future_retail_id>
FROM rpm_future_retail_tmp;
DROP TABLE rpm_future_retail_tmp;
ALTER TABLE rpm_future_retail_tmp_ RENAME TO rpm_future_retail_tmp;
where
the degree of parallelism might vary depending on your DBMS's source
the statements to reproduce later the privileges(grants) and indexes should
be saved to a place for the table(rpm_future_retail_tmp) before
dropping it

Answer 2

If sequence starts from 1, then you might also consider updating column with a ROWNUM value instead.
This is my 21cXE database, running on MS Windows 10, Intel i5, 8GB RAM. I have a table with ~1 million rows (don't feel like creating one 150 times larger):
SQL> select count(*) from rpm;
COUNT(*)
----------
1033616
Elapsed: 00:00:00.02
Updating it with a sequence takes ~12 seconds:
SQL> update rpm set id = seq.nextval;
1033616 rows updated.
Elapsed: 00:00:12.98
SQL> update rpm set id = seq.nextval;
1033616 rows updated.
Elapsed: 00:00:12.39
SQL> update rpm set id = seq.nextval;
1033616 rows updated.
Elapsed: 00:00:10.56
Let's try rownum; it takes less time (average of 3 runs is ~4 seconds):
SQL> update rpm set id = rownum;
1033616 rows updated.
Elapsed: 00:00:07.51
SQL> update rpm set id = rownum;
1033616 rows updated.
Elapsed: 00:00:02.89
SQL> update rpm set id = rownum;
1033616 rows updated.
Elapsed: 00:00:02.87
SQL>
I understand that your system is different from mine and timings depend on various things, but I guess it can't harm if you try another approach.
For future inserts into the ID column (via database trigger?), just (re)create the sequence:
SQL> select max(id) from rpm;
MAX(ID)
----------
1033616
SQL> drop sequence seq;
Sequence dropped.
SQL> create sequence seq start with 1033617;
Sequence created.
SQL>