oracle insert into very slow

Question

I have the following Oracle query, that executes quickly (in a few seconds):

select contract_id_fk as ct,
       max(trip_id) as tid,
       max(consumed_mileage) as cum
from trip
where to_date > to_date('20-12-2016','DD-MM-YYYY')
and contract_id_fk is not null
and vehicle_id_fk is not null
and trip_stop_status is null
group by contract_id_fk

the 'trip' table has a large number of row (more than 20 million) Now, I want to insert the result of this query in a table, using:

INSERT INTO lst
select contract_id_fk as ct, 
       max(trip_id) as tid, 
       max(consumed_mileage) as cum 
from trip
where to_date > to_date('20-12-2016','DD-MM-YYYY') 
and contract_id_fk is not null 
and vehicle_id_fk is not null 
and trip_stop_status is null 
group by contract_id_fk

This is extremely slow. So slow that the transaction times out (more than 30 seconds in my case). Any idea why this is so slow, and how it could be optimized?

Answer 1

It's very hard to say what's the reason, as there could be multiple things, such as:

• insufficient memory, or redo log size
• suboptimal table settings (PCTFREE etc)
• hardware performance issue

Two things I'd recommend:

• check what are the wait events
• try insert with APPEND hint

INSERT /*+ APPEND */ INTO lst
select contract_id_fk as ct,
max(trip_id) as tid, max(consumed_mileage) as cum
from trip where to_date > to_date('20-12-2016','DD-MM-YYYY')
and contract_id_fk is not null and vehicle_id_fk is not null and
trip_stop_status is null group by contract_id_fk

Answer 2

Try to select with the cursor, and insert in the loop. Like that;

DECLARE
       CURSOR SCURSOR IS
        select contract_id_fk as ct, 
           max(trip_id) as tid, 
           max(consumed_mileage) as cum 
        from trip
        where to_date > to_date('20-12-2016','DD-MM-YYYY') 
        and contract_id_fk is not null 
        and vehicle_id_fk is not null 
        and trip_stop_status is null 
        group by contract_id_fk ;
    BEGIN
       FOR RECS IN SCURSOR
       LOOP
            INSERT INTO lst 
            SELECT RECS.ct , RECS.tid , RECS.cum FROM DUAL;
            COMMIT;
       END LOOP;
    END;

Answer 3

检查缓冲区现金，我认为问题可能出在缓冲区现金或尝试收集表统计信息

Answer 4

As they are still answers being made for this almost 4 years old question ... The root cause for the slowness for a saturation in IO (RDS instance on AWS...) plus way too many group by/sort. More a database modeling problem than anything else. Thank you for all those that tried to help!