I have a table with about 100k records and I want to delete some rows, The problem is that the DELETE
statement is running very slowly - it didn\'t finish in 30 min
When DML operations take a long time, create new table with the remaining rows and drop the previous table, instead of deleting.
I mean,
create table NEW_TABLE as
select * from daily_au_by_service_summary
where summary_ts <= to_date('09-04-2012','dd-mm-yyyy');
This will be faster, especially when you are deleting a substantial number of rows. (%10 of total rows, for example.)
Obviously, a delete operation will take longer than a select, but that doesn't account for the difference you see.
It sounds like additional code is being run upon the delete, which indecates there may be triggers on the table that are also running. Can you check this?
To delete means to change the table's content. And this means, that after each deleted row all indexes must be updated and all foreign-key references must be checked. This can take a very long time!
Maybe this helps:
Make a copy of that table without any references, triggers and additional indexes. Then do this:
insert into new_table (field1, field2, ...) values (
select field1, field2, ...
from daily_au_by_service_summary
where summary_ts < to_date('09-04-2012','dd-mm-yyyy')
);
If the fields in the tabels are defined in identical order, this might work too:
insert into new_table values (
select *
from daily_au_by_service_summary
where summary_ts < to_date('09-04-2012','dd-mm-yyyy')
);
After that:
truncate daily_au_by_service_summary
and then:
insert into daily_au_by_service_summary (field1, field2, ...) values (
select field1, field2, ...
from new_table;
);
New Table is not needed any longer:
drop new_table;
There can be many reasons:
SELECT
is fast)LOB
s, many columns).If the foreign keys are the problem, the usual solution is to add indexes on the foreign column: For each delete, Oracle needs to check whether this would violate a foreign key relation.