On Wed, Apr 11, 2018 at 1:33 PM, Adrian Klaver <adrian.klaver@xxxxxxxxxxx> wrote:
A test case here confirms it sends individual INSERTS:
test_(postgres)# insert into fdw_test_table select * from fdw_test;
INSERT 0 3
Where fdw_test_table is the remote table and fdw_test is the local one.
postgres-2018-04-11 11:29:23.812 PDT-0LOG: statement: insert into fdw_test_table select * from fdw_test;
postgres-2018-04-11 11:29:23.812 PDT-0LOG: execute pgsql_fdw_prep_2: INSERT INTO public.fdw_test_rcv(id, fld_1, fld_2) VALUES ($1, $2, $3)
postgres-2018-04-11 11:29:23.812 PDT-0DETAIL: parameters: $1 = '1', $2 = 'one', $3 = 't'
postgres-2018-04-11 11:29:23.813 PDT-10140LOG: execute pgsql_fdw_prep_2: INSERT INTO public.fdw_test_rcv(id, fld_1, fld_2) VALUES ($1, $2, $3)
postgres-2018-04-11 11:29:23.813 PDT-10140DETAIL: parameters: $1 = '2', $2 = 'two', $3 = 'f'
postgres-2018-04-11 11:29:23.813 PDT-10140LOG: execute pgsql_fdw_prep_2: INSERT INTO public.fdw_test_rcv(id, fld_1, fld_2) VALUES ($1, $2, $3)
postgres-2018-04-11 11:29:23.813 PDT-10140DETAIL: parameters: $1 = '3', $2 = 'three', $3 = 'f'
So much for that idea(:=
Yeah, I saw the same with a 132 row insert. Now imagine that with a monthly 50 million row insert or delete. :p Thanks for the confirmation!
I'm definitely leaning towards the copy/load/delete method.
Don.
Don Seiler
www.seiler.us
www.seiler.us
