On Mon, Mar 20, 2023 at 09:30:50AM +0900, Michael Paquier wrote: > Did you check dump and restore flows with partition > trees and --no-table-access-method? Perhaps there should be > some regression tests with partitioned tables?
I was looking at the patch, and as I suspected the dumps generated
are forgetting to apply the AM to the partitioned tables. For
example, assuming that the default AM is heap, the patch creates
child_0_10 with heap2 as AM, which is what we want:
CREATE ACCESS METHOD heap2 TYPE TABLE HANDLER heap_tableam_handler;
CREATE TABLE parent_tab (id int) PARTITION BY RANGE (id) USING heap2;
CREATE TABLE child_0_10 PARTITION OF parent_tab
FOR VALUES FROM (0) TO (10);
However a dump produces that (content cut except for its most relevant
parts):
CREATE ACCESS METHOD heap2 TYPE TABLE HANDLER heap_tableam_handler;
SET default_tablespace = '';
CREATE TABLE public.parent_tab (
id integer
)
PARTITION BY RANGE (id);
SET default_table_access_method = heap2;
CREATE TABLE public.child_0_10 (
id integer
);
This would restore the previous contents incorrectly, where parent_tab
would use heap and child_0_10 would use heap2, causing any partitions
created after the restore to use silently heap. This is going to
require a logic similar to tablespaces, where generate SET commands
on default_table_access_method so as --no-table-access-method in
pg_dump and pg_restore are able to work correctly. Having tests to
cover all that is a must-have.
--
Michael
signature.asc
Description: PGP signature
