Re: GSoC proposal - "make an unlogged table logged"

On 2014-03-04 01:10:50 -0300, Fabrízio de Royes Mello wrote:
> Today I do something like that:
> 
> 1) create unlogged table tmp_foo ...
> 2) populate 'tmp_foo' table (ETL scripts or whatever)
> 3) start transaction
> 4) lock table tmp_foo in access exclusive mode
> 5) update pg_class set relpersistence = 'p' where oid = 'tmp_foo':regclass
> 6) drop table foo; -- the old foo table
> 7) alter table tmp_foo rename to foo;
> 8) end transaction
> 9) run pg_repack in table 'foo'
> 
> I know it's very ugly, but works... and works for standbys too... :-)

No, it doesn't work. It just may happen to not fail loudly/visibly in
some cases. You're absolutely risking corruption of this *and* other
relations when doing so.

Greetings,

Andres Freund

-- Andres Freund                       http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training &
Services