Francisco Reyes wrote:
> I knew it sounded too good to be true.
> 1- The trigger was not set in the master (ie nothing went to the children).
> 2- The master had no index and no RI.. so it was a straight insert.
>
> I corrected (ie set the trigger in the master and RI in the children).
> Has been running for 10 hours and has not finished.
FWIW it tends to be faster to do the bulk load first and add the
indexes and constraints later. (Though obviously you must be prepared
to cope with the failing rows, if any). However, if you do this
INSERT/SELECT thing frequently, this is probably not very workable.
--
Alvaro Herrera http://www.CommandPrompt.com/
The PostgreSQL Company - Command Prompt, Inc.