Gordan Bobic <gordan@bobich.net> writes:
> I'm going to have to rant now. The "dump" and "restore" which use the COPY
> method are actually totally useless for large databases. The reason for
> this is simple. Copying a 4 GB table with 40M rows requires over 40GB of
> temporary scratch space to copy, due to the WAL temp files. That sounds
> totally silly. Why doesn't pg_dump insert commits every 1000 rows or so???
Yup, "rant" is the correct term. The WAL file growth problem has been
fixed as of 7.1.3, so your premise is false.
regards, tom lane