"Alexander Todorov" <alexx.todorov@gmail.com> writes:
> On 7/2/07, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> The fresh-initdb approach is more likely to work without any strange
>> corner cases. If you try a setup where the system catalogs are on
>> persistent storage but you have a tablespace on ramdisk, then after
>> restart you'll have pg_class entries referencing files that don't exist
>> anymore, which I believe will provoke errors.
> I believe error will occur if trying to access these objects. To avoid
> this pg_dump/pg_restore may be useful and recreating the
> tables/indexes after restart.
You might as well start with a freshly initdb'd cluster (all on ramdisk)
and do pg_restore from a full dump instead of a data-only dump. The
former will probably be faster as well as more foolproof.
regards, tom lane