On Wed, 15 Jan 2003, Roman Fail wrote:
> The disks are not reading at max speed during the query - when I ran a
> VACUUM ANALYZE (after data migration), sar -b was consistently 100,000
> block reads/sec. It does not seem like the hardware is holding back
> things here. I read something about 'fsync' recently, would changing
> that setting apply in this case?
You ran vacuum analyze, but some of the explain still looks suspiciously
like it's using default statistics (dr and cr for example, unless they
really do have 1000 rows).
What are the actual query and table definitions for the query?