On Mon, 2007-07-23 at 17:56 +0200, Csaba Nagy wrote:
> Now I don't put too much hope I can convince anybody that the limit on
> the delete/update commands has valid usage scenarios, but then can
> anybody help me find a good solution to chunk-wise process such a buffer
> table where insert speed is the highest priority (thus no indexes, the
> minimum of fields), and batch processing should still work fine with big
> table size, while not impacting at all the inserts, and finish in short
> time to avoid long running transactions ? Cause I can't really think of
> one... other than our scheme with the delete with limit + trigger +
> private temp table thing.
Use partitioning: don't delete, just drop the partition after a while.
--
Simon Riggs
EnterpriseDB http://www.enterprisedb.com