On Tue, 2006-01-03 at 17:38 -0500, Bruce Momjian wrote:
> Right, the DML will be single-threaded and fsync of all dirty pages will
> happen before commit of each transaction.
heap_sync() would occur at end of statement, as it does with CTAS. We
could delay until EOT but I'm not sure I see why; in most cases they'd
be the same point anyway.
I'd been toying with the idea of making the freshly added blocks live
only in temp_buffers to avoid the shared_buffers overhead, but that was
starting to sounds too wierd for my liking.
Best Regards, Simon Riggs