On 2013-11-30 12:22:16 -0500, Noah Misch wrote:
> On Sat, Nov 30, 2013 at 05:00:58PM +0100, Andres Freund wrote:
> > The problem with log_newpage_buffer() is that we'd quite possibly issue
> > one such call per item on a page. And that might become quite
> > expensive. Logging ~1.5MB per 8k page in the worst case sounds a bit
> > scary.
>
> I had in mind issuing at most one call per page. heap_page_prune() has a
> structure conducive to that.
That, at least as far as I can imagine, would make the fix quite
complicated though. In the first phase heap_page_prune() we aren't in a
critical section and cannot modify the buffer yet, so we would make all
the involved code cope with the unfixed xids and hint bits.
Greetings,
Andres Freund
-- Andres Freund http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training &
Services