On 2013-05-30 15:34:04 +0300, Heikki Linnakangas wrote:
> On 30.05.2013 15:12, Andres Freund wrote:
> >Now, I am far from being convinced its a good idea to get rid of
> >PD_ALL_VISIBLE, but I don't think it does. Except that it currently is
> >legal for the page level ALL_VISIBLE being set while the corresponding
> >visibilitymap one isn't there's not much prohibiting us fundamentally
> >from looking in the vm when we need to know whether the page is all
> >visible, is there?
>
> Hmm, so you're suggesting that the visibility map would be *required* to
> interpret the pages correctly. Ie. if you just zapped the visibility map,
> you'd lose critical information and the heap would appear to be corrupt. I
> guess that's possible, but it makes me quite uneasy. At the moment, it's
> relieving to know that it's always safe to just truncate the visibility map
> in case of emergency.
I didn't say its a good idea, just that I don't think Robert's
conclusion is necessarily valid. But requiring only the few kbytes of
the vm to be written instead of all of the heap during freezeing (or
whatever we would call it) has quite some allure, I admit that.
But I think that should be a separate project to reeingineering how
freezing works.
Greetings,
Andres Freund
-- Andres Freund http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training &
Services