Re: TOAST performance (was Re: [GENERAL] Delete Performance)

> Bruce Momjian <pgman(at)candle(dot)pha(dot)pa(dot)us> writes:
> > Could we add code to detect a partial write when we recover from one
> > using WAL so we can know if these partial writes are ever
> > happening?
>
> What's your point? It clearly *can* happen during power-failure
> scenarios. All the monitoring in the world won't disprove that.

My point is uh, um, eh, I think it is a very important point that I
should make ... um. :-)

Seriously, how do OS's handle partial page write, especially to
directories?

Another item I was considering is that INSERT and UPDATE, because they
append to the tables, don't really cause lots of pre-page writes, while
DELETE could affect all page in a table and would require pre-page
writes on all of them.

However, deletes are only marking the XID status of the rows.
Unfortunately I can't think of a way of recording those new XID's in WAL
and preventing a possible failure while the XID's are written to the
page. Can someone help me here?

--
Bruce Momjian | http://candle.pha.pa.us
pgman(at)candle(dot)pha(dot)pa(dot)us | (610) 853-3000
+ If your life is a hard drive, | 830 Blythe Avenue
+ Christ can be your backup. | Drexel Hill, Pennsylvania 19026