Re: UPDATEDs slowing SELECTs in a fully cached database

"Kevin Grittner" <Kevin(dot)Grittner(at)wicourts(dot)gov> writes:
> ... Jeff does raise a good point, though -- it seems odd
> that WAL-logging of this pruning would need to be synchronous.

Yeah, we need to get to the bottom of that. If there's enough
shared_buffer space then it shouldn't be.

> We
> support asynchronous commits -- why not use that feature
> automatically for transactions where the only writes are this sort
> of thing. Which raises an interesting question -- what happens to
> the timings if your SELECTs are done with synchronous_commit = off?
> I wonder if it would make any sense to implicitly use async commit
> for a transaction which is declared READ ONLY or which never
> acquires and XID?

Huh? If there was never an XID, there's no commit WAL record, hence
nothing to make asynchronous.

regards, tom lane