Re: [HACKERS] pg_dump and thousands of schemas

Robert Haas <robertmhaas(at)gmail(dot)com> writes:
> On Thu, May 31, 2012 at 10:31 AM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>> I'm not; Jeff Janes is. But you shouldn't be holding your breath
>> anyway, since it's 9.3 material at this point.

> I agree we can't back-patch that change, but then I think we ought to
> consider back-patching some variant of Tatsuo's patch. Maybe it's not
> reasonable to thunk an arbitrary number of relation names in there on
> one line, but how about 1000 relations per LOCK statement or so? I
> guess we'd need to see how much that erodes the benefit, but we've
> certainly done back-branch rearrangements in pg_dump in the past to
> fix various kinds of issues, and this is pretty non-invasive.

I am not convinced either that this patch will still be useful after
Jeff's fix goes in, or that it provides any meaningful savings when
you consider a complete pg_dump run. Yeah, it will make the lock
acquisition phase faster, but that's not a big part of the runtime
except in very limited scenarios (--schema-only, perhaps).

The performance patches we applied to pg_dump over the past couple weeks
were meant to relieve pain in situations where the big server-side
lossage wasn't the dominant factor in runtime (ie, partial dumps).
But this one is targeting exactly that area, which is why it looks like
a band-aid and not a fix to me.

regards, tom lane