Re: [Patch] Optimize dropping of relation buffers using dlist

On Wed, Sep 23, 2020 at 8:04 AM tsunakawa(dot)takay(at)fujitsu(dot)com
<tsunakawa(dot)takay(at)fujitsu(dot)com> wrote:
>
> From: Amit Kapila <amit(dot)kapila16(at)gmail(dot)com>
> > > > > Don't
> > > > > we need to guarantee the cache to be valid while recovery?
> > > > >
> > > >
> > > > One possibility could be that we somehow detect that the value we
> > > > are using is cached one and if so then only do this optimization.
> > >
> > > I basically like this direction. But I'm not sure the additional
> > > parameter for smgrnblocks is acceptable.
> > >
> > > But on the contrary, it might be a better design that
> > > DropRelFileNodeBuffers gives up the optimization when
> > > smgrnblocks(,,must_accurate = true) returns InvalidBlockNumber.
> > >
> >
> > I haven't thought about what is the best way to achieve this. Let us see if
> > Tsunakawa-San or Kirk-San has other ideas on this?
>
> I see no need for smgrnblocks() to add an argument as it returns the correct cached or measured value.
>

The idea is that we can't use this optimization if the value is not
cached because we can't rely on lseek behavior. See all the discussion
between Horiguchi-San and me in the thread above. So, how would you
ensure that if we don't use Kirk-San's proposal?

--
With Regards,
Amit Kapila.