Re: Fix checkpoint skip logic on idle systems by tracking LSN progress

Hello,

At Mon, 21 Nov 2016 14:41:27 +0900, Michael Paquier <michael(dot)paquier(at)gmail(dot)com> wrote in <CAB7nPqSetnFjhGAB+tE2M68Vc_3BwbsEPe+dCMB8xnH0UYw3aA(at)mail(dot)gmail(dot)com>
> On Mon, Nov 21, 2016 at 1:31 PM, Kyotaro HORIGUCHI
> <horiguchi(dot)kyotaro(at)lab(dot)ntt(dot)co(dot)jp> wrote:
> > So, all my original concern were cleared.
>
> Cool. Perhaps this could be marked as ready for committer then?

^^;

> > The last one is
> > resetting by a checkpointer restart.. I'd like to remove that if
> > Andres agrees.
>
> Could you clarify this point? v18 makes sure that the last segment
> switch stays in shared memory so as we could still skip the activity
> of archive_timeout correctly.

I don't doubt that it works. (I don't comment on the comment:) My
concern is complexity. I don't think we wish to save almost no
harm behavior caused by a thing rarely happens. But, if you and
others on this thread don't mind the complexity, It's not worth
asserting myself more.

So, after a day waiting, I'll mark this as ready for committer
again.

reagards,

--
Kyotaro Horiguchi
NTT Open Source Software Center