Re: Refactoring the checkpointer's fsync request queue

On Wed, Mar 13, 2019 at 2:00 PM Shawn Debnath <sdn(at)amazon(dot)com> wrote:
> So ... wondering if there are any other left over items for this patch
> or is it good to go? I imagine there's at least a couple of us who would
> love to see this get in for PG12.

I rebased my WIP undo stuff[1] (targeting 13) on top of this, and that
seemed to go smoothly and the interfaces made sense, which was
reassuring. I do wonder if we'll need to expose a way for eg
pg_rewind and perhaps external backup tools to find paths and offsets
given WAL block references that might in future include an SMGR ID
(well that's one proposal), but that handwavy requirement doesn't seem
to conflict with anything we've done here. I'm planning to do another
round of review and testing. Aside from some refactoring which I
think looks good anyway and prepares for future patches, the main
effect of this patch is to force the checkpointer to open and close
the files every time which seems OK to me. I know Andres wants to
make a pass through it too.

[1] https://www.postgresql.org/message-id/CA%2BhUKGKN42jB%2BubCKru716HPtMbahdia39GwG5pLgWLMZ_y1ng%40mail.gmail.com

--
Thomas Munro
https://enterprisedb.com