Re: Refactoring the checkpointer's fsync request queue

On Fri, Mar 01, 2019 at 04:27:47PM -0500, Robert Haas wrote:
> On Fri, Mar 1, 2019 at 3:35 PM Shawn Debnath <sdn(at)amazon(dot)com> wrote:
> > On Fri, Mar 01, 2019 at 03:03:19PM -0500, Robert Haas wrote:
> > > On Fri, Mar 1, 2019 at 2:36 PM Shawn Debnath <sdn(at)amazon(dot)com> wrote:
> > > > I disagree, at least with combining and retaining enums. Encoding all
> > > > the possible request types with the current, planned and future SMGRs
> > > > would cause a sheer explosion in the number of enum values.
> > >
> > > How big of an explosion would it be?
> >
> > 4 enum values x # of smgrs; currently md, soon undo and slru so 12 in
> > total. Any future smgr addition will expand this further.
>
> I thought the idea was that each smgr might have a different set of
> requests. If they're all going to have the same set of requests then
> I agree with you.

Yeah, in this particular case and at this layer, the operations are
consistent across all storage managers, in that, they want to queue a
new sync request for a specific file, forget an already queued request,
forget a hierarchy of requests, or unlink a specific file.

The fun is at the smgr layer which was discussed in a sub-thread in the
"Drop type smgr" thread started by Thomas. I started on a patch and will
be sending it out after the refactor patch is revised.

--
Shawn Debnath
Amazon Web Services (AWS)