On Wed, 23 Sep 2015 11:46:00 -0400
Robert Haas <robertmhaas@gmail.com> wrote:
> On Wed, Sep 23, 2015 at 11:22 AM, Alvaro Herrera
> <alvherre@2ndquadrant.com> wrote:
> > Robert Haas wrote:
> >> On Tue, Sep 22, 2015 at 5:16 AM, Ildus Kurbangaliev
> >> <i.kurbangaliev@postgrespro.ru> wrote:
> >> > Yes, probably.
> >> > I'm going to change API calls as you suggested earlier.
> >> > How you do think the tranches registration after initialization
> >> > should look like?
> >>
> >> I don't see any need to change anything there. The idea there is
> >> that an extension allocates a tranche ID and are responsible for
> >> making sure that every backend that uses that tranche finds out
> >> about the ID that was chosen and registers a matching tranche
> >> definition. How to do that is the extension's problem. Maybe
> >> eventually we'll provide some tools to make that easier, but
> >> that's separate from the work we're trying to do here.
> >
> > FWIW I had assumed, when you created the tranche stuff, that SLRU
> > users would all allocate their lwlocks from a tranche provided by
> > slru.c itself, and the locks would be stored in the slru Ctl
> > struct. Does that not work for some reason?
>
> I think that should work and that it's a good idea. I think it's just
> a case of nobody having done the work.
>
There is a patch that splits SLRU LWLocks to separate tranches and
moves them to SLRU Ctl. It does some work from the main patch from
this thread, but can be commited separately. It also simplifies
lwlock.c.
--
Ildus Kurbangaliev
Postgres Professional: http://www.postgrespro.com
Russian Postgres Company
