Re: pgsql: Delay commit status checks until freezing executes.

On Wed, Jan 4, 2023 at 10:41 AM Andres Freund <andres(at)anarazel(dot)de> wrote:
> > It's currently possible for VACUUM to set the all-frozen bit while
> > unsetting the all-visible bit, due to a race condition [1]. This is
> > your long standing bug. So apparently nobody is qualified to commit
> > patches in this area.
>
> That's a non-sequitur. Bugs are a fact of programming.

I agree.

> > About a year ago, there was a massive argument over some earlier work
> > in the same general area, by me. Being the subject of a pile-on on
> > this mailing list is something that I find deeply upsetting and
> > demoralizing. I just cannot take much more of it. At the same time,
> > I've made quite an investment in the pending patches, and think that
> > it's something that I have to see through.
>
> I'm, genuinely!, sorry that you feel piled on. That wasn't, and isn't, my
> goal.

Apology accepted.

I am making a simple, practical point here, too: I'm much too selfish
a person to continue to put myself in this position. I have nothing to
prove, and have little to gain over what I'd get out of working in
various other areas. I wasn't hired by my current employer to work on
VACUUM in particular. In the recent past I have found ways to be very
productive in other areas, without any apparent risk of protracted,
stressful fights -- which is something that I plan on getting back to
soon. I just don't have the stomach for this. It just isn't worth it
to me.

> I think the area of code desperately needs work. I complained because I
> didn't like the process and was afraid of the consequences and the perceived
> need on my part to do post-commit reviews.

The work that I did in 15 (in particular commit 0b018fab, the "oldest
extant XID" commit) really isn't very useful without the other patches
in place -- it was always supposed to be one piece of a larger whole.
It enables the freezing stuff because VACUUM now "gets credit" for
proactive freezing in a way that it didn't before. The motivating
examples wiki page shows examples of this [1].

Once the later patches are in place, the 15/16 work on VACUUM will be
complete, and I can walk away from working on VACUUM having delivered
a very useful improvement to performance stability -- a good outcome
for everybody. If you and Robert can find a way to accommodate that,
then in all likelihood we won't need to have any more heated and
protracted arguments like the one from early in 2022. I will be quite
happy to get back to working on B-Tree, likely the skip scan work.

[1] https://wiki.postgresql.org/wiki/Freezing/skipping_strategies_patch:_motivating_examples
--
Peter Geoghegan