Re: Background writer and checkpointer in crash recovery

On Sat, Jul 31, 2021 at 2:16 AM Robert Haas <robertmhaas(at)gmail(dot)com> wrote:
> On Fri, Jul 30, 2021 at 4:42 AM Aleksander Alekseev
> <aleksander(at)timescale(dot)com> wrote:
> > v2-0001 and v2-0002 look fine, but I don't like much the idea of introducing a new GUC in v2-0003. It's for very specific needs, which most of the users, I believe, don't care about. I suggest dealing with v2-0001 and v2-0002 first and then maybe submit and discuss v2-0003 as a separate CF entry.

Thanks.

> Thanks for bumping this thread; I had forgotten all about this effort,
> but having just spent a bunch of time struggling with the thicket of
> cases in StartupXLOG(), I'm now feeling highly motivated to make some
> more progress in simplifying things over there. I am still of the
> opinion that 0001 is a good idea, and I don't have any suggestions for
> how it could be improved,

That's good news, and thanks. Yes, clearly there is much more that
can be simplified here.

> except perhaps that the call to
> PublishStartupProcessInformation() could maybe have a one-line
> comment.

Done. BTW that is temporary, as I'm planning to remove that machinery soon[1].

> Thomas, are you planning to press forward with committing
> this soon? If not, do you mind if I do?

I pushed 0001. Let me think about 0002, and flesh out 0003 a bit more.

[1] https://www.postgresql.org/message-id/flat/CA+hUKGLYRyDaneEwz5Uya_OgFLMx5BgJfkQSD=q9HmwsfRRb-w(at)mail(dot)gmail(dot)com