> On 26 Aug 2021, at 15:09, Bruce Momjian <bruce@momjian.us> wrote:
> On Thu, Aug 26, 2021 at 03:24:33PM +0800, Julien Rouhaud wrote:
>> .. I still think that
>> changing bgworker_should_start_now() is a better option.
>
> I am not sure. We have a lot of pg_upgrade code that turns off things
> like autovacuum and that has worked fine:
True, but there are also conditionals on IsBinaryUpgrade for starting the
autovacuum launcher in the postmaster, so there is some precedent.
> Basically, pg_upgrade has avoided any backend changes that could be
> controlled by GUCs and I am not sure we want to start adding such
> changes for just this.
In principle I think it’s sound to try to avoid backend changes where possible
without sacrificing robustness.
--
Daniel Gustafsson https://vmware.com/