On Fri, Jan 13, 2023 at 9:55 PM Andres Freund <andres@anarazel.de> wrote:
> How about a float autovacuum_no_auto_cancel_age where positive values are
> treated as absolute values, and negative values are a multiple of
> autovacuum_freeze_max_age? And where the "computed" age is capped at
> vacuum_failsafe_age? A "failsafe" autovacuum clearly shouldn't be cancelled.
>
> And maybe a default setting of -1.8 or so?
Attached is a new revision, v5. I'm not happy with this, but thought
it would be useful to show you where I am with it.
It's a bit awkward that we have a GUC (autovacuum_no_auto_cancel_age)
that can sometimes work as a cutoff that works similarly to both
freeze_max_age and multixact_freeze_max_age, but usually works as a
multiplier. It's both an XID age value, an MXID age value, and a
multiplier on XID/MXID age values.
What if it was just a simple multiplier on
freeze_max_age/multixact_freeze_max_age, without changing any other
detail?
--
Peter Geoghegan