Re: Berserk Autovacuum (let's save next Mandrill)

On Wed, 2020-03-25 at 23:19 +0300, Alexander Korotkov wrote:
> On Wed, Mar 25, 2020 at 10:26 PM Andres Freund <andres@anarazel.de> wrote:
> > On 2020-03-25 11:05:21 -0500, Justin Pryzby wrote:
> > > Since we talked about how scale_factor can be used to effectively disable this
> > > new feature, I thought that scale=100 was too small and suggesed 1e10 (same as
> > > max for vacuum_cleanup_index_scale_factor since 4d54543ef).  That should allow
> > > handling the case that analyze is disabled, or its threshold is high, or it
> > > hasn't run yet, or it's running but hasn't finished, or analyze is triggered as
> > > same time as vacuum.
> > 
> > For disabling we instead should allow -1, and disable the feature if set
> > to < 0.
> 
> This patch introduces both GUC and reloption.  In reloptions we
> typically use -1 for "disable reloption, use GUC value instead"
> semantics.  So it's unclear how should we allow reloption to both
> disable feature and disable reloption.  I think we don't have a
> precedent in the codebase yet.  We could allow -2 (disable reloption)
> and -1 (disable feature) for reloption.  Opinions?

Here is patch v11, where the reloption has the same upper limit 1e10
as the GUC.  There is no good reason to have them different.

I am reluctant to introduce new semantics like a reloption value of -2
to disable a feature in this patch right before feature freeze.

I believe there are enough options to disable insert-only vacuuming for
an individual table:

- Set the threshold to 2147483647.  True, that will not work for very
  large tables, but I think that there are few tables that insert that
  many rows before they hit autovacuum_freeze_max_age anyway.

- Set the scale factor to some astronomical value.

Yours,
Laurenz Albe