Thread: Proposal- GUC for max dead rows before autovacuum

Proposal- GUC for max dead rows before autovacuum

From
Michael Lewis
Date:
To mitigate the need for per-table tuning of autovacuum configuration, I'd like to propose a new GUC for autovacuum_vacuum_threshold_max.

Currently, it seems that I can either set autovacuum_vacuum_scale_factor much smaller than default on tables with millions of rows, or set a value globally that means small tables are auto vacuumed rarely.

The default value for this new setting value could be -1 or 0 to disable the feature, or something like 100,000 perhaps so that tables with more than 500,0000 tuples are candidates for an autovacuum before they would with current default values.

Re: Proposal- GUC for max dead rows before autovacuum

From
Laurenz Albe
Date:
On Tue, 2019-11-19 at 15:35 -0700, Michael Lewis wrote:
> To mitigate the need for per-table tuning of autovacuum configuration, I'd like to propose a new GUC for
autovacuum_vacuum_threshold_max.
> 
> Currently, it seems that I can either set autovacuum_vacuum_scale_factor much smaller than default on tables with
millionsof rows,
 
> or set a value globally that means small tables are auto vacuumed rarely.
> 
> The default value for this new setting value could be -1 or 0 to disable the feature, or something like 100,000
perhaps
> so that tables with more than 500,0000 tuples are candidates for an autovacuum before they would with current default
values.

I think this is unnecessary.
Usually you have problems only with a few tables, and it is no problem
to set autovacuum parameters on these individually.

Yours,
Laurenz Albe