Re: Autovacuum not running properly

From: Luca Ferrari <fluca1978(at)gmail(dot)com>
To: Rajesh Kumar <rajeshkumar(dot)dba09(at)gmail(dot)com>
Cc: Pgsql-admin <pgsql-admin(at)lists(dot)postgresql(dot)org>
Subject: Re: Autovacuum not running properly
Date: 2023-09-04 08:25:36
Message-ID: CAKoxK+5hti6wnctLLvrhfq1Bn8wnmsVQDTy+iE=2j+Re15D=iQ@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-admin

On Fri, Sep 1, 2023 at 6:32 PM Rajesh Kumar <rajeshkumar(dot)dba09(at)gmail(dot)com> wrote:
>
> In that case how to be sure of autovacuum is not necessary for this table? On what basis? Do we have any calculation

The idea should be the number of "manipulated tuples" exceed the
thresholds. For example, in the case of insert workload, something
like:

SELECT current_setting( 'autovacuum_vacuum_insert_threshold' )::numeric
+ current_setting( 'autovacuum_vacuum_insert_scale_factor' )::numeric
* reltuples
FROM pg_class
WHERE relname = <your relation> and relkind = 'r';

will give you the amount of tuples that, once inserted, will trigger
an autovacuum. Therefore, unless your table receive more tuples than
the above, autovacuum will not consider vacuuming the table (thn
there's autoanalyze, that is another story but the reasoning is
similar).

Long story short: if your table is pretty much static, than it is
clear that autovacuum is not goint to vacuum over an over the same
data.

Luca

In response to

Responses

Browse pgsql-admin by date

  From Date Subject
Next Message Luca Ferrari 2023-09-04 08:30:08 Re: Why isn't there a IF NOT EXISTS clause on constraint creation?
Previous Message Erik Wienhold 2023-09-02 13:27:57 Re: DBLINK Error