Hi Again,

Is there any difference in the way vacuum is handled in postgres9.6 and 
postgres12.9, We are noticing the below issue of waiting process only after 
upgrading to postgres12.5

$ ps -ef | grep 'waiting'
postgres  8833 62646  0 Jul28 ?        00:00:00 postgres: postgres cgms [local] 
VACUUM waiting
postgres 18437 62646  0 Jul27 ?        00:00:00 postgres: postgres cgms [local] 
VACUUM waiting

Regards,
Karthik

From: Julien Rouhaud <rjuju...@gmail.com>
Date: Monday, 7 November 2022 at 7:06 PM
To: Laurenz Albe <laurenz.a...@cybertec.at>
Cc: Karthik Jagadish (kjagadis) <kjaga...@cisco.com>, Dave Page 
<dp...@pgadmin.org>, pgsql-hack...@postgresql.org 
<pgsql-hack...@postgresql.org>, Chandruganth Ayyavoo Selvam (chaayyav) 
<chaay...@cisco.com>, Prasanna Satyanarayanan (prassaty) <prass...@cisco.com>, 
Jaganbabu M (jmunusam) <jmunu...@cisco.com>, Joel Mariadasan (jomariad) 
<jomar...@cisco.com>
Subject: Re: Postgres auto vacuum - Disable
Hi,

On Mon, Nov 07, 2022 at 02:22:56PM +0100, Laurenz Albe wrote:
> On Mon, 2022-11-07 at 12:12 +0000, Karthik Jagadish (kjagadis) wrote:
> > I have follow-up question where the vacuum process is waiting and not doing 
> > it’s job.
> > When we grep on waiting process we see below output. Whenever we see this 
> > we notice
> > that the vacuum is not happening and the system is running out of space.
> >
> > [root@zpah0031 ~]# ps -ef | grep 'waiting'
> > postgres  8833 62646  0 Jul28 ?        00:00:00 postgres: postgres cgms 
> > [local] VACUUM waiting
> > postgres 18437 62646  0 Jul27 ?        00:00:00 postgres: postgres cgms 
> > [local] VACUUM waiting
> >
> >
> > What could be the reason as to why the vacuum is not happening? Is it 
> > because some lock is
> > present in the table/db or any other reason?
>
> Look in "pg_stat_activity".  I didn't check, but I'm sure it's the 
> intentional break
> configured with "autovacuum_vacuum_cost_delay".  Reduce that parameter for 
> more
> autovacuum speed.

Really?  An autovacuum should be displayed as "autovacuum worker", this looks
like plain backends to me, where an interactive VACUUM has been issued and is
waiting on a heavyweight lock.

Reply via email to