Re: pg_ctl promote causes error "could not read block" (PG 9.5.0 and 9.5.4)

From: Michael Paquier <michael(at)paquier(dot)xyz>
To: raj 1988 <raj888389(at)gmail(dot)com>
Cc: pgsql-general(at)lists(dot)postgresql(dot)org
Subject: Re: pg_ctl promote causes error "could not read block" (PG 9.5.0 and 9.5.4)
Date: 2018-03-29 00:31:13
Message-ID: 20180329003113.GE2102@paquier.xyz
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-general

On Wed, Mar 28, 2018 at 09:36:11AM -0700, raj 1988 wrote:
> Are we hitting some bug? tried to look around but not able to confirm if we
> are hitting a bug or not. For us this is happening consistently on
> different servers whenever we do pg_ctl promote and then it block WRITE on
> that table.

This has the strong smell of the FSM bug fixed in 9.5.5:
https://www.postgresql.org/docs/devel/static/release-9-5-5.html

So, in order to get things right:
1) Update to the latest version of Postgres 9.5.
2) Make sure that your cluster gets in a clean state. There are
instructions here:
https://wiki.postgresql.org/wiki/Free_Space_Map_Problems

> As of now we get rid of the error either by doing vacuum full or CTAS, but
> i am afraid what we will do in case this happens to our few TB tables.

This rebuilds the file-space map, which is why it goes away. You really
want to do the work I am mentioning above to get back to a clean state.
--
Michael

In response to

Browse pgsql-general by date

  From Date Subject
Next Message Ken Beck 2018-03-29 02:29:14 Re: Postgresql 9.3 Server will not start after Ubuntu Upgrade
Previous Message Adrian Klaver 2018-03-28 22:58:26 Re: Postgresql 9.3 Server will not start after Ubuntu Upgrade