Date: Thu, 27 Jun 2019 19:47:35 -0400
From: Alvaro Herrera <alvherre@2ndquadrant.com>
To: Jenn <smith.rents@gmail.com>
Cc: pgsql-admin@lists.postgresql.org
Subject: Re: Need help to debug ERROR: MultiXactId 808530957 has not been
 created yet -- apparent
Message-ID: <20190627234735.GA9854@alvherre.pgsql>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: 
 <CADR+diih7uRdTtL-9r9a1ExZx32yMrVjYL7uhDaVCM-X_O6oyg@mail.gmail.com>
User-Agent: Mutt/1.9.4 (2018-02-28)
Precedence: bulk

On 2019-Jun-28, Jenn wrote:

> Getting "MultiXactId 808530957 has not been created yet -- apparent" error
> when trying to VACUUM FREEZE big table, about 263m rows. Normal VACUUM runs
> fine.
> Also tried to recreate the table and there were no errors "create table
> my_table_copy as
>   select * from my_table".
> 
> 1.Is it data corruption problem?
> 2.how to diagnose the error and possibly fix?

Yeah, it seems like data corruption -- that multixact value is not
anywhere near a reasonable range.  Can you find out what page it is
complaining about, and get a heap_page_items() from that page?

instructions for part 1:
1. in command line: gdb -p <pid>, where PID comes from
   "select pg_backend_pid()" in psql
2. in gdb, do "break errfinish", then "cont".  It should wedge.
3. in psql, execute the query that throws the error
4. gdb becomes alive.  Do "bt full".  A page number is displayed
somewhere in the resulting stack trace.  If you don't know where to
look, paste the output here.

part 2:
1. create extension pageinspect
2. select * from heap_page_items(get_raw_page('yourtable'::regclass, <page number>));
3. paste output here

-- 
Álvaro Herrera                https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services