Postgres Pro
Downloads
Home   >   mailing lists

Thread: ERROR: missing chunk number 0 for toast value 1821556134 in pg_toast_17881

ERROR: missing chunk number 0 for toast value 1821556134 in pg_toast_17881

From
Tory M Blue
Date:
Greetings and salutations.

I've got some weirdness. 

Current:
Postgres 9.3.4 
Slony 2.2.3
CentOS 6.5

Prior running Postgres 9.1.2 w/slony 2.1.3 CentOS 6.2

I found that if I tried to run a vacuum full on 1 table that I recently reindexed (out of possibly 8 tables) that I get this error:

# vacuum full table.ads;

ERROR:  missing chunk number 0 for toast value 1821556134 in pg_toast_17881

If I run a vacuum analyze it completes fine, but I can't run a vacuum full without it throwing an error. I seem to be able to query the table and I seem to be able to add data to the table and slony seems fine as does postgres. 

I'm unclear why the vacuum full is failing with this error. I've done some searching and there are hints to prior bugs, but I didn't catch anything in 9.3.3 to 9.3.7 that talks about this.

My next steps without your fine assistance, will be to drop the table from slon and re-add it (meaning it will drop the table completely from this db and recreate it from the master (there we can do a vacuum full without failure)..

I have already tried to remove the indexes and just create those, but no luck. 

Ideas?

Thanks

Tory



Re: ERROR: missing chunk number 0 for toast value 1821556134 in pg_toast_17881

From
Josh Berkus
Date:
On 05/27/2015 12:50 AM, Tory M Blue wrote:
> Greetings and salutations.
>
> I've got some weirdness.
>
> Current:
> Postgres 9.3.4
> Slony 2.2.3
> CentOS 6.5
>
> Prior running Postgres 9.1.2 w/slony 2.1.3 CentOS 6.2
>
> I found that if I tried to run a vacuum full on 1 table that I recently
> reindexed (out of possibly 8 tables) that I get this error:
>
> # vacuum full table.ads;
>
> ERROR:  missing chunk number 0 for toast value 1821556134 in pg_toast_17881
>
> If I run a vacuum analyze it completes fine, but I can't run a vacuum
> full without it throwing an error. I seem to be able to query the table
> and I seem to be able to add data to the table and slony seems fine as
> does postgres.

Is this happening on the Slony master or the replica?

There have been serveral bugs since 9.1.2 which could have caused this
particular error.  Are you certain that this is a recent problem?

Note that this error affects just one compressed value or row, so you're
not losing other data, unless it's a symptom of an ongoing problem.

--
Josh Berkus
PostgreSQL Experts Inc.
http://pgexperts.com