Date: Mon, 10 Sep 2018 11:05:24 -0700 (MST)
From: chrisjonmcdonald <chris@woollysweater.com>
To: pgsql-admin@postgresql.org
Message-ID: <1536602724722-0.post@n3.nabble.com>
Subject: invalid toast page in block 1500 of relation
 pg_tblspc/16387/PG_10_201707211/16392/80912520.
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Precedence: bulk

Hi,

I am running postgres 10.3 server on armv71 linux which i've built from
source. The server host is a raspberry pi with postgres server running
inside a docker container. All filestore is over iscsi. I use postgres to
store data for mediawiki. This is a very low throughput environment. I've
had no problems with this configuration over recent years and upgraded to
postgresql to 10.3 about a year ago with no problems since.

About a week ago i found i could not take a backup using pg_dump and the
failure message was:

chunk 1 for toast value 80945156 in pg_toast_80849380
invalid page in block 1500 of relation
pg_tblspc/16387/PG_10_201707211/16392/80912520.

I have a filesystem backup of the failing instance.
I've tried a reindex which fails with the same error
I've tried using zero_damaged_pages which allows the reindex to succeed but
then reports the same error still when i attempt to select from the table.

What can i do to minimise my data-loss and fixup my instance (if anything)?
I have a db backup but its a month old and i'd like to try and recover what
data i can from that month.

TIA

chris


--
Sent from: http://www.postgresql-archive.org/PostgreSQL-admin-f2076596.html