Обсуждение: Easy way to know the original table row/column given a pg_toast.pg_toast_XXXX chunk_id ?

Hello,
just a general question regarding toast, we have run across various instances of problems
having to do with corrupted toast data as a consequence of hardware failures.
The ERRORs were of the type : "missing chunk number %d for toast value %u in %s"
The usual way people try to find the row in the original table is to try a (potentially large)
number of queries in a effort to narrow down the problem, and finally find the row
with the corrupted TOAST.
(yes i know writing a program to do it would be handy, unfortunately the said server of ours
lies somewhere in some ocean with limited ISDN connectivity, so in general deployments are a big PITA)

I understand the chunk_id is the "oid" of the "value" to be toasted, is there any way to get to the row/column
of the toasted table, faster than the aforementioned exhaustive search?

What i did in order to research this (not in a context of a specific problem)
was in my test server, to create a new row in some toasted table, and then find the newest additions (chunk_id) to the
corresponding toast table. But i could not find a way (via pg_* relations) to somehow link the new row of the original
table
to the set of new rows (chunk_id, chunk_seq) in the toast table.

--
Achilleas Mantzios
Head of IT DEV
IT DEPT
Dynacom Tankers Mgmt

On Fri, Jan 24, 2014 at 7:03 AM, Achilleas Mantzios
<achill@matrix.gatewaynet.com> wrote:
[...]
> I understand the chunk_id is the "oid" of the "value" to be toasted, is
> there any way to get to the row/column
> of the toasted table, faster than the aforementioned exhaustive search?

I guess there are no way to do it except the search. Try to look at
this article [1], it perfectly describes what one needs to do in such
cases as yours, and contains all the automation stuff also. Hope you
will find it useful.

[1] http://www.databasesoup.com/2013/10/de-corrupting-toast-tables.html

--
Kind regards,
Sergey Konoplev
PostgreSQL Consultant and DBA

http://www.linkedin.com/in/grayhemp
+1 (415) 867-9984, +7 (901) 903-0499, +7 (988) 888-1979
gray.ru@gmail.com

On 25/01/2014 01:01, Sergey Konoplev wrote:
> On Fri, Jan 24, 2014 at 7:03 AM, Achilleas Mantzios
> <achill@matrix.gatewaynet.com> wrote:
> [...]
>> I understand the chunk_id is the "oid" of the "value" to be toasted, is
>> there any way to get to the row/column
>> of the toasted table, faster than the aforementioned exhaustive search?
> I guess there are no way to do it except the search. Try to look at
> this article [1], it perfectly describes what one needs to do in such
> cases as yours, and contains all the automation stuff also. Hope you
> will find it useful.
>
> [1] http://www.databasesoup.com/2013/10/de-corrupting-toast-tables.html
>

Very useful, Thank you!

--
Achilleas Mantzios
Head of IT DEV
IT DEPT
Dynacom Tankers Mgmt