Re: weird pg_statistic problem - Mailing list pgsql-general

From Enrico Sirola
Subject Re: weird pg_statistic problem
Date
Msg-id BA8A1A86-0349-4F3F-9B61-455D899825E1@gmail.com
Whole thread Raw
In response to Re: weird pg_statistic problem  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-general
Hello Tom,

Il giorno 11/nov/2011, alle ore 22.05, Tom Lane ha scritto:

> Enrico Sirola <enrico.sirola@gmail.com> writes:
>> this morning I experienced a weird problem with our pgsql database (9.0.3):
>> while performing a simple query, I receive the following error:
>> Nov 11 10:24:09 <host> postgres[23395]: [7-1] ERROR:  missing chunk number 0 for toast value 550556127 in
pg_toast_2619
>
> Was this a transient error, or repeatable?
>
> If it was transient, it's probably a recently-fixed issue:
> http://archives.postgresql.org/pgsql-hackers/2011-10/msg01366.php
> http://archives.postgresql.org/pgsql-committers/2011-11/msg00014.php

apparently, it's transient

>> ERROR:  duplicate key value violates unique constraint "pg_statistic_relid_att_inh_index"
>> DETAIL:  Key (starelid, staattnum, stainherit)=(531526103, 7, f) already exists.
>
> This seems unrelated.  Can you repeat this one?  If so, try REINDEX'ing
> that index and see if the problem goes away.
>
> It'd be worth your while to update to 9.0.5 --- we fixed a fair number
> of potential data-corruption issues since January.

while examinigs the db state, I found many other inconsistencies here and there (e.g. tables with multiple records with
thesame PK), at the end I restored a backup.  
I'll be installing today 9.0.5 and let you know if it happens again. Thanks a lot for your help,
Enrico

P.S.
by the way, I have a streaming replica server that has been corrupted as well.

pgsql-general by date:

Previous
From: Frank van Vugt
Date:
Subject: = any((select )) needs cast, why?
Next
From: Tom Lane
Date:
Subject: Re: = any((select )) needs cast, why?