Re: BUG #12812: invalid attribute number <number> for <table_name>

Thanks John,
I also try to re index for pg_attribute and that particular table, but that
is also failed. We will check the reason for Postgres itself crash. I hope
upcoming version may address this problem.
On 28-Feb-2015 9:28 pm, "Tom Lane" <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:

> John R Pierce <pierce(at)hogranch(dot)com> writes:
> > On 2/28/2015 1:50 AM, IPN Bala GSS TVL wrote:
> >> ERROR: invalid attribute number 24932 for transaction_set
> >> LINE 1: SELECT * FROM AD_GSS_DMS_P.TRANSACTION_SET
>
> > that looks like data corruption may have occurred.
>
> Yeah :-(. The only occurrence of that message text in 9.3 is in
> RelationBuildTupleDesc, and it's saying it found a row in pg_attribute
> with that attnum and the mentioned table's attrelid. The attnum is
> obviously bogus for any row in pg_attribute, ergo something has corrupted
> data in pg_attribute. It's unlikely that only one row has been corrupted,
> too :-(
>
> Depending on how valuable that database is, you might want to go back
> to your last backups, or you might want to seek professional help
> --- there are assorted consulting companies that specialize in dealing
> with data-corruption situations.
>
> In any case you should try to identify the cause of the problem.
> It's more than likely a hardware or OS problem --- there are few
> known reasons for Postgres itself to cause this sort of failure.
>
> regards, tom lane
>