Re: pg_dump or hardware?

Thanks for the quick reply

Tom Lane wrote:
> andrea suisani <andrea.suisani@opinioni.net> writes:

[cut]

> Hmm ... what do you get from
>
> select oid from pg_class where relname = 'nominativi';
   oid
--------
  561644
(1 row)

afaics it seems weird.... does this mean that another postgresql
object screw up? this is the "\d nominativi" output:

          Table "public.nominativi"
    Column   |       Type       | Modifiers
------------+------------------+-----------
  nome_cogno | character(70)    |
  indirizzo  | character(40)    |
  cap        | character(5)     |
  citta      | character(35)    |
  prov       | character(2)     |
  prefisso   | character(4)     |
  telefono1  | character(13)    |
  telefono2  | character(13)    |
  note       | character(100)   |
  idpersona  | integer          |
  estrazione | date             |
  num_estra  | integer          |
  occupato   | boolean          |
  cod_prov   | integer          |
  cod_com    | integer          |
  cod_reg    | integer          |
  capoluo    | integer          |
  rand       | double precision |
Indexes:
     "citta1_idx" btree (prov, citta) WHERE (((prov = 'BL'::bpchar) OR (prov =
'PD'::bpchar) OR (prov = 'RO'::bpchar) OR (prov = 'VE'::bpchar) OR (prov = 'TV'::bpchar)
OR (prov = 'VR'::bpchar)) AND (cod_com IS NULL))
     "cod_com_ndx1" btree (cod_com)
     "codprov_capo_rand1" btree (cod_prov, capoluo, rand)
     "pre_tel_index1" btree (prefisso, telefono1)


and none of nominativi's indexes has an oid like 201327173

select oid
from pg_class
where relname in ('citta1_idx','cod_com_ndx1','codprov_capo_rand1','pre_tel_index1');
    oid
----------
  46788374
  40916657
  40916656
  40916658
(4 rows)


> select relname from pg_class where oid = 201327173;
  relname
---------
(0 rows)

while if I "ask" for oid  561644 this is what I get:

select relname from pg_class where oid = 561644;
   relname
------------
  nominativi
(1 row)


> If the first returns 201327173 while the second doesn't return anything,
> then I'd wonder about corruption of pg_class's OID index.  REINDEXing it
> might help.
> I don't remember whether 7.4 makes you use a standalone
> backend to reindex system catalogs --- see its REINDEX man page for
> details.

from what I can see from REINDEX man page, I can reindex shared system catalogs
in stand-alone mode

[cut]

Regards,


Andrea