missing chunk number 0 for toast value ...

Hi list,

I am working with postgres 8.0.3 on a HP Pro Liant Xeon box under SuSE
Linux 9.1

When dumping a database, I got the error

pg_dump: SQL command to dump the contents of table "bearbeitungsschritt"
failed: PQendcopy() failed.
pg_dump: Error message from server: ERROR:  missing chunk number 0 for
toast value 32152544
pg_dump: The command was: COPY public.bearbeitungsschritt (objid,
fach_rfg, klasse, anlass_e, anlass_s, anlass_w, datum_e, datum_s,
datum_w, datum_u, datum_l, beziehung, bezeichnung_des_bearbeitungssch,
anfangsdatum, abgabedatum, sachbearbeitung, naechste_sachbearbeitung,
status_der_bearbeitung) TO stdout;

so I could not dump the table.
It seemed that only some few rows of a single column seemed infected ,
and those were unreadable only in a single column, everything else could
be dumped out and reconstructed.
The errors seemed to happen in the column "beziehung" of the table
described below. I should note that the column is of type text[] and the
array can be very long and can have severals of hundreds of entries.


Moreover, in some few cases, when selecting
SELECT oid, some_other_rows from bearbeitugsschritt where oid =
some_special_value,

I did not get back one value but two values (oid should be unique?)
Meanwhile the errors are fixed and a new database has been set up.



              Table "public.bearbeitungsschritt"
              Column              |     Type      | Modifiers
---------------------------------+---------------+-----------
  objid                           | character(16) | not null
  fach_rfg                        | real          | not null
  klasse                          | text          | not null
  anlass_e                        | character(16) | not null
  anlass_s                        | character(16) |
  anlass_w                        | character(16) |
  datum_e                         | integer       | not null
  datum_s                         | integer       |
  datum_w                         | integer       |
  datum_u                         | integer       | not null
  datum_l                         | integer       | not null
  beziehung                       | text[]        |
  bezeichnung_des_bearbeitungssch | text          | not null
  anfangsdatum                    | integer       | not null
  abgabedatum                     | integer       | not null
  sachbearbeitung                 | text          |
  naechste_sachbearbeitung        | text          |
  status_der_bearbeitung          | character(16) |
Triggers:
     "RI_ConstraintTrigger_6084155" AFTER DELETE ON bearbeitungsschritt
FROM bearbe_kosten NOT DEFERRABLE INITIALLY IMMEDIATE FOR EACH ROW
EXECUTE PROCEDURE "RI_FKey_cascade_del"('bearbe_kosten_fk1',
'bearbe_kosten', 'bearbeitungsschritt', 'UNSPECIFIED', 'bearbe_objid',
'objid')
     "RI_ConstraintTrigger_6084156" AFTER UPDATE ON bearbeitungsschritt
FROM bearbe_kosten NOT DEFERRABLE INITIALLY IMMEDIATE FOR EACH ROW
EXECUTE PROCEDURE "RI_FKey_noaction_upd"('bearbe_kosten_fk1',
'bearbe_kosten', 'bearbeitungsschritt', 'UNSPECIFIED', 'bearbe_objid',
'objid')
Inherits: objektkopf

No Im looking for the causes for the crash (as I dont want to do the
ugly repair more often...)
a) hardware? (NEW and strong machine!)
b) kernel?
c) something with arrays (overflows?)
Can anybody help me?
Or is there a simpler way to repair this?

Thank you,

With kind regards

Hubert Fröhlich

-------------------------------------------------------------------------------
Dr.-Ing. Hubert Fröhlich
Bezirksfinanzdirektion München
Alexandrastr. 3, D-80538 München, GERMANY
Tel. :+49 (0)89 / 2190 - 2980
Fax  :+49 (0)89 / 2190 - 2997
hubert dot froehlich at bvv dot bayern dot de