"Marc G. Fournier" <scrappy@hub.org> writes:
>> Hmm. That suggests a clobbered index (as in, first page zeroed out).
> I've rebuilt the whole data structure, as I fear it is a residual of the
> crash we had last week, but, if it means anytyhing, we had done a 'drop
> database/create database' before we tried to rebuild teh database ... is
> it possible that the pg_log got corrupted and could affect this? or maybe
> template1, so that when the create database was run, it was pulling in
> corrupted system tables?
pg_log has no indexes, so it couldn't provoke this. But dropped bits in
an index in template1 could, since createdb just does 'cp -r' to
replicate template1 into a new database.
If you suspect template1 is hosed, you could try doing initdb in a
scratch directory and then do a diff of the resulting template1 against
your working copy.
regards, tom lane