Re: database corruption help

I'm running ubuntu and can see references to only one log file which i'm
assuming is the postmaster log..

Anyway the only relevant bits are:

GMT LOG:  relation "pg_class" TID 15538/4: dead HOT-updated tuple ---
cannot shrink relation
2009-02-12 21:06:40 GMT STATEMENT:  VACUUM FULL VERBOSE ANALYZE pg_class
2009-02-12 21:06:40 GMT WARNING:  index "pg_class_relname_nsp_index"
contains 1818 row versions, but table contains 1813 row versions

If that helps...

Thanks
> John Lister <john.lister-ps@kickstone.com> writes:
>
>> I seem to have more dead rows now..
>> doing a vacuum full on pg_class gives me
>>
>
>
>> INFO:  vacuuming "pg_catalog.pg_class"INFO:  "pg_class": found 37
>> removable, 1845 nonremovable row versions in 18905 pages
>> DETAIL:  27 dead row versions cannot be removed yet.
>> Nonremovable row versions range from 160 to 229 bytes long.
>> There were 933834 unused item pointers.
>> Total free space (including removable row versions) is 150368692 bytes.
>> 18839 pages are or will become empty, including 0 at the end of the table.
>> 0 pages containing 0 free bytes are potential move destinations.
>>
>
> Hmm, that last seems to indicate that do_shrinking has been reset.
> It looks like the cases where that happens are reported with elog(LOG)
> which means they probably only went to the postmaster log (maybe we
> should fix things so that vacuum verbose reports those).  What have
> you got in the postmaster log?
>
>             regards, tom lane
>