Niall Litchfield schreef:
> Frank van Bortel wrote:
>> Chuck schreef:
>>> Oracle 9.2.0.7
>>>
>>> Is there any way determine authoritatively if an ORA-1555 "snapshot too
>>> old" error was caused by delayed block cleanout? I suspect it is
>>> happening on overnight processing but am not 100% sure. The query starts
>>> at 2am and failed at 3:30. Undo retention is 7 hours so that's not the
>>> problem, and a query of v$undostat for that time period shows only about
>>> 1g of undo being written during that time. The undo tablespace is 4g.
>>>
>>> TIA
>> There was a recent discussion where an ora-01555 occurred during
>> an export - similar undo retention and undostats. One comment
>> was that in 9i, undo retention is not guaranteed.
>
> It was my comment, and yes that's correct
>
>> Another option is an other session, that uses commits in a loop,
>> during an update.
>
> or indeed the overnight batch process itself committing frequently in a
> loop.
>
>
>
As Chuck mentions "the query starts at 2 am", I ruled out the
option of updates, but maybe "query' is a more generalized
term for Oracle processes of any type where he lives; English
is not my native tongue, so every now and again, I misunderstand
or misinterpret things.
--
Regards,
Frank van Bortel
Top-posting is one way to shut me up...
Received on Sun Mar 04 2007 - 03:41:24 CST
