FREE SPACE Inconsistency

Question

Hi,
I am using the following SQL to report on free space from a production database and it has reported free space incorrectly two days in a row even though nothing was deleted
SELECT NVL(b.tablespace_name, NVL(a.tablespace_name,'UNKNOWN')) "Tablespace",
ROUND(Mbytes_alloc,0) "Allocated MB",
ROUND(Mbytes_alloc-NVL(Mbytes_free,0),0) "Used MB",
ROUND(NVL(Mbytes_free,0),0) "Free MB",
ROUND( ((Mbytes_alloc-NVL(Mbytes_free,0))/Mbytes_alloc * 100 ),2) "% Used",
ROUND(Mbytes_maxalloc,0) "Max Allocated MB"
FROM
(SELECT SUM(bytes)/1024/1024 Mbytes_free,
MAX(bytes) /1024/1024 largest,
tablespace_name
FROM sys.dba_free_space
GROUP BY tablespace_name
) a,
(SELECT SUM(bytes)/1024/1024 Mbytes_alloc,
SUM(maxbytes) /1024/1024 Mbytes_maxalloc,
tablespace_name,
COUNT(*) data_files
FROM sys.dba_data_files
GROUP BY tablespace_name
)b
WHERE a.tablespace_name (+) = b.tablespace_name
ORDER BY 5 DESC;
On the Thursday, running the script reported the USERS tablespace as:
Thursday:
allocated mb: 44266
used mb:     43977
free mb: 289
% Used: 99.35%
maxsize mb: 69632
On Friday, the script reported:
allocated mb: 44266
used mb: 43412
free mb: 854
% Used: 98.07%
maxsize mb: 69632
Why is it reporting a difference in FREE MB over two days when nothing has been deleted?
Any help is appreciated.
Thanks.

Answer 1

You can prove no objects have been deleted by 1 - implementing auditing on tables and indexes so that all create, drops, and alters will be logged and available via the dba_audit_trail and associated views or 2 - You can take a snapshot of the objects assigned to the tablespace, store it in a table, and then the next day compare the contents of dba_segmetns verse the snapshot displaying any dropped or new objects. You could also capture size changes allowing you to show where free changes came from and also where free extents were consumed.
I think DDL auditing is the best approach and that most all DDL should be audited on production databases. You should not be creating and dropping tables as part of normal production processing. I am not a big fan of DML auditing. However I have seen developers who write code that creates and drops tables. Generally I get the drop and create replaced with a call to a truncate procedure.
You customer may not trust the audit results so you may want to set up auditing and perform option 2 as proof. Once you have proved your point then all you should need going forward is auditing; however, you may find running a option 2 process that tracks object extent changes useful as a space management tool especially if you do not have the EM Diagnostic and Performance Pack licenses so you cannot access the AWR (10g+).
HTH -- Mark D Powell --