Shrink space of datafile

Question

Oracle 11.1.0.7:
I need to give some space back to the storage. I am trying to figure out what process should I follow to shrink the datafiles and drop the datafiles.
1. Should I first do shrink space on tablespace and then drop few datafiles?
2. How do I know how much space can be shrunk?
3. How do I know if after shrinking which datafiles are ready for drop?
Is there any other better way?

Answer 1

I"ve just gone through a very similar procedure.
First i had to identify the high water mark in each datafile, and then shrink it back to that or just above that value.
NOTE: It's best to do this when your DB is not in use, specifically growing or if it is some of the commands might fail saying that there is data beyond that size.
Now most likely your data is spread across the datafiles in some sort of "optimal" way by oracle, so it's unlikely that you'll be able to get any "clean" datafiles that you can just drop.
So what i did was to create a new tablespace big enough to hold the objects, MOVED the tables and REBUILT the indexes in the new TS.
Now one option here is to just keep the new TS's and drop the old ones, if needed you can even rename your tablespace (10g and above)
ALTER TABLESPACE old_ts RENAME TO new_ts;
However, you must follow the rules when renaming a tablespace (also need to separetly rename the df if you want to keep your naming standard) :
You cannot rename the SYSTEM or SYSAUX tablespaces.
You cannot rename an offline tablespace.
You cannot rename a tablespace that contains offline datafiles.
Here we wanted to keep the file names locations so i ended up copying everything back into the originals, but before i did that , i could querried dba_segments to get the minimum size required, and I then I shrunk and dropped the datafiles in the original TS as appropriate.
To get the hight water mark, i used this script :
SELECT 'ALTER DATABASE DATAFILE ''' || FILE_NAME || ''' RESIZE ' || HWMM || ';' CL_DYN_SQL
FROM (
SELECT DF.FILE# FILE_ID, DF.NAME FILE_NAME, NVL(DFD.MAX_BLOCK_COUNT,1)*DF.BLOCK_SIZE HWM,
DECODE(SIGN((1024)-(CEIL(NVL(DFD.MAX_BLOCK_COUNT,1)*DF.BLOCK_SIZE/(1024*1024)))),
          1,1024, CEIL(NVL(DFD.MAX_BLOCK_COUNT,1)*DF.BLOCK_SIZE/(1024*1024))) || 'M' HWMM,
          DF.BYTES
FROM V$DATAFILE DF, V$TABLESPACE TS,
SELECT FILE_ID, MAX(BLOCK_ID+BLOCKS-1) MAX_BLOCK_COUNT
          FROM DBA_EXTENTS DE
          WHERE DE.TABLESPACE_NAME IN ('TS_D', 'TS_I')
          GROUP BY FILE_ID
) DFD          
WHERE DF.FILE# = DFD.FILE_ID (+)
AND DF.TS# = TS.TS#
AND TS.NAME IN ('TS_D', 'TS_I')
WHERE HWM < BYTES
ORDER BY 1;