SQL Server DO NOT enforce the .mdf/.ndf/.ldf File Name extensions.
For more info, please refer:
http://sqlserverzest.com/2011/11/23/can-a-database-have-more-than-one-primary-data-mdf-file/

whereas NoTruncate rearranges all the pages and brings unallocated pages to the end of the file but DOES NOT release the free space back to the operating system.(***file does not appear to have shrunk, it remains the same size)

BEAR IN MIND

TRUNCATEONLY:- NO REARRANGE just Release empty space at the end of the file

NOTRUNCATE:- ONLY REARRANGE but does not release space to the OS. the size of the file remains the same..

so do a NOTRUNCATE to get the free pages to the end of the file and then do a TRUNCATEONLY to release the space back to the OS(i mean..SHRINK THE FILE PHYSICALLY)

So to my understanding, if I do a DBCC SHRINKFILE without a NOTRUNCATE or a TRUNCATEONLY it will both rearrange the pages and release the space at the end of the file? Is that correct?

Thanks again. Your input really helped us out.

Coy

The command you are using to shrinkfile only returns free space at the end of the file to the operating system, so it is possible that the database does not have any free space at the end of the file.

instead the data file may have free space(unallocated pages) at the beginning or in the middle of the file, so in order to get this free space to the end of the file we have to do a

DBCC SHRINKFILE(*****,*****, NOTRUNCATE)

We have to use the NOTRUNCATE option to get the free space to the end of the data file.

NOTRUNCATE rearranges all the objects in the database and pushes the unallocated pages to the end of the file

After you do a NOTRUNCATE you then run the following Command to release the free space at the end of the file to the OS

DBCC SHRINKFILE(****,****,TRUNCATEONLY)

***Please note SHRINKFILE(***,***,NOTRUNCATE) is not advisable as it rearranges all the pages in a database (actually shriking is not advisable!!)

***NOTRUNCATE option is also only applicable to Data file not Log files.

NOTE:-I would advise you to rebuild all your indices(index’s) after you perform this.

bear in mind shrink NOTRUNCATE would take loads of time as it rearranges all the pages in a database ….

We are trying to shrink our database from 2.4TB to 1.5TB. Most of the data is on one data file. We ran the operation below and we specify using TRUNCATEONLY so that the shrink operation doesn’t get logged.

The operation completed but the OS or SQL Server has not given back the space.

Any feedback on why the SHRINK operation would not fully complete?

We’ve investigated many angles, but still haven’t been able to find the bottleneck.

Thanks,

Coy Bernardo

USE [UMG_GIMM_EDW]
GO
DBCC SHRINKFILE (N’UMG_GIMM_EDW_FG1_Dat1′ , 2331220, TRUNCATEONLY)
GO

When using partitions in a circular rotation, it is desirable to shrink emptied partitions to a minimum size until they are next in the rotation. I my test example I have 12 partitions (1 per month). I only wish to keep 3-4 months of data online, deleting 1 day at a time and shrinking the partition when it is empty (where is TRUNCATE table (partition_id) when you need it? MIA and no where in sight – what else should one expect from MS). This approach reduces the physical disk space requirements by ca. 2/3.

My issue now is that in my long running test scenario, it is now in fact impossible to shrink some of the empty partititions, and I am left with wasted space that I cannot reclaim. For some reason the ability to reclaim space has broken in this test database.

I am running 2008R2 (10.50.1797)

If you have any clues why an empty partition might not shrink other than SQLServer being broken, I would be happy to hear them.

The .ndf extension is suggested by microsoft as a easy way to identify your secondary data files from primary data files (.mdf), you can use any extension you want in those files, always trying to be consistent.

In my humble opinion i recommend you to use the “standard” extensions (.mdf, .ndf, .ldf) for your DB files in case of any other DBA get his hands on them, that will be easier to undestand.

I am experimenting with one of our (critical and huge) databases that initially was created with two physical files, .mdf and ldf. But now I wanted to split the things up a bit, and I added several logical/physical files, both in the same filegroup, and in additional filegroup. When adding those files, I chose their extensions to be .mdf, and there are several .mdf files and one .ldf file now – no .ndf’s.

I would also like to post a question about the possibilities for partitioning a database (i.e. spread it across several physical files) – I experimented with partitioning some tables, but that seems to be too complicated…

Thanks in advance!

Regards, Zoran