Move data to separate partition

There are 2 methods.  Both involve downtime.

1) For each database, "Detach" the database, move the log file to the new partition, then "Re-attach" the database.  When you do, you'll need to specify the new location of the log file.

2) Update the location of the database files and set the log file location to the location where the log file will moved to.  You can do this for all of your databases.  Then shutdown your MS SQL server process, move all of the log files, and bring it back up again.  If you did it correctly, all your databases will be back up and running.


I usually recommend method 1.  Your DBs are only down one at a time and you know right away if you have a problem with a specific database.  You *must* use method 2 for system databases (msdb, model, distribution, and tempdb).  To move the master files requires yet another different method.

I'll post some screenshots and SQL code in a second.

And here is the TSQL code to the detach/attach (I prefer to use this instead of the GUI, since I script out all my moves beforehand):

-- Detach the old database
USE [master]
GO
-- If needed: this will kick off any users
-- ALTER DATABASE [nemtest1] SET SINGLE_USER WITH ROLLBACK IMMEDIATE
GO
EXEC master.dbo.sp_detach_db @dbname = N'nemtest1'
GO


-- ----------------
-- Move file....
-- ----------------

-- Re-attach the database at the new location
USE [master]
GO
CREATE DATABASE [nemtest1] ON
( FILENAME = N'C:\Project\MSSQL2012\Data\nemtest.mdf' ),
( FILENAME = N'C:\New\Location\of\log\file\nemtest_log.ldf' )
FOR ATTACH
GO

Select allOpen in new window

Here's the SQL for method 2.  Looks much simpler, but remember, you have to shut down your SQL server, move the files, then bring it back up.  This could be a significant amount of downtime.  Also, you have to look up the names of your log files (not necessarily databasename_log, as is the case even in my example).

1) Run this for all the database log files you want to move:

ALTER DATABASE [nemtest1]
MODIFY FILE ( NAME = nemtest_log
        , FILENAME = 'C:\New\Location\of\log\file\nemtest_log.ldf'
);
GO

Select allOpen in new window

2) Shut DOWN your MS SQL server process
3) Move all your log files (may take a long time)
4) Bring up your MS SQL server
5) Hope you did everything correctly.

To find/check your DATA and LOG file locations, I use this SQL:

USE master
GO
SELECT db.database_id AS id
        , db.name AS 'dbname'
        , type_desc
        , mf.name AS 'filename'
        , physical_name
        , mf.state_desc AS 'fstate'
        , mf.is_read_only AS 'fro'
        , db.state_desc AS 'dbstate'
        , db.is_read_only AS 'dbro'
FROM sys.master_files AS mf
        JOIN sys.databases AS db
                ON mf.database_id = db.database_id
ORDER BY db.name, type_desc DESC
;
GO

Select allOpen in new window

You don't like the DETACH method?  Works perfectly for me.  Can you explain *why* you don't like it?

Sure:

1) detach changes the security settings on the file -- often *only* the *specific* person that detached it can reattach it.  If that person has something else come up and can't finish the attach, it can be a royal pita to get the file reattached (w/o doing serious fiddling with file securities, which is not my specialty)

2) if you detach a damaged db, it will almost never reattach, and you've a major problem ... so now you've got to run a DBCC CHECK before the detach, and that can take a long time on a large db



Can you explain *why* you suggest stopping SQL just to move a log file for a selected db(s) to another drive?

I know #1 happens (I consider it a bug) and again, didn't think that would be relevant here - assuming the same user would be doing the entire transfer process.

Curious though...  What if I'm connecting using an SQL Server account and not an AD account?  What if my AD account has access to the SQL Server, but has no permissions to the underlying Windows Server?


#2 very true, but I thought it irrelevant in this context.  If you've got a bad database, you've got other problems. ;-)


As for stopping SQL, you are 100% correct.  There is no need to do it.  I didn't realize taking a database individually offline would do the same thing.  I had assumed, incorrectly, that the database file locations were only read at SQL Server startup (or during an attach/detach).  Of course, you would want to stop your database if you're moving any of the system databases (model, msdb, tempdb, and/or distribution - and of course, master, which has other caveats as well).  When I've had to do migrations in the past, this was the case (ie system files had to be moved as well).

Quite true: you can't take system dbs offline, so the method wouldn't apply there.

#1 is not a bug -- it's (officially) a "security enhancement".  I've even seen some DBAs that agree with this.  But it can be a royal pain.  You can detach using SSMS on your local machine, remote into the main server box, and ACK, you can't attach the db *you* detached because of security issues ... GRRRR, frustrating!

#2 detach seems to have a problem with ANY error, including on nonclus indexes.  Offline / Online seems more resilient to such issues.

Awesome. Thank you much for the info!