March 17, 2003 at 8:25 am
Hello,
I currently have a database in which all of the objects of every application reside in the Primary filegroup. For the sake of recovery operations, I am considering dividing the database into filegroups that correspond to each application that I have.
My concern is that under the current set up, if I have to ever recover the database to an earlier point in time because of a problem with one application (I use full recovery), users of another particular application will experience an unnecessary loss of data.
Right now, I have four main applications in the database that I would categorize as Central Data, Multiple Measures, CELDT, and Music Inventory. If I understand Filegroups correctly, one way to approach this would be to create four Filegroups for each of the four applications, then setup one or more NDF files for each of the filegroups. I might have a filegroup called ‘Central_Data’, and it might contain ‘Central_Data1.NDF’ and Central_Data2.NDF, and these files might reside on separate disks or be part of a RAID 5 arrangement. The other applications would be set up in a similar manner.
Is it correct that when an object is created for an application on an NDF file that resides in a particular filegroup, that the object is assigned to the filegroup and not a particular NDF file within the filegroup? I remember reading that NDF files within filegroups are filled proportionally.
Is it correct that if I were to split up my database into different filegroups this way and needed to recover objects/data specific to an application, I could recover by naming a specific filegroup, and the associated transaction log file for the database? What effect of a recovery like this have on the other filegroups?
If I needed to recover the whole database, and had it split up into filegroups, would I need to name each filegroup in the recovery steps?
I have read several sections from BOL on the topic of filegroups and their recovery, but have not experimented with it yet. Your input on these questions would be very helpful.
Regards,
CSDunn
March 17, 2003 at 8:33 am
I don't think what you are aiming for can be done via filegroups.
If your applications have unrelated database objects thyen put them in separate databases then they can be backed up and restored independantly.
With filegroups the database is still a single entity it just aids means that you can backup the files at different times.
Cursors never.
DTS - only when needed and never to control.
Cursors never.
DTS - only when needed and never to control.
Viewing 2 posts - 1 through 1 (of 1 total)
You must be logged in to reply to this topic. Login to reply