Data Backup and Recovery

How to utilize SnapShots and SnapMirror with Oracle Database?

LSOTECHUSER
12,512 Views

We will be migrating our database (Oracle 11g RAC) from an older Equallogic storage system to a NetApp FAS 2240 in the coming months.  Currently the database is using ASM and is connected via iSCSI.  When we move to the NetApp, at least initially we will use iSCSI and continue to use ASM. We would like to take advantage of Snapshots and Snapmirror when the database is moved to the NetApp.  We will not be using SnapManager for Oracle at this time.

I am not quite sure how the NetApp SnapShot process integrates with the Oracle and what pieces are required to make this work.

How are Snapshots and Snapmirror handled so that the database is backed up properly?  In particularly, how is the backup scheduled so that the database is put into hot or online backup mode when the snapshot and snapmirror runs?  Is everything handled from the Oracle host side via a script that can make the database ready and then trigger the snapshot on NetApp or do you need to coordinate the snapshot on NetApp separately with Oracle?

thanks

1 ACCEPTED SOLUTION

billshaffer
12,511 Views

There are many snapmanager products - the basic ones are snapmanager for windows/unix.  All these do is intall some commands through which you can manage the snapshots on the controller - for example, snap list, snap create, snap delete, etc.  The connection is over one of the API channels.  The other snapmanager products (for Oracle, Exchange, SharePoint, etc.)  utilize the basic snapmanager products for talking to the controllers, but they also talk to the application in question, and will handle all the scheduleing and integration.

Remember that, in a snapmirror relationship, you're copying the source volume and snapshots.  What you say about updating a snapmirror relationship creating a snap that is copied to the destination is technically correct, but that process is part of the internal workings, and I find it better just to focus on the volume and non-snapmirror snapshot data.  So if you put your DB in hot standby, then issue a snapmirror update, wait for it to finish, then take your DB out of hot standby, you've got a consistent copy of your DB at DR that you could snapmirror back if needed.

If, however, you put your DB in hot standby, take a snapshot, take your DB out of hot standby, and THEN inititate the snapmirror update - which you don't need to wait for to finish - then your destination _volume_ is inconsistent from a DB perspective, because the DB was active when it was copied.  The snapshot you took when the DB was in hot standby, though, also gets copied - and that is consistent.  Instead of just reversing the direction of the snapmirror, you'd have to break the snapmirror, restore the volume to the desired snapshot on the destination, then mirror it back.

One caveat would be that a snapmirror reversal (as in my first scenario) would only copy incremental changes, while the second scenario would be a full copy - irrelevant in the event of a real disaster,but possibly impactful when faced with corruption.  Another benefit of doing just the snapshots would be that you could keep multiple consistent copies around as well.

Make sense?

Bill

View solution in original post

15 REPLIES 15
Public