Enhancement request for selecting the Media server to perform the restore from an OST disk pool when multiple Media Servers have access to the disk pool.

Abstract:  Enhancement request for selecting the Media server to perform the restore from an OST disk pool when multiple Media Servers have access to the disk pool.

A Data Domain OST disk pool can be accessed by multiple Media servers that can send backup data the DD storage server.  The access to DD is controlled by credentials.   Any Media server that has credentials to the DD can send/retrieve backup images from the DD disk pool.   By design NetBackup will select any Media Server with access to the Data Domain OST disk pool where the backup image resides to perform the restore for a NetBackup client.  This selection is based on a load balancing algorithm which selects the least busy NetBackup server.  This is explained in NetBackup 6.5 System Administrator's Guide I.

Issue: More than one Media Servers may have access to the same OpenStorage disk pool. Under normal circumstances, NetBackup selects Media Server for the restore based on Media Server Load Balancing algorithm. This may result in a non-desired Media Server to be selected for the restore causing the restore to fail.

Point Solutions:  As of Netbackup 6.5.4 and newer versions Symantec has provided point solutions and options to control the selection of  the Media server used for the restore.  In summary these options include:

Option 1: Touch file USE_BACKUP_MEDIA_SERVER_FOR_RESTORE under /usr/openv/netbackup/db/config on the Master.   When this touch file is present NetBackup will use the same Media Server that performed the backup to handle the restore.

Option 2:  Media host override (also known as FORCE_RESTORE_MEDIA_SERVER configuration) can be used to force a specific media server to perform restores of backups done by another media server. 

Option 3: Use command line bprestore with the qualifier –disk_media_server – this works on a per job level and allows the user to specify the media server required for the specific restore job.  Unlike the other two options this is a dynamic setting that can be applied when needed.

Option 1 has precedence over options 2 and 3 and option 2 has precedence over option 3.

The point solutions outlined above although can alleviate most restore issues for small or medium environments, for the size of the Bank, with multiple NBU domains spanning sites (primary and EV/Dr) these options do not address every restore situation.

For example, if option 1 is implemented then a SAN Media Server cannot restore  the backups of a another SAN Media Server.  This is specially the case with VLDB that are deployed as Media Server.  It will require the Netbackup admin to remove the touch file and either use option 2 or instruct the user (if possible) to use option 3.  Using one option for fixing restore for some requests  it is certain to break the restores for some other requests.  In such large multi-site NBU environment these options are operationally cumbersome if not impossible to maintain.

Enhancement Request:

Below are some of the items my orginization is looking for improvements into the NetBackup product. 

1. Provide flexibility for end-users to select the SAN Media Server for the restore without having to contact the NetBackup admin to modify touch files or update media override entries
2. Ability for end-users to identify which Media Server performed the backup for a client/user backup.
3. Ability to specify the Media server to perform the restore from the GUIs (Java and Windows)
4. Ability to specify the Media Server to perform the restore for all NetBackup agents (currently limited to Oracle, DB2)
5. Ability to create group of restore servers for groups of clients

Thanks.