Minimizing timeout failures on large database restores

A large SQL Server restore may fail with a Client Read Timeout error before any data has been read from the NetBackup media. This error occurs because the SQL Server may need to pre-write the database files before the restore operation begins. The time that is required for this process is a function of certain factors: the size of the database files and the speed at which your host machine can write to disk. For example, consider that your system can perform disk writes at the rate of 60 megabytes per second and you have a 2.4 terabyte database. Then it takes at least 12 hours for SQL Server to prep the disk before the actual restore can begin. In reality, the delay may be even longer than what you calculate by as much as 20% to 40%.

The timeout problem can be resolved by increasing the NetBackup Client Read Timeout setting. Use the NetBackup Administration Console on the server to change the properties of each client that contains a database you may need to restore. The default for the Client Read Timeout setting is 300 seconds (5 minutes). If you have any clients which contain large SQL Server databases, you may need to set this value much higher.

If you use SQL Server 2005 or later, you can eliminate file initialization during SQL Server restores.

More Information

Instant data file initialization