Each SQL Server® restore scenario is implemented using one or more restore operations, which is called a restore sequence. A restore sequence moves data through one or more phases of restore. The possible phases of a restore include the data copy, redo (roll forward), and undo (roll back) phases, which are detailed in the following topics.
Redo is the process of reapplying logged changes to the data in the roll-forward set to bring the data forward in time. To accomplish the redo, the SQL Server® Database Engine processes Transaction Log backups as they are restored. The engine starts with the log contained in the data copy backups listed in the preceding topic and any Incremental Transaction Log backups that are then restored. Roll forward stops at the recovery point; this point is either the end of the log or a point specified in the SQL Server restore options for Incremental Transaction Log backups. Plug‑in for SQL Server supports both time-based and marked transaction-based PIT recovery.
After the database is transactionally consistent, recovery brings the database online.
The following restore scenarios are supported under Simple Recovery Model:
|
• |
Complete database restore: With a Simple Recovery Model database, the goal of a complete database restore is to restore the entire database. This type of restore typically involves restoring a Full Database backup or a Full Database backup followed by a Differential Database backup. The entire database is offline during a complete database restore. |
|
• |
File restore: In a file restore for a Simple Recovery Model database, the goal is to restore one or more damaged read-only files without restoring the entire database. This type of restore involves a Full File and Filegroup backup. |
|
• |
Partial database restore: A partial restore for a Simple Recovery Model database includes restoring the primary filegroup and all read/write filegroups that were included in a Partial Database or Differential Partial Database backup. Restores of the read-only filegroups with a Full File and Filegroup backup are required only if the read-only filegroups are damaged. |
The following restore scenarios are supported under the Full and Bulk-Logged Recovery Models:
|
• |
Complete database restore: Under the Full and Bulk-Logged Recovery Models, the goal is to restore the entire database to the point-of-failure. A complete restore sequence typically includes: |
|
1 |
Perform Tail-Log backup. |
|
2 |
Restore Full Database or Full File and Filegroup backup while selecting With NO RECOVERY restore option. |
|
3 |
Restore Differential Database or Differential File and Filegroup backup while selecting With NO RECOVERY restore option. |
|
4 |
Restore all subsequent Incremental Transaction Log backups in order while selecting the With NO RECOVERY restore option. |
|
5 |
|
• |
File restore: In a file restore for a Full or Bulk-Logged Recovery Model database, the goal is to restore the entire database while leaving the option available to restore damaged files without restoring the entire database. This type of restore involves the following restore sequence: |
|
1 |
Perform Tail-Log backup. |
|
2 |
|
3 |
|
4 |
Restore all subsequent Incremental Transaction Log backups in order while selecting the With NO RECOVERY restore option. |
|
5 |
|
• |
Point-in-time (PIT) recovery: You can perform a PIT recovery on the transaction logs that are restored during an Incremental Transaction Log restore. Plug‑in for SQL Server supports both time-based — using hours:minutes:seconds — and marked-transaction-based PIT recovery. PIT recovery options are available when restoring Incremental Transaction Log and Tail-Log backups. |
|
• |
Time-Based Point-in-Time Recovery: Time-based PIT recovery is useful when the time that the data corruption occurred is known. For example, if a developer dropped a table at 6:00:00 A.M., PIT recovery can be performed with a stop time of 5:55:00 A.M. Time-based PIT recovery can only be used for unrelated databases; therefore, when multiple SQL Server® databases are required to be logically consistent, marked-transaction PIT recovery should be used. |
|
• |
Marked-Transaction-Based Point-in-Time Recovery: SQL Server 2005 and later support the ability to insert named marks into a transaction log to allow recovery to a specific mark. This functionality is beneficial if you have two or more databases that must be logically consistent. You can implement special procedures to ensure the recoverability of these databases after a restore. Because you can only recover related databases to a marked transaction, your application must be configured to insert named marks into the transaction log before performing backups. For more information, see Ensuring Recoverability of Related Databases and the Recovering to a Marked Transaction in the SQL Server Books Online. |
You can only restore system databases from backups that are created on the version of SQL Server® that the instance is running. For example, to restore a system database on an instance that is running on SQL Server 2005 SP1, use a backup that was created after the instance was upgraded to SQL Server 2005 SP1.
The master database must be accessible and at least partly usable for starting a SQL Server® Instance. If the master database becomes unusable, you can return the database to a usable state by either restoring the master from a current database backup or rebuilding the master.
You can restore the master database from a Full Database backup only if you are able to start the server instance. If severe damage to the master database prevents you from starting the SQL Server, rebuilding the master is the only option left. For more information, see Rebuilding the master Database in the SQL Server Books Online.
The restoration of the master database includes:
|
• |
|
• |
Because the changes made to master since the last backup were lost, re‑create them by performing those steps again. |
|
• |
For more information, see Restoring data to an alternate SQL Server®.
Restoring the model or msdb database from a backup is required under the following circumstances:
|
• |
The master database has been rebuilt. In this case, restore model from a backup because rebuilding master re-creates model. |
|
• |
Restoring model or msdb is the same as performing a Complete Database restore of a user database.
Before restoring msdb, the SQL Server Agent Service in the Control Panel must be stopped. Quest also recommends that you shut down applications, because the restore procedure disconnects users when necessary.
Replication supports restoring replicated databases to the same SQL Server® Instance and database from which the backup was created. If you restore a replicated database from a backup to a different instance, replication settings cannot be preserved. In this case, re-create all publications and subscriptions after backups are restored.
For more information, see Strategies for Backing Up and Restoring Snapshot and Transactional Replication and Strategies for Backing Up and Restoring Merge Replication sections in the SQL Server Books Online.
© ALL RIGHTS RESERVED. Conditions d’utilisation Confidentialité Cookie Preference Center
