Services or capabilities described in Amazon Web Services documentation might vary by Region. To see the differences applicable to the China Regions, see Getting Started with Amazon Web Services in China (PDF).

Backup and Restore

SQL Server Usage

The term backup refers to both the process of copying data and to the resulting set of data created by the processes that copy data for safekeeping and disaster recovery. Backup processes copy SQL Server data and transaction logs to media such as tapes, network shares, cloud storage, or local files. You can copy these backups back to the database using a restore process.

SQL Server uses files, or filegroups, to create backups for an individual database or subset of a database. Table backups aren’t supported.

When a database uses the FULL recovery model, transaction logs also need to be backed up. Transaction logs allow backing up only database changes since the last full backup and provide a mechanism for point-in-time restore operations.

Recovery model is a database-level setting that controls transaction log management. The three available recovery models are SIMPLE, FULL, and BULK LOGGED. For more information, see Recovery Models (SQL Server) in the SQL Server documentation.

The SQL Server RESTORE process copies data and log pages from a previously created backup back to the database. It then triggers a recovery process that rolls forward all committed transactions not yet flushed to the data pages when the backup took place. It also rolls back all uncommitted transactions written to the data files.

SQL Server supports the following types of backups:

SQL Server also supports media families and media sets that you can use to mirror and stripe backup devices. For more information, see Media Sets, Media Families, and Backup Sets (SQL Server) in the SQL Server documentation.

SQL Server 2008 Enterprise edition and later versions, support backup compression. Backup compression provides the benefit of a smaller backup file footprint, less I/O consumption, and less network traffic at the expense of increased CPU utilization for running the compression algorithm. For more information, see Backup Compression (SQL Server) in the SQL Server documentation.

A database backed up in the SIMPLE recovery mode can only be restored from a full or differential backup. For FULL and BULK LOGGED recovery models, you can restore transaction log backups to minimize potential data loss.

Restoring a database involves maintaining a correct sequence of individual backup restores. For example, a typical restore operation may include the following steps:

For large databases, a full restore, or a complete database restore, from a full database backup isn’t always a practical solution. SQL Server supports data file restore that restores and recovers a set of files and a single Data Page Restore, except for databases using the SIMPLE recovery model.

Syntax

SQL Server uses the following backup syntax.

Backing Up a Whole Database
BACKUP DATABASE <Database Name> [ <Files / Filegroups> ] [ READ_WRITE_FILEGROUPS ]
  TO <Backup Devices>
  [ <MIRROR TO Clause> ]
  [ WITH [DIFFERENTIAL ]
  [ <Option List> ][;]

BACKUP LOG <Database Name>
  TO <Backup Devices>
  [ <MIRROR TO clause> ]
  [ WITH <Option List> ][;]

<Option List> =
COPY_ONLY | {COMPRESSION | NO_COMPRESSION } | DESCRIPTION = <Description>
| NAME = <Backup Set Name> | CREDENTIAL | ENCRYPTION | FILE_SNAPSHOT | { EXPIREDATE =
<Expiration Date> | RETAINDAYS = <Retention> }
{ NOINIT | INIT } | { NOSKIP | SKIP } | { NOFORMAT | FORMAT } |
{ NO_CHECKSUM | CHECKSUM } | { STOP_ON_ERROR | CONTINUE_AFTER_ERROR }
{ NORECOVERY | STANDBY = <Undo File for Log Shipping> } | NO_TRUNCATE
ENCRYPTION ( ALGORITHM = <Algorithm> | SERVER CERTIFICATE = <Certificate> | SERVER
ASYMMETRIC KEY = <Key> );

SQL Server uses the following restore syntax.

RESTORE DATABASE <Database Name> [ <Files / Filegroups> ] | PAGE = <Page ID>
FROM <Backup Devices>
[ WITH [ RECOVERY | NORECOVERY | STANDBY = <Undo File for Log Shipping> } ]
[, <Option List>]
[;]

RESTORE LOG <Database Name> [ <Files / Filegroups> ] | PAGE = <Page ID>
[ FROM <Backup Devices>
[ WITH [ RECOVERY | NORECOVERY | STANDBY = <Undo File for Log Shipping> } ]
[, <Option List>]
[;]

<Option List> =
MOVE <File to Location>
| REPLACE | RESTART | RESTRICTED_USER | CREDENTIAL
| FILE = <File Number> | PASSWORD = <Password>
| { CHECKSUM | NO_CHECKSUM } | { STOP_ON_ERROR | CONTINUE_AFTER_ERROR }
| KEEP_REPLICATION | KEEP_CDC
| { STOPAT = <Stop Time>
| STOPATMARK = <Log Sequence Number>
| STOPBEFOREMARK = <Log Sequence Number>

Examples

Perform a full compressed database backup.

BACKUP DATABASE MyDatabase TO DISK='C:\Backups\MyDatabase\FullBackup.bak'
WITH COMPRESSION;

Perform a log backup.

BACKUP DATABASE MyDatabase TO DISK='C:\Backups\MyDatabase\LogBackup.bak'
WITH COMPRESSION;

Perform a partial differential backup.

BACKUP DATABASE MyDatabase
  FILEGROUP = 'FileGroup1',
  FILEGROUP = 'FileGroup2'
  TO DISK='C:\Backups\MyDatabase\DB1.bak'
  WITH DIFFERENTIAL;

Restore a database to a point in time.

RESTORE DATABASE MyDatabase
  FROM DISK='C:\Backups\MyDatabase\FullBackup.bak'
  WITH NORECOVERY;

RESTORE LOG AdventureWorks2012
  FROM DISK='C:\Backups\MyDatabase\LogBackup.bak'
  WITH NORECOVERY, STOPAT = '20180401 10:35:00';

RESTORE DATABASE AdventureWorks2012 WITH RECOVERY;

For more information, see Backup Overview (SQL Server) and Restore and Recovery Overview (SQL Server) in the SQL Server documentation.

PostgreSQL Usage

Amazon Aurora PostgreSQL-Compatible Edition (Aurora PostgreSQL) continuously backs up all cluster volumes and retains restore data for the duration of the backup retention period. The backups are incremental and you can use them to restore the cluster to any point in time within the backup retention period. You can specify a backup retention period from one to 35 days when creating or modifying a database cluster. Backups incur no performance impact and don’t cause service interruptions.

Additionally, you can manually trigger data snapshots in a cluster volume that you can save beyond the retention period. You can use Snapshots to create new database clusters.

Restoring Data

You can recover databases from Amazon Aurora automatically retained data or from a manually saved snapshot. Using the automatically retained data significantly reduces the need to take frequent snapshots and maintain Recovery Point Objective (RPO) policies.

The Amazon Relational Database Service (Amazon RDS) console displays the available time frame for restoring database instances in the Latest Restorable Time and Earliest Restorable Time fields. The Latest Restorable Time is typically within the last five minutes. The Earliest Restorable Time is the end of the backup retention period.

Database Cloning

Database cloning is a fast and cost-effective way to create copies of a database. You can create multiple clones from a single DB cluster. You can also create additional clones from existing clones. When first created, a cloned database requires only minimal additional storage space.

Database cloning uses a copy-on-write protocol. Data is copied only when it changes either on the source or cloned database.

Data cloning is useful for avoiding impacts on production databases. For example:

Copying and sharing snapshots

You can copy and share database snapshots within the same Amazon Region, across Amazon Regions, and across Amazon accounts. Snapshot sharing allows an authorized Amazon account to access and copy snapshots. Authorized users can restore a snapshot from its current location without first copying it.

Copying an automated snapshot to another Amazon account requires two steps:

Backup Storage

In all Amazon RDS regions, backup storage is the collection of both automated and manual snapshots for all database instances and clusters. The size of this storage is the sum of all individual instance snapshots.

When an Aurora PostgreSQL database instance is deleted, all automated backups of that database instance are also deleted. However, Amazon RDS provides the option to create a final snapshot before deleting a database instance. This final snapshot is retained as a manual snapshot. Manual snapshots aren’t automatically deleted.

The Backup Retention Period

Retention periods for Aurora PostgreSQL DB cluster backups are configured when creating a cluster. If not explicitly set, the default retention is one day when using the Amazon RDS API or the Amazon CLI. The retention period is seven days if using the Amazon Console. You can modify the backup retention period at any time with values of one to 35 days.

Disabling automated backups

You can’t disable automated backups on Aurora PostgreSQL. The backup retention period for Aurora PostgreSQL is managed by the database cluster.

Migration Considerations

Migrating from a self-managed backup policy to a Platform as a Service (PaaS) environment such as Aurora PostgreSQL is a complete paradigm shift. You no longer need to worry about transaction logs, file groups, disks running out of space, and purging old backups.

Amazon RDS provides guaranteed continuous backup with point-in-time restore up to 35 days.

Managing a SQL Server backup policy with similar RTO and RPO is a challenging task. With Aurora PostgreSQL, all you need to set is the retention period and take some manual snapshots for special use cases.

Examples

The following walkthrough describes how to change Aurora PostgreSQL DB cluster retention settings from one day to seven days using the Amazon RDS console.

For more information, see Maintenance Plans.

Summary

For more information, see Managing an Amazon Aurora DB cluster in the User Guide for Aurora.