THIS TOPIC APPLIES TO: SQL Server (starting with 2016)Azure SQL DatabaseAzure SQL Data Warehouse Parallel Data Warehouse
This topic describes backup-and-restore considerations for several features in SQL Server 2017. These features include: file restore and database startup, online restore and disabled indexes, database mirroring, and piecemeal restore and full-text indexes.
In this Topic:
This section is relevant only for SQL Server databases that have multiple filegroups.
When a database is started, only filegroups whose files were online when the database was closed are recovered and brought online.
If a problem is encountered during database startup, recovery fails, and the database is marked as SUSPECT. If the problem can be isolated to a file or files, the database administrator can take the files offline and try to restart the database. To take a file offline, you can use the following ALTER DATABASE statement:
ALTER DATABASE database_name MODIFY FILE (NAME ='filename', OFFLINE)
If startup succeeds, any filegroup that contains an offline file remains offline.
This section is relevant only for databases that have multiple filegroups and, for the simple recovery model, at least one read-only filegroup.
In these cases, when a database is online, the index can be created, dropped, enabled or disabled only if all filegroups holding any part of the index are online.
For information about restoring offline filegroups, see Online Restore (SQL Server).
This section is relevant only for full-model databases that have multiple filegroups.
The database mirroring feature will be removed in a future version of Microsoft SQL Server. Avoid using this feature in new development work, and plan to modify applications that currently use this feature. Use Always On availability groups instead.
Database mirroring is a solution for increasing database availability. Mirroring is implemented on a per-database basis and works only with databases that use the full recovery model. For more information, see Database Mirroring (SQL Server).
To distribute copies of a subset of the filegroups in a database, use replication: replicate only those objects in the filegroups you want to copy to other servers. For more information about replication, see SQL Server Replication.
Creating the Mirror Database
The mirror database is created by restoring, WITH NORECOVERY, backups of the principal database on the mirror server. The restore must keep the same database name. For more information, see Prepare a Mirror Database for Mirroring (SQL Server).
You can create the mirror database by using use a piecemeal restore sequence, where supported. However, you cannot start mirroring until you have restored all the filegroups and, typically, restored log backups to get the mirror database close enough in time with the principal database. For more information, see Piecemeal Restores (SQL Server).
Restrictions on Backup and Restore During Mirroring
While a database mirroring session is active, the following restrictions apply:
Backup and restore of the mirror database are not allowed.
Backup of the principal database is allowed, but BACKUP LOG WITH NORECOVERY is not allowed.
Restoring the principal database is not allowed.
This section is relevant only for databases that contain multiple filegroups and, for the simple-model databases, only for read-only filegroups.
Full-text indexes are stored in database filegroups and can be affected by a piecemeal restore. If the full-text index resides in the same filegroup as any of the associated table data, piecemeal restore works as expected.
To view the filegroup ID of the filegroup that contains a full-text index, select the data_space_id column of sys.fulltext_indexes.
Full-Text Indexes and Tables in Separate Filegroups
If a full-text index resides in a separate filegroup from all of the associated table data, the behavior of piecemeal restore depends on which of the filegroups is restored and brought online first:
If the filegroup that contains the full-text index is restored and brought online before the filegroups that contain the associated table data, full-text search works as expected as soon as the full-text index is online.
If the filegroup that contains the table data is restored and brought online before the filegroup that contains the full-text index, full-text behavior might be affected. This is because Transact-SQL statements that trigger a population, rebuild the catalog, or reorganize the catalog fail until the index is brought online. These statements include CREATE FULLTEXT INDEX, ALTER FULLTEXT INDEX, DROP FULLTEXT INDEX, and ALTER FULLTEXT CATALOG.
In this case, the following factors are significant:
If the full-text index has change tracking, user DML will fail until the index filegroup is brought online. Delete operation will also fail until the index filegroup is online.
Regardless of change tracking, full-text queries fail because the index is not available. If a full-text query is tried when the filegroup that contains the full-text index is offline, an error is returned.
Status functions (such as FULLTEXTCATALOGPROPERTY) succeed only when they do not have to access full-text index. For example, access to any online full-text metadata would succeed, but uniquekeycount, itemcount would fail.
After the full-text index filegroup is restored and brought online, the index data and table data are consistent.
As soon as both the base table filegroup and the full-text index filegroup are online, any paused full-text population is resumed.
SQL Server supports NTFS file system data compression of read-only filegroups and read-only databases.
Restoring files in a read-only filegroup is supported on compressed NTFS files. Backup and restore of these filegroups works essentially as it would for any read-only filegroup, with the following exceptions:
Restoring a read-write file (including the primary or log files of a read-write database) to a compressed volume fails and displays an error.
Restoring a read-only database to a compressed volume is allowed.
Log files of read/write databases should never be placed on compressed file systems.