Using FILESTREAM with Other SQL Server Features
Because FILESTREAM data is in the file system, this topic provides some considerations, guidelines, and limitations for using FILESTREAM with the following features in SQL Server:
SQL Server Integration Services (SSIS) handles FILESTREAM data in the data flow like any other BLOB data by using the DT_IMAGE SSIS data type.
You can use the Import Column transformation to load files from the file system into a FILESTREAM column. You can also use the Export Column transformation to extract files from a FILESTREAM column to another location in the file system.
You can work with FILESTREAM data through distributed queries and linked servers by treating it as varbinary(max) data. You cannot use the FILESTREAM PathName() function in distributed queries that use a four-part name, even when the name refers to the local server. However you can use PathName() in the inner query of a pass-through query that uses OPENQUERY().
SQL Server does not support database snapshots for FILESTREAM filegroups. If a FILESTREAM filegroup is included in a CREATE DATABASE ON clause, the statement will fail and an error will be raised.
When you are using FILESTREAM, you can create database snapshots of standard (non-FILESTREAM) filegroups. The FILESTREAM filegroups are marked as offline for those database snapshots.
A SELECT statement that is executed on a FILESTREAM table in a database snapshot must not include a FILESTREAM column; otherwise, the following error message will be returned:
Could not continue scan with NOLOCK due to data movement.
A varbinary(max) column that has the FILESTREAM attribute enabled at the Publisher can be replicated to a Subscriber with or without the FILESTREAM attribute. To specify the way in which the column is replicated, use the Article Properties - <Article> dialog box or the @schema_option parameter of sp_addarticle or sp_addmergearticle. For more information about how to set schema options, see How to: Specify Schema Options (SQL Server Management Studio) and How to: Specify Schema Options (Replication Transact-SQL Programming). Data that is replicated to a varbinary(max) column that does not have the FILESTREAM attribute must not exceed the 2-GB limit for that data type; otherwise, a run-time error is generated. We recommend that you replicate the FILESTREAM attribute, unless you are replicating data to SQL Server 2005. Replicating tables that have FILESTREAM columns to SQL Server 2000 Subscribers is not supported, regardless of the schema option that is specified. For more information about how to replicate data to earlier versions of SQL Server, see Using Multiple Versions of SQL Server in a Replication Topology.
Replicating large data values from SQL Server 2008 to SQL Server 2005 Subscribers is limited to a maximum of 256 MB data values. For more information, see Maximum Capacity Specifications.
Considerations for Transactional Replication
If you use FILESTREAM columns in tables that are published for transactional replication, note the following considerations:
If any tables include columns that have the FILESTREAM attribute, you cannot use values of database snapshot or database snapshot character for the @sync_method property of sp_addpublication.
The max text repl size option specifies the maximum amount of data that can be inserted into a column that is published for replication. This option can be used to control the size of FILESTREAM data that is replicated. For more information, see max text repl size Option.
If you specify the schema option to replicate the FILESTREAM attribute, but you filter out the uniqueidentifier column that FILESTREAM requires or you specify not to replicate the UNIQUE constraint for the column, replication does not replicate the FILESTREAM attribute. The column is replicated only as a varbinary(max) column.
Considerations for Merge Replication
If you use FILESTREAM columns in tables that are published for merge replication, note the following considerations:
Both merge replication and FILESTREAM require a column of data type uniqueidentifier to identify each row in a table. Merge replication automatically adds a column if the table does not have one. Merge replication requires that the column have the ROWGUIDCOL property set and a default of NEWID() or NEWSEQUENTIALID(). In addition to these requirements, FILESTREAM requires that a UNIQUE constraint be defined for the column. These requirements have the following consequences:
If you add a FILESTREAM column to a table that is already published for merge replication, make sure that the uniqueidentifier column has a UNIQUE constraint. If it does not have a UNIQUE constraint, add a named constraint to the table in the publication database. By default, merge replication will publish this schema change, and it will be applied to each subscription database. For more information about schema changes, see Making Schema Changes on Publication Databases.
If you add a UNIQUE constraint manually as described and you want to remove merge replication, you must first remove the UNIQUE constraint; otherwise, replication removal will fail.
By default, merge replication uses NEWSEQUENTIALID() because it can provide better performance than NEWID(). If you add a uniqueidentifier column to a table that will be published for merge replication, specify NEWSEQUENTIALID() as the default.
Merge replication includes an optimization for replicating large object types. This optimization is controlled by the @stream_blob_columns parameter of sp_addmergearticle. If you set the schema option to replicate the FILESTREAM attribute, the @stream_blob_columns parameter value is set to true. This optimization can be overridden by using sp_changemergearticle. This stored procedure enables you to set @stream_blob_columns to false. If you add a FILESTREAM column to a table that is already published for merge replication, we recommend that you set the option to true by using sp_changemergearticle.
Enabling the schema option for FILESTREAM after an article is created can cause replication to fail if the data in a FILESTREAM column exceeds 2 GB and there is a conflict during replication. If you expect this situation to arise, it is recommended that you drop and re-create the table article with the appropriate FILESTREAM schema option enabled at creation time.
Merge replication can synchronize FILESTREAM data over an HTTPS connection by using Web Synchronization. This data cannot exceed the 50 MB limit for Web Synchronization; otherwise, a run-time error is generated.
Log shipping supports FILESTREAM. Both the primary and secondary servers must be running SQL Server 2008, or a later version, and have FILESTREAM enabled.
Full-text indexing works with a FILESTREAM column in the same way that it does with a varbinary(max) column. The FILESTREAM table must have a column that contains the file name extension for each FILESTREAM BLOB. For more information, see Querying varbinary(max) and xml Columns (Full-Text Search), Full-Text Search Filters,Full-Text Indexing and Querying Process, and sys.fulltext_document_types (Transact-SQL).
The full-text engine indexes the contents of the FILESTREAM BLOBs. Indexing files such as images might not be useful. When a FILESTREAM BLOB is updated it is reindexed.
For failover clustering, FILESTREAM filegroups must be put on a shared disk. FILESTREAM must be enabled on each node in the cluster that will host the FILESTREAM instance. For more information, see How to: Set Up FILESTREAM on a Failover Cluster.