Deploying and Versioning Applications
The following information is included as reference for existing applications that support offline scenarios only. For developing new applications, please see Architecture and Classes for Database Synchronization.
This topic describes deployment and versioning of client and server applications. For information about initializing peer databases, see "Initializing a Server Database" in How to: Synchronize Other ADO.NET Compatable Databases.
The way in which you deploy a client and server application that uses Sync Framework depends on whether the application is two-tier or N-tier. In a two-tier application, all the synchronization code is on the client, and the application can be deployed like any other application that is based on SQL Server Compact. Typically, deployment involves either using ClickOnce to publish the application or packaging the application and then copying it to each client. For more information about how to deploy applications that use SQL Server Compact, see the topics "Deploying Applications" and "How to: Build and Deploy an Application to a Desktop" in SQL Server Compact Books Online. In an N-tier application, you must also deploy code to the computer that hosts the server-side synchronization components.
For two-tier and N-tier applications, you must decide how to create and initialize the client database that has the schema and data that the application requires. There are two main approaches:
Use Sync Framework to create the database, and download the table schema and data during the first synchronization. This is the simplest approach. However, it might require a lot of network traffic and server processing if you have many clients and a large set of data.
Pre-create the database that has the table schema and an initial data set. The database, which is in the form of an .sdf file, can be transported or downloaded by the client application. If the application downloads incremental changes, it can bring the client database up-to-date during the first synchronization. If you use this approach, set an appropriate value for the CreationOption property.
If the application includes uploads and you want to know where changes originated, the application can set the ClientId property to uniquely identify the client database to the server; or Sync Framework will generate a new ID for the database.
Pre-creating the client database is not suitable for applications in which data is filtered differently for each client. It can cause missed changes during the initial synchronization.
For more information about how to initialize the client database, see How to: Initialize the Client Database and Work with Table Schema.
After an application has been developed and deployed, new versions of the application might be developed. There are two areas to consider for application versioning:
The parts of the application that are affected by a new version
Whether a new version must coexist with earlier versions
Generally, an N-tier application is easier to update for Sync Framework because you can update the server-side components separately from the client-side components. This enables you to expose a new service that includes updated functionality, and shields clients from server-side changes until they are ready to have client-side components updated.
Versioning for N-tier Applications
For Sync Framework, the main application update to consider is a schema change to the server database. Sync Framework does not automatically propagate schema changes from the server to the client. However, it does support schema changes as long as you update the appropriate parts of the application. For a list of schema changes that are supported in the client database, see Working with Table Schemas in the Client Database. If a schema change is made in the server database, you should first decide whether it must be propagated to the client at all. For example, if you add a credit card column to a table in the server database but you have a business rule that disallows employees from taking credit card information offline, you do not have to propagate the schema change to the client. If you decide to propagate the schema change, how you do this depends on the type of change.
Common Schema Changes
The two most common schema changes are adding a column to an existing table and adding a new table, or deciding to include an existing table that was not originally part of the application. If the application requires a new table to be synchronized, you must add a SyncAdapter for the table. After the server components are updated, the client application or the user of the application can determine whether to synchronize the new table. If the decision is to synchronize the new table, the client components must be changed to include a SyncTable for the new table. These changes must be pushed out to each client application in some way. ClickOnce technology enables you to publish application updates, and applications can check for any updates before they start. If you add a new column to a table and want that column to be synchronized, you must add the column to the client database and update the commands that are used to synchronize changes. For schema changes other than adding a new column, drop the table on the client and re-create it by using Sync Framework or another mechanism.
In some scenarios, the new and old versions of a client application might have to coexist. In such cases, you can take two different approaches:
Maintain multiple versions of the server components so that an older version of the client components can continue to access the appropriate version of the server components. The best way to approach versioning is to expose a new service with updated functionality. Clients could be upgraded by using ClickOnce technology. When the upgrade is complete, the client would be redirected to the new web service.
Design the client and server components to use an application version parameter. This approach enables the client application to notify the server about which version it expects. Consider the following example in which the stored procedure that selects inserts from the server accepts an @appversion parameter. A parameter is added to the SelectIncrementalInsertsCommand, and then a value is supplied by using a SyncParameter in the SyncAgent. When synchronization occurs, the application version is sent to the server, and the stored procedure selects the correct version of the table schema.
this.Configuration.SyncParameters.Add( new SyncParameter("@appversion", 1)); customerIncrUpdates.Parameters.Add("@appversion", SqlDbType.Int);
Me.Configuration.SyncParameters.Add( _ New SyncParameter("@appversion", 1)) customerIncrUpdates.Parameters.Add("@appversion", SqlDbType.Int);
CREATE PROCEDURE usp_CustomerSelectIncrementalInserts( @sync_last_received_anchor timestamp, @sync_new_received_anchor timestamp, @appversion int ) AS IF @appversion = 1 BEGIN SELECT CustomerId, CustomerName, SalesPerson FROM Sales.Customer WHERE InsertTimestamp > @sync_last_received_anchor AND InsertTimestamp <= @sync_new_received_anchor END ELSE IF @appversion = 2 BEGIN SELECT CustomerId, CustomerName, SalesPerson, CustomerType FROM Sales.Customer WHERE InsertTimestamp > @sync_last_received_anchor AND InsertTimestamp <= @sync_new_received_anchor END ELSE BEGIN PRINT('Unrecognized version') RETURN END