Setting the initial synchronization priorities for inbound connections

Applies To: Windows Server 2003, Windows Server 2003 R2, Windows Server 2003 with SP1, Windows Server 2003 with SP2

Setting the initial synchronization priority for inbound connections

It is often advantageous to control the sequence of synchronization during replication. Using Distributed File System (DFS) snap-in and the File Replication Service (FRS), you can configure synchronization priorities. You can determine which inbound connection will go first in the initial synchronization that occurs when you add a new member to a replica set or when you restore one of your replica set members. For example, you can adjust synchronization priorities to ensure that computers across a high-bandwidth network connection synchronize first, followed by synchronization with computers that are communicating through a lower-bandwidth connection. By adjusting the synchronization priorities, you can often reduce the time needed to perform the synchronization process.

Using FRS, you can choose from three synchronization priorities:

  • High: All connections marked High must successfully synchronize before a synchronization is attempted on Medium-priority connections.

  • Medium: At least one Medium priority connection must successfully complete an initial synchronization before any synchronization is attempted on Low-priority connections. FRS attempts to synchronize all connections in the Medium priority level, but only one must be successful before FRS attempts to synchronize Low priority connections.

  • Low: FRS attempts to synchronize with each computer on a Low priority connection one time, but any failure will not delay other synchronization attempts.

Low is the default synchronization priority for all connections. Typically, you should configure high-bandwidth or local connections to have a higher synchronization priority than low-bandwidth connections. Therefore, depending on the scenario, a priority of either Medium or High is appropriate for your local connections. However, be aware that if you have a High priority connection that fails, no other synchronization will take place.

The following two scenarios describe how you might use synchronization priorities:

The use of High and Low priorities

In the first example, you are using a hub and spoke topology. Computer X at a branch site fails and needs to be restored. Computer X has inbound connections from Computers A and B, which are located at a hub site. Computer X also has an inbound connection from Computer Y, which is located at the same branch site as Computer X. You should set the synchronization priority from Computer Y to Computer X to High. Set synchronization priorities from Computers A and B at the hub site to Computer X at the branch site to Low.

The reason for these settings is that under normal conditions, you would not want to do a full synchronization over the slow link between Computer X and Computer A or Computer B. A synchronization from the hub site would take too long and consume all of the bandwidth of the slow link. Instead, the recovery of Computer X must occur first from the local Computer Y. Computer X can then try to synchronize from either Computer A or Computer B, but if communication problems prevent the synchronization, Computer X can still come online with the same content as Computer Y.

The use of Medium and Low priorities

For the second example, you are using a custom topology that consists of a two-level hub and spoke architecture. Computer A is at the hub site. The second level consists of Computers J, K, L, and M at the regional offices. At the bottom level are several departmental computers in branch offices.

Computer A at the hub site has high bandwidth bi-directional connections to Computers J, K, L and M. At the regional offices, Computers J, K, L and M are connected in a medium bandwidth bi-directional ring. The branch office computers are each connected to two of the regional computers (J, K, L and M) by a low-bandwidth connection.

Computers J, K, L and M are all expected to be up to date. Therefore, for Computer A, you should set the inbound connections from Computers J, K, L and M to Medium priority. You should also set the priority for the ring of connections between the regional office computers to Medium. The priority from the branch office computers to Computers J, K, L and M should be set to Low.

Consider Computer K. Its inbound connections from A, J and L should all be set to Medium priority because a synchronization with any one of them would be sufficient for Computer K to come online. If the hub site (Computer A) is unavailable, Computers J, K, L and M can still complete a recovery without first synchronizing with Computer A. The justification for the Low priority from the branch office computers is the same as described in the first example.

Overriding the schedule for initial synchronization

Typically, synchronization will only take place at times defined in the synchronization schedule that you configure. However, you can override the schedule for initial synchronization or the initial creation of replica set member. By selecting the Override Schedule check box in the Priorities dialog box, you can cause the initial replication to occur immediately.

For more information on setting the priority, see Modify replication topology. For more information on setting the replication schedule, see Set the replication schedule.