Synchronization Reliability

There are several scenarios that can cause a synchronization failure. For example, the network disconnection of a single server or having a single target server fail and be unreachable can cause a partial failure but would not cause the synchronization to quit. A complete failure is only likely to happen if there is a cluster-wide network or TCP/IP failure.

NOTE

---

If the cluster controller fails when synchronization is occurring, the synchronization will fail and the content will not be applied. After you bring the controller back online (or designate a new controller) we recommend that you initiate a full synchronization to ensure that the entire cluster is in synchronization.

Transport Recovery

Application Center uses the existing TCP/IP and DCOM mechanisms for transport recovery if there is a mid-point transfer failure. The only driver responsible for recovery is the File System driver because it implements its own transport.

Apply Time Recovery

Apply time recovery is the responsibility of the individual drivers because it occurs when a driver is writing its data stream. If the target is locked, the driver is responsible for retrying the write. Each content driver determines the retry count and interval for its own data. If the apply can't be done within its retry parameters, the driver skips that set of changes and generates an event notification that's sent to the replication engine.

Each synchronization will have an end state—that is, a degree of success—which generates different events and performs subsequent actions depending on the end state. (See Table 6.1.) A "success" end state occurs when all content is replicated and applied to all the targets correctly. Because changes to content are typically set to automatic synchronization mode, full synchronization will be applied only where necessary.

Table 6.1 Replication Results

Result	Corrective action	Events generated
Success—All changes are replicated and applied to all servers.	None	Successful replication.
Partial Success—One or more, but not all, servers fail to receive or have changes applied.	Resynchronize the failed items to the targets.	Failure events for specific target(s).
Complete Failure—Every server fails to receive or have changes applied.	Repeat the synchronization.	Failure events for every target.

NOTE

---

Because the Application Center Synchronization Service runs in a separate process from IIS, even a synchronization failure that requires a restart won't affect the cluster's ability to serve content.