Under normal operating conditions, application writes are logged to the Primary Replicator Log and synchronously replicated to the Bunker node and any other synchronous Secondary sites. By default, the replication to the Bunker node is in the synchronous override mode. Thus, in the case of proper network availability the replication happens in synchronous mode. However, if the network becomes unavailable, replication to the Bunker Secondary happens asynchronously. During normal replication, the Bunker node functions as a Secondary. However, if a disaster occurs at the Primary, the Bunker node must be converted to a Primary and the data in its Replicator Log can be used to bring the Secondary up-to-date.

A write is completed to the application as soon as it is logged to the Primary Replicator Log, the Bunker Replicator Log, and the other synchronous Secondary Replicator Logs. Volume Replicator asynchronously writes the data to the Primary data volume and sends it to the asynchronous Secondary site. When the Secondary acknowledges the writes, the Replicator Log header is updated to indicate the status of the Secondary.

In a typical asynchronous replication setup, the network bandwidth is provisioned for average application write rate. Therefore, in the case of high write rates, the Bunker Replicator Log may contain some writes that are considered complete by the application but are still to be applied to the asynchronous Secondary. The network bandwidth for synchronous replication must therefore be provisioned for peak application write rate. The Replicator Log protection (srlprot) for the RLINK between the Primary and Bunker is set to off, by default. If for some reason the Primary replicator overflows for this RLINK, then the RLINK is detached.