Atlassian Corporation
System and method for consistent mirroring in a cluster

Abstract:

Described herein is a computer implemented method for consistent distributed mirroring of data maintained by a primary cluster. The method comprises removing, by one mirror node of multiple mirror nodes, a change notification from a change queue shared between multiple mirror nodes; initiating an update process for the data corresponding to the change notification; and fetching one or more objects corresponding to the change notification from the primary cluster. Designated temporary slave mirror nodes notify a designated temporary master mirror node that fetching objects is complete, and the temporary master mirror node notifies the temporary slave mirror nodes that the update process is complete once a notification is received from each of the temporary slave mirror nodes.