Map Update

Network disruption, disk failure, or a crash on a storage node causes IO to fail on the affected node (See Failure recovery). Once the failure is resolved, the node must resynchronize before it can service IOs again. First, the recovering node fetches an updated dirty map from a NORMAL storage server — this mechanism is called map update. After the map update completes, IOs can be serviced again. RMR-server syncs the dirty data both while servicing IOs and through the sync thread.