Reliable communication is crucial to the correct functioning of distributed systems. We propose a multi-ring communication structure and a reconfiguration algorithm that tolerate multiple link failures before the network divides into more than one partition. In case of partitioning, each partition is reconfigured to allow communication among the sites within the partition. The algorithm handles recovery of links and merges partitions once links become operational again. The algorithm itself is fault-tolerant, and it is fully distributed and does not require global knowledge about the status of the network at any one site.