Underreplicated

Underreplicated is a term used in distributed storage systems, notably the Hadoop Distributed File System (HDFS), to describe a data block that has fewer replicas than the configured replication factor. Under such conditions, data availability and durability can be compromised because a fraction of the expected copies of the block are missing.

Underreplication can occur when DataNodes fail or are taken offline, during maintenance, or when network partitions prevent replication from completing. Other causes include disk failures, decommissioning of nodes, corruption or failed block reports, and delays in heartbeat or block report communication that cause existing replicas to be discarded or not newly created.

In HDFS, the NameNode monitors the replication counts of every block and marks a block as underreplicated when the number of live replicas falls below the required replication factor. The system maintains an under-replicated blocks queue and uses the cluster’s replication policy to schedule additional replicas on healthy DataNodes. By default, the replication factor is typically 3.

Remediation involves automatically re-creating missing replicas where possible. If enough healthy DataNodes are available, the system will re-replicate blocks to meet the target factor. If not, blocks may remain underreplicated, reducing fault tolerance and increasing the risk of data loss in failure scenarios. Administrators can repair clusters by bringing nodes back online, decommissioning or reconfiguring replication factors, or manually triggering replication.

Related concepts include over-replication, replication factor, NameNode, DataNode, and block management in distributed file systems.