Understanding NetApp MetroCluster: Architecture, Data Synchronization, and High‑Availability Solutions

NetApp MetroCluster (MCC) is a dual‑active storage solution built on Data ONTAP that connects two data‑center sites using dedicated FC/VI adapters, FibreBridge, and either direct fiber switches (under 500 m) or DWDM links (up to 100 km) to provide high‑availability.

The architecture evolved to place two FAS/V dual‑controller arrays at each site, forming cross‑site clusters that maximize resource utilization while allowing remote access to controllers in case of failures.

Version 8.3 introduced a 4‑controller active‑active MetroCluster supporting distances up to 200 km. Two HA pairs form local clusters, which are then combined into a four‑node cluster. NVRAM stores mirrored logs, and SyncMirror replicates writes to both sites, ensuring continuous service during node or site failures.

Key components include:

Clustered Failover – administrator‑initiated failover between primary and disaster‑recovery storage.

SyncMirror – real‑time remote data copy that allows access to data from the remote site during a failover.

ClusterRemote – management mechanism that detects disasters and initiates remote storage takeover.

Data synchronization occurs over three distinct networks:

Cluster configuration sync – a redundant TCP/IP network using CRS (Configuration Replication Service) to replicate configuration changes between clusters.

NVRAM log sync – redundant FV‑VI/FC‑VI links with RDMA and QoS to synchronize NVRAM logs and heartbeat messages, reducing write I/O overhead.

Backend disk double‑write – FC network with FibreBridge converting SAS to FC, connecting dedicated switches at both sites for disk‑level replication.

The four‑node MetroCluster operates in an AP mode where only one node of an HA pair serves data at a time; failover to the partner node occurs upon failure, and site‑wide failover can be triggered manually via CFOD commands or automatically by the TieBreak arbitration software.

Memory data mirroring splits each node’s NVRAM into four sections for local logs, HA‑pair partner logs, and remote HA‑pair logs, ensuring rapid recovery after controller or site failures.

Metadata changes are stored in MDV (Metadata Volume) and replicated to RDB (Replicated Database) via CRS. After link restoration, MDV logs are replayed to synchronize configuration differences.

SyncMirror operates at the Aggregate level, where each Aggregate consists of two Plexes (local and remote). Writes are committed to both Plexes before acknowledgment, and Plex failures are recovered using Aggregate snapshots.

The TieBreak software runs on Linux hosts, monitoring SSH sessions to HA pairs and clusters, detecting failures within 3‑5 seconds and retrying every 3 seconds.

Overall, MetroCluster integrates NetApp Cluster, SyncMirror, RAID‑DP, and other protection features to deliver SAN and NAS dual‑active storage with near‑zero RTO/RPO, supporting enterprise applications such as Oracle RAC, VCS/MSFC, and SAP HANA.