Persistent data backend architecture

This page introduces the architecture of the backend that Consul server agents use to store Raft index data.

Raft index

Consul uses the Raft protocol to manage server consensus, to maintain a reliable, fault-tolerant state across all servers. This consensus mechanism ensures consistent service discovery and health monitoring, even when individual servers fail or become temporarily disconnected from the cluster.

The Raft index provides a record of the cluster's state. It tracks interactions between Consul servers as they conduct elections, register service instances into the Consul catalog, and update the catalog with the results of service node health checks.

You can also use Consul's snapshot agent to save a copy of the entire Raft index. This snapshot lets you restore a datacenter from a backup in the event of an outage or catastrophic failure. You can save snapshot to a cloud storage bucket to ensure data persistence.

Data directory

Consul writes the Raft index to the data directory specified in the agent configuration with the data_dir parameter or -data_dir CLI flag. The data directory is a requirement for all agents, and should be durable across reboots.

Write-ahead log (WAL) LogStore backend

Consul logs the Raft index with the write-ahead log (WAL) LogStore backend. The WAL backend implements a traditional log with rotating, append-only log files, and it retains logs without affecting a cluster's write performance at scale.

Previous versions of Consul used BoltDB as the default LogStore backend. Refer to the WAL LogStore backend overview for more information. To use BoltDB instead of WAL, refer to Revert storage backend to BoltDB from WAL.