High Availability

This section covers the configuration of High Availability (HA) features in Apache Ozone.

📄️ SCM HA Configuration

Configuration

📄️ OM HA Configuration

Ozone Manager (OM) High Availability ensures that there is no single point of failure for the metadata-manager node responsible for key space management. In HA mode, the internal state is replicated via RAFT (with Apache Ratis) across multiple Ozone Manager instances.

📄️ Client Failover

Overview

📄️ Listener OM

A Listener Ozone Manager (Listener OM) is a read-only member of an OM high availability (HA) group. It applies the same Ratis log as voting members so its metadata stays current, but it does not vote in elections or on the Raft write path. You can direct read traffic to Listener OMs to spread load and improve read-heavy workloads.

📄️ OM Follower Read

OM follower read lets Ozone clients send eligible read-only OM requests to non-leader OMs in an HA service. This can reduce read load on the OM leader while preserving the normal leader path for writes and for read requests that are not eligible for follower routing.