replication_group_members

replication_group_members is a view exposed by performance_schema that provides a real-time snapshot of the membership in a MySQL Group Replication cluster. Each row represents one node and reports its current state and role within the distributed consensus protocol. On a standalone server or one using traditional replication (asynchronous or semisync), the view returns no rows.

How it works #

The view aggregates information exchanged through the group communication plugin (GCS/Paxos). The main columns are:

SELECT MEMBER_HOST, MEMBER_PORT, MEMBER_STATE, MEMBER_ROLE
FROM performance_schema.replication_group_members;

A node in RECOVERING state is applying transaction backfill received from the donor. UNREACHABLE means the other members have lost contact but have not yet expelled the node from the group.

Operational context #

This view is the first checkpoint during maintenance operations on a Group Replication cluster: rolling patching, planned failover, post-restart verification. Before stopping a node, confirm that all members are ONLINE and that exactly one PRIMARY exists (Single-Primary mode). In Multi-Primary mode, all ONLINE nodes show PRIMARY.

The view does not replace continuous monitoring: alerting on UNREACHABLE or ERROR states requires polling it periodically or integrating it with an orchestration layer (ProxySQL, MySQL Router, Orchestrator).