At a glance
Replica Set Members (state) is a live table listing every member of the replica set with its hostname and current state string:PRIMARY,SECONDARY,RECOVERING,STARTUP2,ARBITER,DOWN, and so on. It is the single-glance answer to “is the cluster healthy and who is taking writes right now?” A well-formed set shows exactly onePRIMARYand the restSECONDARY(plus anyARBITER); anything else (a missing primary, a member stuck inRECOVERING, or a node reportingDOWN) means replication or failover is mid-flight and writes may be blocked.
What it tracks
The card is populated directly fromrs.status().members, taking each member’s name (host:port) and stateStr and rendering them as rows in real time. There is no aggregation window; it is the instantaneous topology of the set. The healthy shape is one and only one PRIMARY with all data-bearing peers in SECONDARY. Transient states are normal during maintenance: STARTUP2 appears while a freshly added member performs its initial sync, and RECOVERING appears briefly during rollback or after a node rejoins. The states that warrant attention are a persistent RECOVERING (the member cannot catch up), DOWN or (not reachable/healthy) (the member is unreachable from the primary), or the absence of any PRIMARY (the set has lost quorum and is read-only until an election completes). This card carries no numeric alert of its own; its alerting partner is the Replica Set Member Lag >10s or in RECOVERING State card.
Reconciling against the source
Confirm against MongoDB directly by runningrs.status() from mongosh on any member and reading the members[].stateStr field, or rs.isMaster() / db.hello() for the quick “who is primary” answer. On Atlas, the cluster topology view shows the same per-node roles and health. The Nerve Centre table mirrors rs.status() field-for-field, so any difference is a polling-interval timing gap during an in-progress election.