Replica Set Alert Evaluation: Primary‑Only vs Per‑Node Assessment in MongoDB Cloud Manager/Atlas

Answer: Alert scope determines evaluation

In MongoDB Cloud Manager and Atlas, replication‑lag alerts are not hard‑coded to the primary or to each secondary; they are evaluated according to the Scope option you set when creating or editing an alert.

Confirmed facts

• The Alerts → Settings UI shows a Scope dropdown with two choices: Primary only and All members (per‑node).
• When Scope is Primary only, the alert condition is evaluated once per replica set, using the lag value reported by the primary.
• When Scope is All members, the condition is evaluated independently on each host that is currently reachable; a separate alert incident is generated for any member that exceeds the threshold.
• If a secondary steps down or becomes unavailable, a per‑node alert will not fire for that host (no data to evaluate), while a primary‑only alert continues to reflect the primary’s view of lag.

How to verify or change the setting

1. Log in to Cloud Manager or Atlas.
2. Navigate to Alerts → Settings.
3. Find the replication‑lag alert (e.g., “Oplog window ahead of primary” or custom lag metric).
4. Click the alert’s name to edit it.
5. Locate the Scope field; its current value tells you whether the alert is primary‑only or per‑node.
6. Change it as needed and save.

If you are unsure which scope is currently applied to your replication‑lag alert, please let me know the present Scope setting so I can confirm whether the alert will fire on any secondary or only when the primary’s view exceeds the threshold.