Fix Lost Primary

Alert Conditions

You can configure alert conditions in the project-level alert settings page to trigger alerts.

To learn more about the alert condition, see Replica set has no primary.

Common Triggers

• A network partition is preventing all members of the replica set from communicating with each other.

• There are not enough voting members in the replica set to keep a primary. A simple majority is required to have a primary. This can happen if too many hosts are down (or not responding) to replica set heartbeats.

• There is not an electable member when the primary goes down.

Fix the Immediate Problem

• Ensure a majority of votes are available in the replica set. If hosts have been shut down permanently, ensure they are removed from the replica set configuration.

• Use basic ping tests from one host to another to ensure they can communicate with each other.

Implement a Long-Term Solution

Make sure there is more than one data-bearing member with a priority greater than 0.

To learn more, see Replica Set Elections in the MongoDB manual.

Monitor Your Progress

View the following charts to monitor whether the deployment is exhausting resources:

• Normalized System CPU

  Monitor CPU usage to determine whether data is retrieved from disk instead of memory.

• Disk IOPs

  Monitor whether disk IOPS approaches the maximum provisioned IOPS. Determine whether the deployment can handle future workloads.

• Connections

  Monitor connections to determine whether the current connection limits are sufficient

To learn more, see View Deployment Metrics.