An Atlas alert is an automated notification triggered when a configured database, user, account, or cluster condition is met. Alerts help administrators catch performance and security problems without continually checking metrics. When triggered, Atlas displays a warning icon on the cluster and sends a notification through the configured channel. section 21.2 section 21.3.1
Choose conditions that call for a response, then set thresholds and notification channels:
- Watch capacity and contention: high CPU steal can signal strain on burstable shared-CPU environments; more than 100 queued operations can signal delays; connections nearing 80%–90% of capacity can signal a scaling need. section 21.3.1
- Watch query targeting for inefficient queries. A high value can point to queries that need optimization. section 21.3.1
- Watch security-sensitive activity, such as unusual access patterns, failed logins, and changes to critical settings. section 20.7
atlas alerts settings create --event OUTSIDE_METRIC_THRESHOLD --enabled \ --metricName QUERY_TARGETING_SCANNED_OBJECTS_PER_RETURNED \ --metricOperator GREATER_THAN --metricThreshold 1000 \ --metricUnits RAW --notificationType EMAIL --notificationEmailEnabled \ --notificationEmailAddress your-email@example.com \ --notificationIntervalMin 15
This Atlas CLI command creates an enabled query-targeting metric alert. It triggers when scanned objects per returned result exceed 1,000 and sends email notifications to the supplied address at a 15-minute interval. Replace the example address and choose a threshold appropriate to the condition you want to catch. section 21.3.1
Use `atlas alerts settings list` to inspect configured alerts and `atlas alerts settings describe` with an alert ID to inspect one setting. Pair notifications with deployment metrics and logs: metrics provide performance history, while logs provide activity history for troubleshooting. Have the team responsible for reliability review and respond to alerts promptly. section 21.3.1 section 21.3 section 21.2
