CloudWatch Metrics and Alarms for Spring Boot

Spring Boot's Micrometer integration publishes metrics to CloudWatch automatically — which means every service gets CPU, memory, HTTP latency, and JVM metrics in CloudWatch without writing any CloudWatch SDK code.

Publishing metrics from Spring Boot

• Add micrometer-registry-cloudwatch2 to the dependencies and configure management.metrics.export.cloudwatch.namespace — Micrometer pushes all registered metrics to CloudWatch in the configured namespace every 60 seconds
• Spring Boot auto-configures JVM metrics (GC pauses, heap usage, thread count), HTTP request metrics (latency, count, error rate), and HikariCP pool metrics — these appear in CloudWatch without any custom code

Custom metrics with Micrometer

• Counter.builder("orders.placed").register(meterRegistry).increment() adds a custom counter; Timer.builder("payment.processing").register(meterRegistry).record(duration) records timing
• Tag metrics with dimensions like service, env, and region — which means CloudWatch dashboards can filter and group by service or environment without creating separate metric namespaces

CloudWatch Alarms

• Define alarms on metrics with thresholds: HTTPServerRequests{status=5xx} above 1% of total requests; jvm.memory.used above 80% of max; P99 latency above 500ms
• Alarms can trigger SNS notifications (which notifies PagerDuty or Slack via webhook), Auto Scaling policies (scale out when CPU > 70%), or Lambda functions for automated remediation

CloudWatch Logs Insights

• Spring Boot JSON logs shipped to CloudWatch Logs can be queried with Logs Insights — fields @timestamp, level, message | filter level = "ERROR" | stats count() by bin(5m) counts errors per 5-minute window, useful for correlating spikes in logs with metric alarms