Amazon CloudWatch
- What it is
- Monitoring and observability service. Collects metrics, logs, and events and turns them into dashboards, alarms, and automated actions.
- Core pieces
- Metrics — numeric time-series data (CPU, request count, custom metrics). Basic monitoring is 5-minute intervals; detailed is 1-minute (extra cost).
- Logs — centralized log storage from EC2 (via agent), Lambda, VPC Flow Logs, and applications. Query with Logs Insights.
- Alarms — trigger on a threshold. Actions: notify via Amazon Simple Notification Service (SNS), trigger AWS Auto Scaling, stop/terminate/reboot an instance.
- Dashboards — cross-service, cross-Region visualization.
- Application Signals and Container Insights for application-level and container observability.
- The critical distinction
- Amazon CloudWatch = performance and operational health — “is it working, how fast, how much?”
- AWS CloudTrail = API audit trail — “who did what, when, from where?”
- AWS Config = configuration state and change history — “what does this resource look like, and did it drift?”
- Note on memory and disk
- Memory and disk usage are not default EC2 metrics — they require the CloudWatch agent inside the instance.