Metrics and Alarms
- Standard metrics come from AWS services; custom metrics are published with PutMetricData or agents.
- High-resolution custom metrics can use 1-second granularity, but they cost more than standard resolution.
- Datapoints-to-alarm lets you require sustained breaches, such as 3 out of 3 periods above a threshold.
- Composite alarms combine alarm states and help reduce alert noise.
Logs
- CloudWatch Logs stores log groups and log streams from applications, agents, and AWS services.
- Metric filters transform matching log events into CloudWatch metrics that can drive alarms.
- Logs Insights is used for ad hoc queries across log groups.
- Subscription filters stream logs to Lambda, Kinesis Data Streams, or Firehose.
Exam Cues
- Need API audit history: use CloudTrail, not CloudWatch metrics.
- Need application log search: use CloudWatch Logs Insights.
- Need dependency tracing: use AWS X-Ray.
- Need one alert only when several alarms are in ALARM: use a composite alarm.
Practice CloudWatch Questions
Put your knowledge to the test with practice questions.