The completion time of a daily pipeline is gradually getting later.
Detect the delay before a failure and notify the owner.
Which monitoring approach is most appropriate?
Publish processing time, queue depth, and unprocessed records as CloudWatch metrics and alarm on useful thresholds or anomalies.
Detailed explanation
Time-series metrics and alarms reveal degradation before a binary success check fails.
Time-series metrics and alarms reveal degradation before a binary success check fails.
A binary result hides gradual performance degradation.
A binary result hides gradual performance degradation.
API count is not a direct measure of processing delay or backlog.
API count is not a direct measure of processing delay or backlog.
Storage transitions do not detect processing latency.
Storage transitions do not detect processing latency.
Check official documentation
Compare the following points with the official documentation.
Amazon CloudWatch custom metrics、alarms、anomaly detectionの公式説明を確認する。
What to confirm
成功/失敗だけの監視と遅延メトリクスの違いを説明できる。
Key points
- CloudWatch
- Metric
- Latency
- Alarm
Notes
- Environment: AWS公式DEA-C01試験ガイドと各サービスの公式ドキュメントの確認
- Confirm current AWS service behavior in the official DEA-C01 documentation.
Foundation review
SLI selection
Choose completion time, record age, throughput, and error rate close to user impact.
Noise control
Use a useful threshold or anomaly model so brief normal variation does not create alert fatigue.