See the list of services requiring explicit metric enablement:
| Service | Enablement method | Reference |
|---|---|---|
Amazon ECS (EC2 launch type) | Enable Container Insights | |
Amazon EKS | Enable Container Insights | |
Amazon S3 | Enable Request Metrics per bucket | |
Amazon CloudFront | Enable Additional Metrics per distribution | |
Amazon API Gateway v2 HTTP APIs | Enable Detailed Metrics per stage | |
Amazon API Gateway v2 WebSocket APIs | Enable Detailed Metrics per stage | |
Amazon API Gateway v1 REST APIs | Enable Detailed Metrics per stage | |
Amazon RDS | Enable Enhanced Monitoring | |
Amazon EC2 | Enable Detailed Monitoring | |
Amazon EMR | Deploy CloudWatch agent for enhanced metrics | |
Amazon SageMaker 1 | Configure metric emission | |
Amazon Glue | Enable metrics via job config or CLI | |
Amazon Route 53 | Create Health Checks |
Enablement required for BYOC, model monitor, multi-model endpoints.
See the specifics for SageMaker:
| SageMaker component | Enablement method | Reference |
|---|---|---|
Training Jobs | Metrics are emitted automatically, but you can define custom metrics in your training script. You can also use SageMaker Debugger for fine-grained metrics. | |
Endpoints (Real-time Inference) | Metrics are emitted automatically for standard endpoints. For multi-model endpoints, metrics like | |
Inference Pipelines | Metrics are emitted automatically, but you must configure each container properly to ensure full metric visibility. | |
Model Monitor (Quality Monitoring) | You must create a monitoring schedule and configure the job to emit metrics to CloudWatch. | |
Bring Your Own Container (BYOC) | You must emit metrics manually from your container to CloudWatch using the |