A platform team runs a microservices application on Amazon EKS. They want a single Amazon Managed Grafana workspace where on-call engineers can correlate two kinds of telemetry side by side: application time-series metrics already exposed in Prometheus exposition format, and the end-to-end distributed traces that show how long each request spends in every downstream service. The team does not want to run or patch any of the back-end stores themselves, and they want Grafana to query each telemetry store in place without copying the data into the workspace. Which two telemetry back ends should they add as Grafana data sources to satisfy this requirement? (Select TWO.)
- AAmazon Managed Service for Prometheus, ingesting the Prometheus-format metrics and queried by Grafana for the time-series panels. Correct
- BAmazon CloudTrail, queried by Grafana to provide the per-request latency breakdown across the downstream microservices.
- CAWS X-Ray, supplying the distributed traces so Grafana can show the time each request spends in every downstream service. Correct
- DA self-managed Prometheus server on Amazon EC2 scraping the pods, with Grafana querying that instance for the metrics panels.
- EAmazon CloudWatch Logs Insights, queried by Grafana to reconstruct the cross-service trace spans for each slow request.
Why A is correct: Amazon Managed Service for Prometheus stores the Prometheus-exposition metrics at scale with no servers to patch, and it is a native Managed Grafana data source for the time-series panels.
Why B is wrong: CloudTrail records management and data API activity for audit, so it is tempting as a central log store, but it does not produce per-request distributed traces or service latency breakdowns.
Why C is correct: AWS X-Ray captures end-to-end distributed traces and is a managed back end with no servers to run, and Grafana can query it directly to render the per-service latency breakdown.
Why D is wrong: A self-managed Prometheus on EC2 can store the metrics, but it leaves the team patching and scaling the time-series database, which the stated requirement explicitly rules out.
Why E is wrong: Logs Insights queries log events and is a valid Grafana source for logs, but it does not assemble cross-service trace spans, so it cannot deliver the distributed latency breakdown the stem needs.