Add default Prometheus EMF metric_declaration when target allocator is enabled

Author: musa-asad

charts/amazon-cloudwatch-observability/templates/_helpers.tpl

@@ -89,6 +89,24 @@ Logic:
   {{- $needsLogs = true -}}
   {{- $_ := set $metricsCollected "kubernetes" (dict "enhanced_container_insights" true) -}}
 {{- end -}}
+{{/* Prometheus EMF: when the Target Allocator is enabled, wire the awsemf path for the
+     Prometheus pipeline by adding logs.metrics_collected.prometheus.emf_processor with
+     metric_declarations. Without this, scraped Prometheus values reach CloudWatch Logs
+     but produce no CloudWatch metrics (Phase-3 findings F1/F5). */}}
+{{- if $ctx.Values.agent.prometheus.targetAllocator.enabled -}}
+  {{- $emf := $ctx.Values.agent.prometheus.emfProcessor -}}
+  {{- if and $emf $emf.metricDeclarations -}}
+    {{- $needsLogs = true -}}
+    {{/* CloudWatch Agent input schema key is "metric_declaration" (singular array);
+         emitting "metric_declarations" (plural) fails config-translator validation
+         ("Additional property metric_declarations is not allowed"). */}}
+    {{- $emfProcessor := dict "metric_declaration" $emf.metricDeclarations -}}
+    {{- if $emf.metricNamespace -}}
+      {{- $_ := set $emfProcessor "metric_namespace" $emf.metricNamespace -}}
+    {{- end -}}
+    {{- $_ := set $metricsCollected "prometheus" (dict "emf_processor" $emfProcessor) -}}
+  {{- end -}}
+{{- end -}}
 {{- if $needsLogs -}}
   {{- $_ := set $config "logs" (dict "metrics_collected" $metricsCollected) -}}
 {{- end -}}

charts/amazon-cloudwatch-observability/values.yaml

@@ -763,6 +763,41 @@ agent:
   otelConfig: # optional YAML config that can be provided to combine with the build-default-otel-config output
   prometheus:
     config:
+    ## EMF (Embedded Metric Format) processor for the Prometheus pipeline.
+    ## When the Target Allocator is enabled (agent.prometheus.targetAllocator.enabled=true),
+    ## the CloudWatch Agent converts scraped Prometheus metrics into CloudWatch metrics via
+    ## EMF using these metric_declarations. WITHOUT metric_declarations the scraped values
+    ## are written to CloudWatch Logs but carry no `_aws.CloudWatchMetrics` block, so NO
+    ## CloudWatch metrics are produced. These declarations wire the awsemf path for the
+    ## Prometheus pipeline. This block is only rendered into the agent config when the Target
+    ## Allocator is enabled, so default behavior for non-Prometheus users is unchanged.
+    ## Override `emfProcessor.metricDeclarations` to scrape additional jobs/metrics, and
+    ## `emfProcessor.metricNamespace` to change the destination CloudWatch namespace.
+    emfProcessor:
+      metricNamespace: ContainerInsights/Prometheus
+      metricDeclarations:
+        # node-exporter (PodMonitor "node-exporter") — host CPU/memory/disk/network
+        - source_labels: ["job"]
+          label_matcher: ".*node-exporter.*"
+          dimensions:
+            - ["ClusterName"]
+            - ["ClusterName", "Namespace"]
+          metric_selectors:
+            - "^node_cpu_seconds_total$"
+            - "^node_memory_MemAvailable_bytes$"
+            - "^node_memory_MemTotal_bytes$"
+            - "^node_filesystem_avail_bytes$"
+            - "^node_filesystem_size_bytes$"
+            - "^node_network_receive_bytes_total$"
+            - "^node_network_transmit_bytes_total$"
+        # nginx (ServiceMonitor "nginx") — application request/connection metrics
+        - source_labels: ["job"]
+          label_matcher: ".*nginx.*"
+          dimensions:
+            - ["ClusterName"]
+            - ["ClusterName", "Namespace"]
+          metric_selectors:
+            - "^nginx_.*$"
     targetAllocator:
       enabled: false
       image: