Add default Prometheus EMF metric_declaration when target allocator is enabled

[musa-asad]

Summary

When the Target Allocator is enabled (agent.prometheus.targetAllocator.enabled=true), the chart renders .spec.prometheus with only scrape_configs and no EMF metric-extraction block. The agent scrapes targets but produces 0 CloudWatch metrics — without metric_declaration, there's no instruction telling the agent which series to export.

This change injects a default emf_processor.metric_declaration block gated behind the TA toggle. Default behavior for non-TA users is unchanged.

Changes

• templates/_helpers.tpl — render prometheus.emf_processor.metric_declaration when TA enabled and emfProcessor declarations are present
• values.yaml — add default agent.prometheus.emfProcessor with metricNamespace + metricDeclarations for node-exporter and nginx

Test Output

helm lint:

$ helm lint charts/amazon-cloudwatch-observability --set clusterName=test-cluster --set region=us-west-2
==> Linting charts/amazon-cloudwatch-observability
[INFO] Chart.yaml: icon is recommended

1 chart(s) linted, 0 chart(s) failed

Gating proof (EMF block renders only when TA enabled):

$ helm template ... --set agent.prometheus.targetAllocator.enabled=true | grep -c metric_declaration
1
$ helm template ... --set agent.prometheus.targetAllocator.enabled=false | grep -c metric_declaration
0

CloudWatch metrics before/after:

ContainerInsights/Prometheus metrics (Helm path):   0 → 7
Same default on add-on install path:                0 → 7

OTEL CI flag matrix (no regression):

$ bash charts/amazon-cloudwatch-observability/tests/flag_matrix.sh
[State #1] enabled=false logs=false containerLogs=false  PASS
[State #2] enabled=false logs=false containerLogs=true   PASS
[State #3] enabled=false logs=true  containerLogs=false  PASS
[State #4] enabled=false logs=true  containerLogs=true   PASS
[State #5] enabled=true  logs=false containerLogs=false  PASS
[State #6] enabled=true  logs=false containerLogs=true   PASS
[State #7] enabled=true  logs=true  containerLogs=false  PASS
[State #8] enabled=true  logs=true  containerLogs=true   PASS
All 8 cases passed.

Related

• Fix Target Allocator startup crashes, PrometheusCR watcher, and Prometheus config pod restart aws/amazon-cloudwatch-agent-operator#386 — TA startup crash fix + pod restart on config change
• Do not warn on target allocator config load for non-TA Prometheus configs aws/amazon-cloudwatch-agent#2157 — agent-side fix for spurious TA warning

[musa-asad]

Closing — this adds an opinionated default metric_declaration rather than fixing a bug. Users are expected to provide their own metric_declaration in configuration-values when enabling Prometheus scraping, as documented in the EKS add-on usage guide.