fix: worker metrics crash + scrape d2-worker for Prometheus

infra/k8s/base/kustomization.yaml

@@ -13,5 +13,7 @@ resources:
   - api-ingress-tls.yaml
   - api-service.yaml
   - api-servicemonitor.yaml
+  - worker-metrics-service.yaml
+  - worker-servicemonitor.yaml
   - migrate-job.yaml
   - worker-deployment.yaml

infra/k8s/base/worker-deployment.yaml

@@ -30,6 +30,9 @@ spec:
         - name: worker
           image: ghcr.io/healthcare-monitoring-system/d2-data-intelligence:sha-a280ecc715a943d47c8b3ed635dc61e7f9f69704
           imagePullPolicy: Always
+          ports:
+            - name: metrics
+              containerPort: 8001
           command:
             - python
             - -m

infra/k8s/base/worker-metrics-service.yaml

@@ -0,0 +1,15 @@
+# Headless Prometheus scrape target for worker /metrics (ML + Kafka pipeline counters).
+apiVersion: v1
+kind: Service
+metadata:
+  name: d2-worker-metrics
+  labels:
+    app: d2-worker
+    d2-metrics: worker
+spec:
+  selector:
+    app: d2-worker
+  ports:
+    - name: metrics
+      port: 8001
+      targetPort: metrics

infra/k8s/base/worker-servicemonitor.yaml

@@ -0,0 +1,16 @@
+# Prometheus Operator: scrape d2-worker /metrics when release label matches the kube-stack.
+apiVersion: monitoring.coreos.com/v1
+kind: ServiceMonitor
+metadata:
+  name: d2-worker
+  labels:
+    release: monitoring
+spec:
+  selector:
+    matchLabels:
+      app: d2-worker
+      d2-metrics: worker
+  endpoints:
+    - port: metrics
+      path: /metrics
+      interval: 30s

src/d2/config.py

@@ -85,7 +85,7 @@ def influx_configured(self) -> bool:
     # Observability
     worker_metrics_enabled: bool = True
     worker_metrics_host: str = "0.0.0.0"
-    worker_metrics_port: int = 9102
+    worker_metrics_port: int = 8001
 
     # Optional Keycloak checks (D4 will supply values)
     keycloak_well_known_url: str | None = None
@@ -104,9 +104,6 @@ def ensure_asyncpg_driver_in_database_url(self) -> Self:
             self.database_url = "postgresql+asyncpg://" + raw.removeprefix("postgres://")
         return self
 
-    # Observability — port the worker exposes Prometheus metrics on. Set to 0 to disable.
-    worker_metrics_port: int = 8001
-
 
 @lru_cache
 def get_settings() -> Settings: