feat: Optional direct OpenMetrics scraping for both remote and within-cluster monitoring

.github/workflows/test-remote-monitoring.yaml

@@ -13,6 +13,11 @@ on:
         required: true
         default: true
         type: boolean
+      use_direct_scraping:
+        description: "Scrape Memgraph's OpenMetrics endpoint directly instead of via mg-exporter (requires Memgraph >= 3.11)."
+        required: false
+        default: false
+        type: boolean
   pull_request:
     branches:
       - main
@@ -23,6 +28,12 @@ on:
       - "scripts/verify-remote-monitoring.sh"
       - ".github/workflows/test-remote-monitoring.yaml"
 
+# Default to the mg-exporter scrape path (works with the released chart appVersion).
+# Direct OpenMetrics scraping requires Memgraph >= 3.11, so it's a manual opt-in via
+# use_direct_scraping=true; pull_request runs always use mg-exporter.
+env:
+  USE_MG_EXPORTER: ${{ inputs.use_direct_scraping != true }}
+
 jobs:
   test-ha-remote-monitoring:
     name: HA remote monitoring
@@ -81,10 +92,15 @@ jobs:
 
       - name: Install HA chart with remote monitoring enabled
         run: |
+          if [[ "${USE_MG_EXPORTER}" == "true" ]]; then
+            echo "Scrape path: mg-exporter"
+            SCRAPE_ARGS=(--set prometheus.enabled=true --set prometheus.namespace=monitoring --set prometheus.serviceMonitor.enabled=false)
+          else
+            echo "Scrape path: direct OpenMetrics"
+            SCRAPE_ARGS=(--set prometheus.enabled=false --set scrapeMemgraphDirectly=true)
+          fi
           helm install memgraph-db ./charts/memgraph-high-availability \
-            --set prometheus.enabled=true \
-            --set prometheus.namespace=monitoring \
-            --set prometheus.serviceMonitor.enabled=false \
+            "${SCRAPE_ARGS[@]}" \
             --set vmagentRemote.enabled=true \
             --set vmagentRemote.namespace=monitoring \
             --set vmagentRemote.remoteWrite.url=http://monitoring-gateway.remote-monitoring.svc.cluster.local/api/v1/write \
@@ -104,7 +120,9 @@ jobs:
 
       - name: Wait for monitoring side components
         run: |
-          kubectl rollout status deployment/mg-exporter -n monitoring --timeout=300s
+          if [[ "${USE_MG_EXPORTER}" == "true" ]]; then
+            kubectl rollout status deployment/mg-exporter -n monitoring --timeout=300s
+          fi
           kubectl rollout status deployment/memgraph-db-memgraph-high-availability-vmagent -n monitoring --timeout=300s
 
       - name: Verify metrics and logs reach monitoring cluster
@@ -174,13 +192,18 @@ jobs:
 
       - name: Install standalone chart with remote monitoring enabled
         run: |
+          if [[ "${USE_MG_EXPORTER}" == "true" ]]; then
+            echo "Scrape path: mg-exporter"
+            SCRAPE_ARGS=(--set prometheus.enabled=true --set prometheus.namespace=monitoring --set prometheus.serviceMonitor.enabled=false)
+          else
+            echo "Scrape path: direct OpenMetrics"
+            SCRAPE_ARGS=(--set prometheus.enabled=false --set scrapeMemgraphDirectly=true)
+          fi
           helm install memgraph-db ./charts/memgraph \
             --set service.enableHttpMonitoring=true \
             --set service.enableWebsocketMonitoring=true \
             --set-json 'memgraphConfig=["--data-directory=/var/lib/memgraph/mg_data","--also-log-to-stderr=true","--log-level=TRACE"]' \
-            --set prometheus.enabled=true \
-            --set prometheus.namespace=monitoring \
-            --set prometheus.serviceMonitor.enabled=false \
+            "${SCRAPE_ARGS[@]}" \
             --set vmagentRemote.enabled=true \
             --set vmagentRemote.namespace=monitoring \
             --set vmagentRemote.remoteWrite.url=http://monitoring-gateway.remote-monitoring.svc.cluster.local/api/v1/write \
@@ -200,7 +223,9 @@ jobs:
       - name: Wait for monitoring side components
         run: |
           kubectl rollout status statefulset/memgraph-db -n default --timeout=300s
-          kubectl rollout status deployment/mg-exporter -n monitoring --timeout=300s
+          if [[ "${USE_MG_EXPORTER}" == "true" ]]; then
+            kubectl rollout status deployment/mg-exporter -n monitoring --timeout=300s
+          fi
           kubectl rollout status deployment/memgraph-db-vmagent -n monitoring --timeout=300s
 
       - name: Verify metrics and logs reach monitoring cluster

README.md

@@ -96,7 +96,9 @@ Both Memgraph charts support optional remote observability:
 This works with VictoriaMetrics/VictoriaLogs, and with other backends that expose compatible Prometheus remote-write and Loki endpoints.
 
 ### Prerequisites
-- Enable chart-level Prometheus exporter (`prometheus.enabled=true`).
+- Provide a metrics source for `vmagentRemote`, either:
+  - the chart-level Prometheus exporter (`prometheus.enabled=true`) — the default, or
+  - direct scraping of Memgraph's OpenMetrics endpoint (top-level `scrapeMemgraphDirectly=true`), which needs no exporter (`prometheus.enabled=false`). See [Scrape Memgraph directly](#scrape-memgraph-directly-without-mg-exporter).
 - Use a secret containing credentials for your remote endpoints (required for `vmagentRemote`; optional for `vectorRemote`).
 - For standalone chart, enable Memgraph monitoring ports:
   - `service.enableHttpMonitoring=true`
@@ -106,6 +108,38 @@ This works with VictoriaMetrics/VictoriaLogs, and with other backends that expos
   - HA chart: add `--monitoring-port=<vectorRemote.websocketPort>` and `--monitoring-address=0.0.0.0` to each instance's `args`
 - If `vmagentRemote.enabled=true` and you only need remote_write, set `prometheus.serviceMonitor.enabled=false` to avoid duplicate scraping of `mg-exporter` by both vmagent and kube-prometheus.
 
+### Scrape Memgraph directly (without mg-exporter)
+By default `vmagentRemote` scrapes the `mg-exporter`, which requires `prometheus.enabled=true`. To scrape Memgraph's OpenMetrics endpoint directly instead — with no exporter deployed — set the top-level `scrapeMemgraphDirectly=true` and leave `prometheus.enabled=false`:
+
+```yaml
+# Top-level master switch: Memgraph serves OpenMetrics and all scrapers hit it directly.
+scrapeMemgraphDirectly: true
+
+prometheus:
+  enabled: false  # mg-exporter not needed
+
+vmagentRemote:
+  enabled: true
+  namespace: monitoring
+  remoteWrite:
+    url: "https://<prom-remote-write>/api/v1/write"
+    basicAuth:
+      secretName: monitoring-basic-auth
+      usernameKey: username
+      passwordKey: password
+  externalLabels:
+    cluster_id: "memgraph-ha"
+    service_name: "Memgraph HA"
+    cluster_env: "dev"
+```
+
+When `scrapeMemgraphDirectly=true`, the chart runs each instance with `--metrics-format=OpenMetrics` and exposes the metrics port automatically (for the standalone chart you do not need to set `service.enableHttpMonitoring`). All other `vmagentRemote` / `vectorRemote` settings are the same as the examples below. The metrics endpoint is served over plain HTTP.
+
+### In-cluster scraping with kube-prometheus-stack
+For an in-cluster Prometheus (e.g. `kube-prometheus-stack`), set `prometheus.serviceMonitor.enabled=true` to provision a `ServiceMonitor`. By default it scrapes the `mg-exporter`; set the top-level `scrapeMemgraphDirectly=true` to scrape Memgraph's OpenMetrics endpoint directly instead (no exporter; requires Memgraph >= 3.11). The same `scrapeMemgraphDirectly` switch governs both the in-cluster ServiceMonitor and the remote `vmagentRemote`.
+
+Set `prometheus.grafanaDashboard.enabled=true` to ship the bundled "Memgraph OpenMetrics" dashboard as a ConfigMap (labelled `grafana_dashboard`) for the Grafana sidecar to auto-load. It uses a datasource template variable, so it binds to Grafana's default Prometheus. The ConfigMap must live in a namespace the sidecar watches — set `prometheus.grafanaDashboard.namespace` to Grafana's namespace (or run the sidecar with `searchNamespace: ALL`).
+
 ### Standalone chart example
 ```yaml
 prometheus: