From 78a56b78780961086a627d5d9983d73df5c55671 Mon Sep 17 00:00:00 2001
From: Giovanni Ferri <giovanni@syscode.uk>
Date: Sun, 13 Sep 2026 10:35:08 +0100
Subject: [PATCH 1/2] feat(unraid): scrape Imp metrics with Alloy

---
 values/base/grafana-k8s-monitoring.yaml       | 80 +++++++++++++++++++
 .../unraid-lab/grafana-k8s-monitoring.yaml    |  8 ++
 2 files changed, 88 insertions(+)
 create mode 100644 values/clusters/unraid-lab/grafana-k8s-monitoring.yaml

diff --git a/values/base/grafana-k8s-monitoring.yaml b/values/base/grafana-k8s-monitoring.yaml
index 33415b8..ee7f907 100644
--- a/values/base/grafana-k8s-monitoring.yaml
+++ b/values/base/grafana-k8s-monitoring.yaml
@@ -136,6 +136,86 @@ collectors:
         scrape_protocols = ["OpenMetricsText1.0.0", "OpenMetricsText0.0.1", "PrometheusText0.0.4"]
         forward_to = [prometheus.relabel.talos_etcd_disk_latency.receiver]
       }
+
+      // Imp publishes three small Prometheus endpoints. Discover only its
+      // ready metric ports; this avoids chart changes and does not broaden
+      // application scraping beyond imp-system.
+      discovery.kubernetes "imp_metrics" {
+        role = "pod"
+        namespaces {
+          names = ["imp-system"]
+        }
+        selectors {
+          role = "pod"
+          label = "app.kubernetes.io/name=imp"
+        }
+      }
+
+      discovery.relabel "imp_metrics" {
+        targets = discovery.kubernetes.imp_metrics.targets
+
+        rule {
+          action = "keep"
+          source_labels = ["__meta_kubernetes_pod_label_app_kubernetes_io_instance"]
+          regex = "imp"
+        }
+        rule {
+          action = "keep"
+          source_labels = ["__meta_kubernetes_pod_label_app_kubernetes_io_component"]
+          regex = "operator|agent|runtime"
+        }
+        rule {
+          action = "keep"
+          source_labels = ["__meta_kubernetes_pod_phase"]
+          regex = "Running"
+        }
+        rule {
+          action = "keep"
+          source_labels = ["__meta_kubernetes_pod_ready"]
+          regex = "true"
+        }
+        rule {
+          action = "keep"
+          source_labels = ["__meta_kubernetes_pod_container_init"]
+          regex = "false"
+        }
+        rule {
+          action = "keep"
+          source_labels = ["__meta_kubernetes_pod_container_port_name"]
+          regex = "metrics"
+        }
+        rule {
+          source_labels = ["__meta_kubernetes_namespace"]
+          target_label = "namespace"
+        }
+        rule {
+          source_labels = ["__meta_kubernetes_pod_name"]
+          target_label = "pod"
+        }
+        rule {
+          source_labels = ["__meta_kubernetes_pod_container_name"]
+          target_label = "container"
+        }
+        rule {
+          source_labels = ["__meta_kubernetes_pod_label_app_kubernetes_io_component"]
+          target_label = "component"
+        }
+        rule {
+          replacement = "integrations/imp"
+          target_label = "job"
+        }
+      }
+
+      prometheus.scrape "imp_metrics" {
+        targets = discovery.relabel.imp_metrics.output
+        scrape_interval = "30s"
+        scrape_timeout = "10s"
+        scrape_protocols = ["OpenMetricsText1.0.0", "OpenMetricsText0.0.1", "PrometheusText0.0.4"]
+        clustering {
+          enabled = true
+        }
+        forward_to = [prometheus.remote_write.grafana_cloud_metrics.receiver]
+      }
   alloy-singleton:
     presets: [singleton]
 telemetryServices:
diff --git a/values/clusters/unraid-lab/grafana-k8s-monitoring.yaml b/values/clusters/unraid-lab/grafana-k8s-monitoring.yaml
new file mode 100644
index 0000000..10cd4d6
--- /dev/null
+++ b/values/clusters/unraid-lab/grafana-k8s-monitoring.yaml
@@ -0,0 +1,8 @@
+# Grafana Kubernetes Monitoring must be allowed to discover Imp's pod metrics.
+# The base profile remains limited to kube-system so OCI is unaffected.
+annotationAutodiscovery:
+  namespaces: [kube-system, imp-system]
+  pods:
+    enabled: true
+    labelSelectors:
+      app.kubernetes.io/name: imp

From 493b0c98e75ea0cb51711074c88001b611935323 Mon Sep 17 00:00:00 2001
From: Giovanni Ferri <giovanni@syscode.uk>
Date: Sun, 13 Sep 2026 10:38:50 +0100
Subject: [PATCH 2/2] fix(unraid): schedule Grafana Alloy operator

---
 values/clusters/unraid-lab/grafana-k8s-monitoring.yaml | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/values/clusters/unraid-lab/grafana-k8s-monitoring.yaml b/values/clusters/unraid-lab/grafana-k8s-monitoring.yaml
index 10cd4d6..54e2986 100644
--- a/values/clusters/unraid-lab/grafana-k8s-monitoring.yaml
+++ b/values/clusters/unraid-lab/grafana-k8s-monitoring.yaml
@@ -6,3 +6,11 @@ annotationAutodiscovery:
     enabled: true
     labelSelectors:
       app.kubernetes.io/name: imp
+
+# Every usable Unraid node is a tainted control-plane node. The Alloy Operator
+# must tolerate that taint before it can create the metrics collector.
+alloy-operator:
+  tolerations:
+    - key: node-role.kubernetes.io/control-plane
+      operator: Exists
+      effect: NoSchedule