From ed7d2aa72780daeb60de593d4f159b87e1f69cd4 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Torjus=20H=C3=A5kestad?= <torjus@usit.uio.no>
Date: Mon, 9 Feb 2026 15:58:28 +0100
Subject: [PATCH] grafana: add deployment metrics to nixos-fleet dashboard

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
---
 services/grafana/dashboards/nixos-fleet.json | 74 ++++++++++++++++++--
 1 file changed, 69 insertions(+), 5 deletions(-)

diff --git a/services/grafana/dashboards/nixos-fleet.json b/services/grafana/dashboards/nixos-fleet.json
index 5a7b61c..733bf39 100644
--- a/services/grafana/dashboards/nixos-fleet.json
+++ b/services/grafana/dashboards/nixos-fleet.json
@@ -29,7 +29,7 @@
       "id": 1,
       "title": "Hosts Behind Remote",
       "type": "stat",
-      "gridPos": {"h": 4, "w": 6, "x": 0, "y": 0},
+      "gridPos": {"h": 4, "w": 4, "x": 0, "y": 0},
       "datasource": {"type": "prometheus", "uid": "prometheus"},
       "targets": [
         {
@@ -63,7 +63,7 @@
       "id": 2,
       "title": "Hosts Needing Reboot",
       "type": "stat",
-      "gridPos": {"h": 4, "w": 6, "x": 6, "y": 0},
+      "gridPos": {"h": 4, "w": 4, "x": 4, "y": 0},
       "datasource": {"type": "prometheus", "uid": "prometheus"},
       "targets": [
         {
@@ -97,7 +97,7 @@
       "id": 3,
       "title": "Total Hosts",
       "type": "stat",
-      "gridPos": {"h": 4, "w": 4, "x": 12, "y": 0},
+      "gridPos": {"h": 4, "w": 3, "x": 8, "y": 0},
       "datasource": {"type": "prometheus", "uid": "prometheus"},
       "targets": [
         {
@@ -124,7 +124,7 @@
       "id": 4,
       "title": "Nixpkgs Age",
       "type": "stat",
-      "gridPos": {"h": 4, "w": 4, "x": 16, "y": 0},
+      "gridPos": {"h": 4, "w": 3, "x": 11, "y": 0},
       "datasource": {"type": "prometheus", "uid": "prometheus"},
       "targets": [
         {
@@ -158,7 +158,7 @@
       "id": 5,
       "title": "Hosts Up-to-date",
       "type": "stat",
-      "gridPos": {"h": 4, "w": 4, "x": 20, "y": 0},
+      "gridPos": {"h": 4, "w": 3, "x": 14, "y": 0},
       "datasource": {"type": "prometheus", "uid": "prometheus"},
       "targets": [
         {
@@ -182,6 +182,70 @@
         "graphMode": "none"
       }
     },
+    {
+      "id": 13,
+      "title": "Deployments (24h)",
+      "type": "stat",
+      "gridPos": {"h": 4, "w": 3, "x": 17, "y": 0},
+      "datasource": {"type": "prometheus", "uid": "prometheus"},
+      "targets": [
+        {
+          "expr": "sum(increase(homelab_deploy_deployments_total{status=\"completed\"}[24h]))",
+          "legendFormat": "Deployments",
+          "refId": "A"
+        }
+      ],
+      "fieldConfig": {
+        "defaults": {
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [{"color": "blue", "value": null}]
+          },
+          "noValue": "0",
+          "decimals": 0
+        }
+      },
+      "options": {
+        "reduceOptions": {"calcs": ["lastNotNull"]},
+        "colorMode": "value",
+        "graphMode": "none"
+      },
+      "description": "Total successful deployments in the last 24 hours"
+    },
+    {
+      "id": 14,
+      "title": "Avg Deploy Time",
+      "type": "stat",
+      "gridPos": {"h": 4, "w": 4, "x": 20, "y": 0},
+      "datasource": {"type": "prometheus", "uid": "prometheus"},
+      "targets": [
+        {
+          "expr": "sum(increase(homelab_deploy_deployment_duration_seconds_sum{success=\"true\"}[24h])) / sum(increase(homelab_deploy_deployment_duration_seconds_count{success=\"true\"}[24h]))",
+          "legendFormat": "Avg Time",
+          "refId": "A"
+        }
+      ],
+      "fieldConfig": {
+        "defaults": {
+          "unit": "s",
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {"color": "green", "value": null},
+              {"color": "yellow", "value": 30},
+              {"color": "red", "value": 60}
+            ]
+          },
+          "noValue": "-"
+        }
+      },
+      "options": {
+        "reduceOptions": {"calcs": ["lastNotNull"]},
+        "colorMode": "value",
+        "graphMode": "none"
+      },
+      "description": "Average deployment duration over the last 24 hours (yellow >30s, red >60s)"
+    },
     {
       "id": 6,
       "title": "Fleet Status",