Ahmad Gul Zadran

Homelab monitoring, Part 3: Ready-made dashboards, sorted into folders

Ready-made Grafana dashboards for Proxmox, your NAS and UniFi, loaded from files, with fixes for panels that show wrong values.

intermediate30 minupdated 2026-10-13dashboardsgrafanamonitoringproxmoxunifiDownload script.sh

Before you start

  • The monitoring stack from Part 1 and the data sources from Part 2.
  • Fill in Your values below and every command on this page updates to match.

Files for this guide

The commands below download these for you. Open one to read it, or import a dashboard by hand.

Your values

Fill these in and every command below updates to match. Nothing is sent anywhere.

Run this at your own risk. These steps worked on my hardware; yours may differ. Take a backup first, read each command before running it, and see the disclaimer.

Your data is now arriving (Parts 1 and 2). Now you make it visible. Instead of building dashboards by clicking in the browser, you save them as files in folders. Grafana loads them automatically. The big advantage: if the container ever breaks, the dashboards come back exactly as they were.

Folder Dashboard Source
Servers Node Exporter Full grafana.com, ID 1860
Servers Proxmox via Prometheus grafana.com, ID 10347, with a CPU fix
Servers NAS Pools & Health download from this site
Network 4 × UniFi-Poller GitHub, with fixes
Dashboards sorted into folders
Dashboards sorted into folders

Everything below runs in the monitoring container (pct enter 140 from the Proxmox host shell).

Step 1: Tell Grafana where the dashboards live

bash
mkdir -p /opt/monitoring/grafana/provisioning/dashboards/files/{Servers,Network}

cat > /opt/monitoring/grafana/provisioning/dashboards/dashboards.yaml <<'EOF'
apiVersion: 1
providers:
  - name: homelab
    orgId: 1
    type: file
    disableDeletion: true
    allowUiUpdates: false
    updateIntervalSeconds: 60
    options:
      path: /etc/grafana/provisioning/dashboards/files
      foldersFromFilesStructure: true
EOF
Setting Meaning
foldersFromFilesStructure Each subfolder (Servers, Network…) becomes a Grafana folder
allowUiUpdates: false The files are the master copy. Changes in the browser can't be saved by accident (see "Changing a dashboard" below).
updateIntervalSeconds: 60 A changed file shows up in Grafana within a minute

Step 2: A small helper

Dashboards from grafana.com contain a placeholder (${DS_PROMETHEUS}) that the browser import normally fills in. This helper does the same in a file. Paste it once per shell session:

bash
fix_ds() {
  jq 'walk(if type == "string" then gsub("\\$\\{DS_PROMETHEUS\\}"; "prometheus") else . end)
      | del(.__inputs, .__requires, .__elements) | .id = null' "$1"
}
cd /tmp

Step 3: Node Exporter Full

The best-known dashboard for Linux servers: CPU, RAM, disks, network, temperatures, all in detail.

bash
curl -fsSL -o 1860.json https://grafana.com/api/dashboards/1860/revisions/latest/download
fix_ds 1860.json > /opt/monitoring/grafana/provisioning/dashboards/files/Servers/node-exporter-full.json

Step 4: Proxmox via Prometheus, with the CPU fix

This dashboard shows every VM and container. But out of the box, its CPU gauges for the host are wrong. Depending on the version, they show only part of the real load, or impossible values like 793 %. The fix takes the host CPU from node_exporter (Part 1) instead, which is correct.

bash
curl -fsSL -o 10347.json https://grafana.com/api/dashboards/10347/revisions/4/download
CPU='100 - avg(rate(node_cpu_seconds_total{mode="idle", instance=~"$instance:.*"}[1m])) * 100'
fix_ds 10347.json | jq --arg cpu "$CPU" '
    (.. | objects | select(.id? == 7 and has("targets"))) |=
        (.targets |= [.[0] | .expr = $cpu]
         | .fieldConfig.defaults.unit = "percent" | .fieldConfig.defaults.max = 100
         | .fieldConfig.defaults.thresholds.steps[1].value = 70
         | .fieldConfig.defaults.thresholds.steps[2].value = 90)
  | (.. | objects | select(.id? == 23 and has("targets"))) |=
        (.targets |= [.[0] | .expr = $cpu]
         | .fieldConfig.defaults.unit = "percent" | .fieldConfig.defaults.min = 0 | .fieldConfig.defaults.max = 100)
  | (.. | objects | select(.id? == 22 and has("targets"))) |=
        (.targets[0].expr = "count(count by (cpu) (node_cpu_seconds_total{instance=~\"$instance:.*\"}))")
' > /opt/monitoring/grafana/provisioning/dashboards/files/Servers/proxmox.json
grep -c node_cpu_seconds_total /opt/monitoring/grafana/provisioning/dashboards/files/Servers/proxmox.json

The last line must print 3 or more: the three panels Current CPU, CPU history and the CPU count are patched.

/revisions/4/ fixes the dashboard to the exact version this fix was written for. A newer version could number its panels differently, and the fix would no longer fit.

Step 5: NAS Pools & Health

A dashboard for the pool metrics from Part 2: usage, free space, health, and a "days until 80 % full" estimate.

bash
curl -fsSL -o /opt/monitoring/grafana/provisioning/dashboards/files/Servers/nas-pools.json \
  https://www.zadran.net/downloads/grafana/dashboards/nas-pools.json

The growth estimate needs a few days of data before it shows a sensible value.

Step 6: UniFi-Poller (4 dashboards)

The official Unpoller dashboards, in a version I have tested. In the UAP (access point) dashboard, five panels need a fix:

Panel Problem in the original Fix
Users, Guests Count stays at 0 with current UniFi versions Count real Wi-Fi clients; guests = clients on an SSID with "guest" in its name
Three pie charts Empty with longer time ranges Show the current value
Clock Needs a plugin we don't install Removed
bash
D=/opt/monitoring/grafana/provisioning/dashboards/files/Network
RAW=https://raw.githubusercontent.com/unpoller/dashboards/adbe342f09d0fe06b1cc152d28ab0bae5ee3d8b9/v2.0.0
for pair in "Client%20Insights:unifi-clients" "Network%20Sites:unifi-sites" "UAP%20Insights:unifi-uap" "USW%20Insights:unifi-usw"; do
  curl -fsSL -o "/tmp/${pair#*:}.json" "$RAW/UniFi-Poller_%20${pair%%:*}%20-%20Prometheus.json"
  fix_ds "/tmp/${pair#*:}.json" > "$D/${pair#*:}.json"
done
jq '
    walk(if type == "object" and has("panels") then .panels |= map(select(.id != 32)) else . end)
  | (.. | objects | select(.id? == 48 and has("targets"))) |=
        (.targets[0].expr = "count(unpoller_client_uptime_seconds{site_name=~\"$Site\", ap_name=~\"$AP\", wired=\"false\"}) or vector(0)")
  | (.. | objects | select(.id? == 60 and has("targets"))) |=
        (.targets[0].expr = "count(unpoller_client_uptime_seconds{site_name=~\"$Site\", ap_name=~\"$AP\", wired=\"false\", essid=~\"(?i).*guest.*\"}) or vector(0)")
  | (.. | objects | select((.id? == 52 or .id? == 54 or .id? == 56) and has("targets"))) |=
        (.targets |= map(.instant = true | .range = false))
' "$D/unifi-uap.json" > /tmp/uap.json && mv /tmp/uap.json "$D/unifi-uap.json"
ls -1 $D

The last line lists the four files.

Step 7: Load them

bash
cd /opt/monitoring && docker compose restart grafana
sleep 20
set -a; . /opt/monitoring/.env; set +a
curl -s -u "admin:$GRAFANA_ADMIN_PASSWORD" 'localhost:3000/api/search?type=dash-db' |
  jq -r '.[] | "\(.folderTitle)\t\(.title)"' | sort | column -t -s $'\t'

Expected:

expected output
Network  UniFi-Poller: Client Insights - Prometheus
Network  UniFi-Poller: Network Sites - Prometheus
Network  UniFi-Poller: UAP Insights - Prometheus
Network  UniFi-Poller: USW Insights - Prometheus
Servers  NAS Pools & Health
Servers  Node Exporter Full
Servers  Proxmox via Prometheus

Open Proxmox via Prometheus in Grafana. Current CPU should match the CPU value on the Proxmox web UI's summary page.

Proxmox dashboard with the corrected CPU gauge
Proxmox dashboard with the corrected CPU gauge

Changing a dashboard

Because the files are the master copy, the Save button shows "Cannot save provisioned dashboard". This is on purpose. To change something:

  1. Make the change in the browser.
  2. Click Export → Export as JSON → Copy to clipboard.
  3. Open the matching file, e.g. nano /opt/monitoring/grafana/provisioning/dashboards/files/Servers/proxmox.json, replace its content, and save.
  4. Within 60 seconds Grafana shows the new version.

To get rid of a dashboard, delete its file.

Troubleshooting

Problem Fix
A dashboard shows "Datasource ${DS_PROMETHEUS} was not found" The file was copied without fix_ds. Repeat its step.
Proxmox: Current CPU empty node_exporter isn't running on the host (Part 1, Step 5), or the Proxmox host is selected under a different IP than node_exporter uses.
UniFi dashboards empty Check that the unpoller target is UP (Part 2), then pick your site in the Site drop-down at the top.
A folder doesn't appear It's empty. Grafana only shows folders that contain at least one dashboard file.

In Part 4, Grafana starts sending alerts to your phone.