// Benutzung · Grundlagen
Reading the dashboard
The Dashboard is the single cross-location view of everything Fleeger
monitors. It loads once, then updates live — a small pulsing dot and a
live label appear top-right when the SSE channel is connected, next to the
last-updated time and a manual Refresh button. Data comes from
VictoriaMetrics (metrics) and ingestion-svc (reachability), proxied through
gateway-svc.
Panels
Locations
One card per location with a reachable / unreachable badge, the resource count, and when it was "last seen". Reachability is derived from metric freshness — see checking reachability. If you see "No metrics received yet", no spoke-agent has reported.
Hub services (meta)
The control plane monitoring itself: each backend service with its version,
uptime, memory, and total request count (scraped from each service's /metrics,
labelled location=meta). See Observability & self-monitoring.
Proxmox nodes
Per node: online/offline, CPU usage (and core count), memory used / total, and uptime.
VMs & containers
Every guest discovered under the Proxmox clusters: name, running/stopped, and — when running — CPU and memory. Stopped guests are dimmed. These rows are auto-discovered (you don't enter them by hand).
Kubernetes
Per cluster: a node table (ready / not-ready, CPU and memory capacity, pod capacity) and a one-line pod summary — running/total plus a count per namespace.
iDRAC / hardware
A live, top-down interactive server view rendered from the real iDRAC sensors:
- Fans spin at their actual RPM — hover for the exact value.
- CPUs are heat-coloured (warm glow as they get hotter) — hover for °C.
- Airflow ripples inlet → exhaust; hover the front/rear edges for inlet / exhaust temps; the exhaust runs warmer-coloured as it heats up.
- PSUs — hover for input voltage and the shared system draw.
Alongside it: a power-draw number with a 60-minute trend chart, and a list of every temperature sensor the hardware reports (no fixed schema — whatever iDRAC exposes shows up).
Tips
- A panel is hidden when there's no data for it (e.g. no Kubernetes panel until a
k8s_clusteris being collected). - "updated HH:MM:SS" shows the last successful refresh; Refresh forces one.
- A name shown more than once, such as two clusters' VM 100 or two locations'
node
pve1, is followed by what tells the rows apart: the location's name and the cluster, e.g.pve1 (homelab · alpha). The charts and the Explorer's panels name their series the same way. A location whose name cannot be loaded is shown by its id, and two locations with the same name get the start of their ids, e.g.pve1 (lab (loc-aaaa) · alpha). - If a panel looks stale, first check the location is reachable — a stopped agent means no fresh metrics.