Proxmox Observability
OpenTelemetry-native metrics for your Proxmox VE estate, arranged as a datacenter → node → guest hierarchy with contextual interlinking, PSI-first pressure visibility, and host-vs-guest correlation. Full cardinality across qemu and LXC, no per-node tax, no proprietary agent — scrape the Proxmox VE Exporter straight into your Collector.
Seven Dashboards - One Datacenter Model
Click any card to enlarge · esc to close
01 · Datacenter overview
Inventory tiles for nodes, guests, and storage above a Needs Attention list of what's actually wrong — stopped guests, PSI-bound VMs, pools past threshold — each with the exact metric that surfaced it. No opaque health score.
02 · Drill-down in three clicks
Click a hot node, land on its deep dive. Click a contended guest, land on it with the host and neighbor VMs already loaded. Time range and entity context carry across every jump — no PromQL, no refilter.
03 · Node deep dive
Per-node saturation, capacity-vs-allocation, and a guest contention table sorted by PSI — so the VMs actually driving the host rise to the top, right next to the host metrics themselves.
04 · Guest + host context
Every guest metric shown side-by-side with the same metric on its host and on the neighbour guests sharing the node — the virtualization question answered without switching dashboards.
05 · PSI-first pressure
CPU, memory, IO, and node load ranked by PSI across the whole fleet. 80% CPU is fine until workloads start waiting — PSI is where that waiting actually shows up.
06 · Storage + growth
Per-pool used % and free GiB tiles for every pool on every node, utilisation over time, growth per week, and a projection of days-to-90% at current rate — interlinked to the guests driving the IO.
07 · Network top talkers
Per-node NIC throughput and per-guest RX/TX ranked. Enter from a guest or a node — the top talkers on that surface load with the same time window, so a noisy neighbour is one click away.
08 · Availability & SLO
Fleet availability, running-vs-stopped counts, worst-guest SLO surfaced first, and per-guest uptime targets — the specialist view for SLO owners, kept out of the default operational dashboard.
Scrape the Proxmox VE Exporter and your PVE node exporters into an OpenTelemetry Collector that writes to your own S3, GCS, or Azure Blob bucket — no vendor storage in the path. Seven interlinked dashboards, per-guest SLOs, qemu- and LXC-level depth, and cluster rollups are on the moment the first object lands.
Step 1
Spin up a Cardinal workspace in under a minute. Free to start, no credit card.
Create account / Log inStep 2
Point your OpenTelemetry Collector at the Proxmox VE Exporter and ship metrics to your own object store. The docs walk you through the config.
Follow the guide