Proxmox monitoring: the tools compared (Zabbix, Checkmk, Grafana, Pulse)

Proxmox VE ships live graphs in its web interface - but no monitoring in the operational sense: nobody gets woken up when a node fails at night, a ZFS pool is degraded or the backup job has been failing for days. That requires an external system. Four approaches have become established: Zabbix with its official template, Checkmk with its special agent, the Prometheus exporter with Grafana, and the lightweight Pulse. This comparison shows what the tools really cover - and where each one has gaps (as of July 2026).

What Proxmox monitoring must cover

Before choosing a tool, look at the checklist. Complete Proxmox monitoring includes:

• Host level: CPU, RAM, swap, disk I/O, network, temperatures and SMART status of the disks
• Virtualization level: status and resources of all VMs and containers, automatic discovery of new guests
• Cluster level: quorum, node reachability, corosync health
• Storage: usage levels, ZFS pool status or Ceph health
• Backups: are the PBS jobs running? How old is the last successful backup?
• Alerting: escalation chains, on-call, maintenance windows - otherwise it is a dashboard, not monitoring

No tool covers all six points alone. The differences lie in how much is missing and how much custom work is needed.

Zabbix: the all-rounder with an official template

Zabbix ships "Proxmox VE by HTTP", an official, vendor-maintained template (available for Zabbix 6.0 through 7.4, Zabbix integration page). It works agentlessly via the PVE REST API (port 8006) with an API token that only needs read permissions (Sys.Audit, Datastore.Audit, VM.Audit - in practice: the PVEAuditor role). Via low-level discovery it automatically detects the cluster including quorum, all nodes (status, CPU, RAM, uptime, version), storages and every VM and LXC container - including triggers for node failure and overload. Since Zabbix 7.4, disk discovery with SMART health via the PVE API, certificate and update monitoring have been added.

The gaps: the template does not cover ZFS pool health or PBS backup jobs. For those you install Zabbix agent 2 on the host (official SMART plugin with smartmontools; proven community templates exist for ZFS) or build custom checks. Zabbix thus offers the best ratio of coverage, alerting depth and operability - provided someone maintains it.

Checkmk: strong on backups

Checkmk provides a dedicated Proxmox VE special agent that connects with a checkmk@pve user and the PVEAuditor role (Checkmk docs). VM and container data reaches the right hosts via the piggyback mechanism. The big plus are the ready-made backup checks: proxmox_ve_vm_backup_status monitors date, age, size and bandwidth of the last successful backup per guest, proxmox_ve_vm_snapshot_age alerts on stale snapshots - exactly the discipline that means custom work elsewhere.

Limitations: automatic host creation for discovered VMs (dynamic host management) is reserved for the commercial editions; in the free Raw Edition that means manual work. The docs also describe username/password login rather than API tokens.

Grafana + Prometheus: the dashboard champions

The prometheus-pve-exporter (actively maintained, version 3.9) exports cluster, node, VM/LXC and storage metrics including backup coverage on port 9221 - again, read-only access (PVEAuditor) is enough. Together with Prometheus and Grafana you get the most flexible visualization, including alerting via Alertmanager.

Alternatively, Proxmox pushes metrics natively to external systems: Graphite, InfluxDB and, since PVE 9.1, OpenTelemetry, configured under Datacenter → Metric Server (Proxmox external metric server documentation). Important context: this is pure metric export - state logic, triggers and escalation must come from the stack behind it. Our Grafana expertise shows how a dashboard stack is built.

Pulse: the fastest overview

Pulse is a lightweight dashboard with real-time views and alerting (Discord, Slack, Telegram, email, webhooks) for Proxmox VE, PBS and PMG - now also for Docker and Kubernetes. Installation via LXC script or Docker takes minutes; the project has around 6,000 GitHub stars. For homelabs and a quick team overview it is unbeatably pragmatic. Relevant for business use: the project now follows an open-core model - the free version retains metrics for 7 days; RBAC, audit logs and 90-day retention are reserved for the paid Pro tier.

The direct comparison

How WZ-IT does it

We run a highly available Zabbix cluster as the monitoring backbone for customers: the official Proxmox template as the base, extended with agent-2 checks for SMART and ZFS, custom checks for PBS backup runs, and alerting with escalation chains up to on-call. We add Grafana for dashboards - the strengths of both worlds. If a monitoring node fails, monitoring continues uninterrupted; the monitoring itself is not a single point of failure.

If you do not want to build and maintain this yourself: Managed Proxmox includes exactly this monitoring (from €179.90 per node per month), and Managed Zabbix is available as a standalone building block without full operations - entry from €79.90 per month. The step-by-step guide for a self-built setup is Monitor Proxmox with Zabbix. We are happy to discuss your setup in a free initial consultation.