Metrics

Every metric starts with cv4pve_. The pages of this section list each one with its type, labels, unit and meaning, grouped by what it describes. This page explains the conventions they share and indexes them all.

Conventions

The `id` label

Guests, storages and nodes are identified by the id label, which is the resource ID used by Proxmox VE itself:

Resource	`id`
Node	`node/pve01`
VM	`qemu/100`
Container	`lxc/101`
Storage	`storage/pve01/local-zfs`, one per node, also for a shared storage
Cluster (only in `cv4pve_up`)	`cluster/pve-cluster`

Per-node metrics use a node label with the node name instead.

Info metrics

Names, tags, versions and other text are not repeated on every series: they are labels of an …_info metric whose value is always 1: cv4pve_guest_info, cv4pve_node_info, cv4pve_storage_info, cv4pve_node_version_info, cv4pve_node_subscription_info, cv4pve_cluster_info. Join them in a query to add the label you need:

# Memory used by each guest, with its name and node
cv4pve_guest_memory_usage_bytes
  * on (id) group_left (name, node) cv4pve_guest_info

State metrics

A state with a fixed set of values (a guest lock, an HA state, a subscription status) is exported as one series per possible value, with a state (or status) label, and value 1 for the current state and 0 for the others. The alert is then a plain comparison, and a guest that stays locked keeps the same series:

cv4pve_guest_lock{state="backup"} == 1

Units and types

Sizes are in bytes, durations in seconds, points in time are Unix timestamps in seconds.
…_ratio metrics go from 0 to 1.
Metrics ending in _total are counters: use them with rate() or increase(). A counter goes back to zero when Proxmox VE resets it (a guest restarted or migrated) or when the exporter restarts; Prometheus handles both.
Everything else is a gauge, except cv4pve_api_request_duration_seconds, a histogram.

Always on, or turned on by a setting

The metrics built from the two cluster-wide calls, /cluster/status and /cluster/resources, are always exported. The others come from a collector, turned on and cached by a setting. A collector that is off exports no series at all.

All metrics

Metric	Type	Setting	Page
`cv4pve_up`	gauge	always	Cluster and nodes
`cv4pve_cluster_info`	gauge	always	Cluster and nodes
`cv4pve_cluster_quorate`	gauge	always	Cluster and nodes
`cv4pve_cluster_nodes`	gauge	always	Cluster and nodes
`cv4pve_node_info`	gauge	always	Cluster and nodes
`cv4pve_node_uptime_seconds`, `cv4pve_node_load_avg1`/`5`/`15`, `cv4pve_node_memory_used_bytes`, `cv4pve_node_memory_total_bytes`, `cv4pve_node_swap_used_bytes`, `cv4pve_node_swap_total_bytes`, `cv4pve_node_root_fs_used_bytes`, `cv4pve_node_root_fs_total_bytes`	gauge	`Node.Status`	Cluster and nodes
`cv4pve_node_cpu_assigned_cores`, `cv4pve_node_memory_assigned_bytes`	gauge	`Node.Status`	Cluster and nodes
`cv4pve_node_version_info`	gauge	`Node.Status`	Cluster and nodes
`cv4pve_node_subscription_info`, `cv4pve_node_subscription_status`, `cv4pve_node_subscription_next_due_timestamp_seconds`	gauge	`Node.Subscription`	Cluster and nodes
`cv4pve_node_disk_health`, `cv4pve_node_disk_wearout`	gauge	`Node.DiskSmart`	Cluster and nodes
`cv4pve_guest_info`, `cv4pve_guest_lock`	gauge	always	Guests
`cv4pve_guest_cpu_usage_ratio`, `cv4pve_guest_cpu_cores`, `cv4pve_guest_memory_size_bytes`, `cv4pve_guest_memory_usage_bytes`, `cv4pve_guest_memory_host_ratio`, `cv4pve_guest_disk_size_bytes`, `cv4pve_guest_disk_usage_bytes`, `cv4pve_guest_uptime_seconds`	gauge	always	Guests
`cv4pve_guest_disk_read_bytes_total`, `cv4pve_guest_disk_write_bytes_total`, `cv4pve_guest_network_receive_bytes_total`, `cv4pve_guest_network_transmit_bytes_total`	counter	always	Guests
`cv4pve_guest_balloon_actual_bytes`	gauge	`Guest.Balloon`	Guests
`cv4pve_storage_info`, `cv4pve_storage_shared`, `cv4pve_storage_size_bytes`, `cv4pve_storage_usage_bytes`	gauge	always	Storage
`cv4pve_ha_state`, `cv4pve_ha_node_state`, `cv4pve_ha_quorate`	gauge	`Cluster.Ha`	HA
`cv4pve_replication_duration_seconds`, `cv4pve_replication_last_sync_timestamp_seconds`, `cv4pve_replication_next_sync_timestamp_seconds`, `cv4pve_replication_fail_count`	gauge	`Node.Replication`	Replication and backup
`cv4pve_not_backed_up_info`, `cv4pve_guests_not_backed_up`	gauge	`Cluster.BackupInfo`	Replication and backup
`cv4pve_scrape_duration_seconds`, `cv4pve_scrape_last_success_timestamp_seconds`	gauge	always	Exporter
`cv4pve_scrape_errors_total`	counter	always	Exporter
`cv4pve_api_request_duration_seconds`	histogram	`ApiInstrumentation`	Exporter
`cv4pve_api_request_errors_total`	counter	`ApiInstrumentation`	Exporter