PCD Metrics Reference
Reference list of host and VM metrics available from the node-exporter and libvirt-exporter on each PCD hypervisor host.
Each Private Cloud Director hypervisor host exposes metrics through two Victoria metrics compatible exporters:
node-exporter: host-level infrastructure metrics, available at
<host-ip>:9388/metrics.libvirt-exporter: VM-level metrics for each virtual machine running on the host, available at
<host-ip>:9177/metrics.
The Victoria metrics agent on each host applies recording rules to these raw metrics and remotely writes the processed output to Victoria metrics on PCD every 5 minutes. Labels identifying the originating host and VM are added during remote write.
For information on how to consume these metrics in your own observability stack, see Exporting PCD Metrics
Hypervisor metrics
These metrics reflect the physical host and are identified by a host label containing the host UUID.
pcd:hyp_cpu_total{host="<host-uuid>", job="ne"}
Total CPU capacity of the hypervisor.
pcd:hyp_cpu_usage{host="<host-uuid>", job="ne"}
CPU usage of the hypervisor, expressed as number of cores consumed.
pcd:hyp_mem_total{host="<host-uuid>", job="ne"}
Total memory capacity of the hypervisor, in bytes.
pcd:hyp_mem_usage{host="<host-uuid>", job="ne"}
Memory currently in use on the hypervisor, in bytes.
pcd:hyp_disk_space{fstype="ext4", host="<host-uuid>"}
Total storage capacity of the hypervisor by filesystem type, in bytes.
pcd:hyp_disk_usage{fstype="ext4", host="<host-uuid>"}
Storage used on the hypervisor by filesystem type, in bytes.
pcd:hyp_disk_read_throughput{device="<device>", host="<host-uuid>"}
Disk read throughput by device, in bytes per second.
pcd:hyp_disk_write_throughput{device="<device>", host="<host-uuid>"}
Disk write throughput by device, in bytes per second.
pcd:hyp_net_rx_throughput{host="<host-uuid>"}
Rate of bytes received over the network for the entire host. Not broken down by interface.
pcd:hyp_net_tx_throughput{host="<host-uuid>"}
Rate of bytes transmitted over the network for the entire host. Not broken down by interface.
pcd:hyp_total{host="<host-uuid>"}
Total count of hypervisors.
Virtual machine metrics
These metrics reflect individual virtual machines and are identified by a domain label containing the VM UUID and a host label containing the host's UUID that the VM runs on.
pcd:vm_cpu_total{domain="<vm-uuid>", host="<host-uuid>"}
Total CPU capacity allocated to the VM.
pcd:vm_cpu_usage{domain="<vm-uuid>", host="<host-uuid>"}
CPU usage of the VM.
pcd:vm_cpu_throttling_percentage{domain="<vm-uuid>", host="<host-uuid>", vcpu="0"}
CPU throttling percentage for the VM, broken down by vCPU core.
pcd:vm_mem_total{domain="<vm-uuid>", host="<host-uuid>"}
Total memory allocated to the VM, in bytes.
pcd:vm_mem_allocated{domain="<vm-uuid>", host="<host-uuid>"}
Memory available inside the VM. Slightly less than pcd:vm_mem_total due to overhead.
pcd:vm_mem_usage{domain="<vm-uuid>", host="<host-uuid>"}
Memory currently in use by the VM, in bytes.
pcd:vm_mem_usage_percent{host="<host-uuid>"}
Memory usage of the VM as a percentage of allocated memory.
pcd:vm_total_storage{domain="<vm-uuid>", host="<host-uuid>"}
Total storage capacity across all disks for the VM, in bytes.
pcd:vm_allocated_storage{domain="<vm-uuid>", host="<host-uuid>"}
Storage allocated to the VM, in bytes.
pcd:vm_used_storage{domain="<vm-uuid>", host="<host-uuid>"}
Storage currently used by the VM, in bytes.
pcd:vm_read_iops{domain="<vm-uuid>", host="<host-uuid>", target_device="<device>"}
Read IOPS for the VM, per disk device.
pcd:vm_write_iops{domain="<vm-uuid>", host="<host-uuid>", target_device="<device>"}
Write IOPS for the VM, per disk device.
pcd:vm_read_throughput{domain="<vm-uuid>", host="<host-uuid>", target_device="<device>"}
Read throughput for the VM, per disk device, in bytes per second.
pcd:vm_write_throughput{domain="<vm-uuid>", host="<host-uuid>", target_device="<device>"}
Write throughput for the VM, per disk device, in bytes per second.
pcd:vm_read_latency{domain="<vm-uuid>", host="<host-uuid>", target_device="<device>"}
Read latency for the VM, per disk device.
pcd:vm_write_latency{domain="<vm-uuid>", host="<host-uuid>", target_device="<device>"}
Write latency for the VM, per disk device.
pcd:vm_rx_bytes{domain="<vm-uuid>", host="<host-uuid>"}
Rate of bytes received over the network for the VM.
pcd:vm_tx_bytes{domain="<vm-uuid>", host="<host-uuid>"}
Rate of bytes transmitted over the network for the VM.
pcd:vm_rx_packets_drop{domain="<vm-uuid>", host="<host-uuid>"}
Rate of inbound packets dropped for the VM.
pcd:vm_tx_packets_drop{domain="<vm-uuid>", host="<host-uuid>"}
Rate of outbound packets dropped for the VM.
pcd:vm_total{host="<host-uuid>"}
Total number of VMs running on the host.
High availability (HA) event metrics
These metrics record VM HA evacuation events on virtualized clusters that have VM HA enabled. They are event-driven and sparse: they emit series only while or after a VM HA event occurs and produce no series during healthy operation. The built-in VMHA Dashboard in Grafana is powered by these metrics.
Unlike the pcd: metrics above, these are not exposed by node-exporter or libvirt-exporter on each hypervisor host. They are emitted by the VM HA service on the Private Cloud Director controller.
vmha_events_generated{phase="<phase>", cluster="<cluster>", host_id="<host-uuid>", vmha_uuid="<event-uuid>"}
One series per VM HA evacuation event. The phase label indicates the event lifecycle stage (for example, completed).
vmha_task_details{phase="<phase>", cluster="<cluster>", vmha_uuid="<event-uuid>", vm_name="<vm-name>"}
One series per VM evacuated within an event. The phase label progresses through queued, inprogress, and completed.
Because these metrics are event-driven, alerting rules built on them must set No Data handling to Normal/OK. Treating "no data" as firing would produce false alerts during healthy operation when no HA events are occurring.
In the metric signatures above, <host-uuid> and <vm-uuid> are the UUID values assigned to each host and VM in your environment, matching the Host ID and VM UUID shown in the Private Cloud Director UI. For HA event metrics, <event-uuid> is the unique identifier for each evacuation event, <vm-name> is the name of an evacuated VM, and <cluster> is the cluster name.
Last updated
Was this helpful?