monitoring

KubeVirtVMGuestMemoryPressure

Meaning

The VM guest OS is under sustained memory pressure (low usable memory with elevated major page faults and/or swap I/O), risking thrashing, swapping, or OOM kills.

Impact

• Performance degradation due to page thrashing
• Increased swap I/O and latency
• Risk of OOM kills and application instability

Diagnosis

• Metrics (Prometheus) 1) Identify the VMI pod (vmi_pod label):

  kubevirt_vmi_info{vm="<vm-name>", namespace="<namespace>",
  phase="running"}
  

  2) Headroom (usable/available) < 5% indicates pressure:

  sum by (name, namespace) (
    kubevirt_vmi_memory_usable_bytes{
      name="<vmi_pod>", namespace="<namespace>"
    }
  ) /
  sum by (name, namespace) (
    kubevirt_vmi_memory_available_bytes{
      name="<vmi_pod>", namespace="<namespace>"
    }
  )
  

  3) Major page faults are elevated:

  sum by (name, namespace) (
    rate(
      kubevirt_vmi_memory_pgmajfault_total{
        name="<vmi_pod>", namespace="<namespace>"
      }[5m]
    )
  )
  

  4) Swap traffic is elevated (bytes/s):

  sum by (name, namespace) (
    rate(
      kubevirt_vmi_memory_swap_in_traffic_bytes{
        name="<vmi_pod>", namespace="<namespace>"
      }[5m]
    ) +
    rate(
      kubevirt_vmi_memory_swap_out_traffic_bytes{
        name="<vmi_pod>", namespace="<namespace>"
      }[5m]
    )
  )
  

Mitigation

• Short term (in guest)
  • If possible, Restart or tune memory-heavy processes; reduce workload.
  • If acceptable, drop caches temporarily.
  • Ensure swap is sized/policy-compliant (or disabled if not desired).
• Increase VM memory (hotplug if supported; otherwise restart may be required)

If you cannot resolve the issue, see the following resources:

• OKD Help
• #virtualization Slack channel

This site is open source. Improve this page.