monitoring

KubeVirtVMGuestMemoryAvailableLow

Meaning

The VM guest OS has very low available memory (<3% headroom) for an extended period with no meaningful swap I/O, approaching OOM conditions.

Impact

• Very limited memory headroom (risk of imminent OOM)
• Potential application failures and degraded performance

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) < 3% sustained:

  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) Swap traffic ~0 over 30m (likely no swap):

  sum by (name, namespace) (
    rate(
      kubevirt_vmi_memory_swap_in_traffic_bytes{
        name="<vmi_pod>", namespace="<namespace>"
      }[30m]
    ) +
    rate(
      kubevirt_vmi_memory_swap_out_traffic_bytes{
        name="<vmi_pod>", namespace="<namespace>"
      }[30m]
    )
  )
  

  4) Major page faults ~0 over 30m:

  sum by (name, namespace) (
    rate(
      kubevirt_vmi_memory_pgmajfault_total{
        name="<vmi_pod>", namespace="<namespace>"
      }[30m]
    )
  )
  

Mitigation

• Immediate actions (in guest)
  • If possible, stop or restart memory-hungry processes to reduce load.
  • If acceptable, drop caches temporarily.
  • Configure/enable swap per policy (or validate it is intentionally disabled).
• 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.