Sync CNV runbook KubeVirtVMGuestMemoryAvailableLow.md (Updated at 2025-09-15 08:41:07 +0000 UTC)

alerts/openshift-virtualization-operator/KubeVirtVMGuestMemoryAvailableLow.md

@@ -0,0 +1,67 @@
+# 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):
+  ```promql
+  kubevirt_vmi_info{vm="<vm-name>", namespace="<namespace>",
+  phase="running"}
+  ```
+  2) Headroom (usable/available) < 3% sustained:
+  ```promql
+  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):
+  ```promql
+  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:
+  ```promql
+  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, log in to the
+[Customer Portal](https://access.redhat.com) and open a support case,
+attaching the artifacts gathered during the diagnosis procedure.