What is Kubernetes Pod Overhead Accounting?

Question

Accepted Answer

**Overhead Accounting** — track resources used by runtime overhead.

```yaml
# RuntimeClass with overhead
apiVersion: node.k8s.io/v1
kind: RuntimeClass
metadata:
  name: kata
handler: kata
overhead:
  podFixed:
    memory: "160Mi"
    cpu: "250m"
---
# Pod requesting resources
apiVersion: v1
kind: Pod
metadata:
  name: kata-pod
spec:
  runtimeClassName: kata
  containers:
  - name: app
    image: myapp:latest
    resources:
      requests:
        cpu: "500m"
        memory: "256Mi"
      limits:
        cpu: "1000m"
        memory: "512Mi"
    # Total allocated: cpu=750m, memory=416Mi
```

```bash
# Check overhead
kubectl get pod kata-pod -o jsonpath='{.spec.overhead}' | jq

# Check node allocation
kubectl describe node node-1 | grep -A 10 "Allocated resources"

# Verify runtime class overhead
kubectl get runtimeclass kata -o jsonpath='{.overhead}' | jq

# Check actual resource usage vs allocated
kubectl top pod kata-pod --containers
kubectl top pod kata-pod

# Compare with non-overhead pod
kubectl top pod regular-pod --containers
kubectl top pod regular-pod

# Check scheduling decisions
kubectl get events --field-selector reason=Scheduled --sort-by=.lastTimestamp | tail -5
```

**Overhead Sources:**
- VM runtime (Kata, gVisor).
- Sidecar containers.
- System containers.

**Accounting:**
- Added to resource requests.
- Counted against ResourceQuota.
- Affects scheduling decisions.
- Cannot be overridden by containers.

**Best Practices:**
- Account for overhead in capacity planning.
- Use lower overhead runtimes for critical workloads.
- Monitor actual vs allocated resources.

DevOps: What is Kubernetes Pod Overhead Accounting?