| name | managing-k8s-hpa-vpa |
| description | Use when working with K8S Hpa Vpa — kubernetes HPA and VPA autoscaling
management. Covers HorizontalPodAutoscaler configurations, scaling history,
target utilization, current vs desired replicas, VerticalPodAutoscaler
recommendations, update modes, and scaling event analysis. Use when debugging
autoscaling behavior, reviewing scaling policies, analyzing resource
right-sizing, or optimizing autoscaler configurations.
|
| connection_type | k8s |
| preload | false |
Kubernetes HPA and VPA Autoscaling Skill
Manage and analyze HorizontalPodAutoscalers and VerticalPodAutoscalers for workload scaling.
MANDATORY: Discovery-First Pattern
Always list HPAs and VPAs before analyzing scaling behavior.
Phase 1: Discovery
#!/bin/bash
echo "=== HorizontalPodAutoscalers (all namespaces) ==="
kubectl get hpa --all-namespaces -o custom-columns='NAMESPACE:.metadata.namespace,NAME:.metadata.name,REFERENCE:.spec.scaleTargetRef.name,TARGETS:.status.currentMetrics[*].resource.current.averageUtilization,MIN:.spec.minReplicas,MAX:.spec.maxReplicas,CURRENT:.status.currentReplicas' 2>/dev/null | head -20
echo ""
echo "=== VerticalPodAutoscalers ==="
kubectl get vpa --all-namespaces -o custom-columns='NAMESPACE:.metadata.namespace,NAME:.metadata.name,MODE:.spec.updatePolicy.updateMode,TARGET:.spec.targetRef.name' 2>/dev/null | head -20
echo ""
echo "=== HPA Details ==="
kubectl get hpa --all-namespaces -o json 2>/dev/null | jq -r '
.items[] |
"\(.metadata.namespace)/\(.metadata.name)\tRef:\(.spec.scaleTargetRef.kind)/\(.spec.scaleTargetRef.name)\tMin:\(.spec.minReplicas // 1)\tMax:\(.spec.maxReplicas)\tCurrent:\(.status.currentReplicas)\tDesired:\(.status.desiredReplicas)"
' | head -20
Phase 2: Analysis
#!/bin/bash
echo "=== HPA Scaling Status ==="
kubectl get hpa --all-namespaces -o json 2>/dev/null | jq -r '
.items[] |
"\(.metadata.namespace)/\(.metadata.name)\tConditions:" +
([.status.conditions[]? | "\(.type)=\(.status)"] | join(","))
' | head -15
echo ""
echo "=== HPAs at Max Replicas ==="
kubectl get hpa --all-namespaces -o json 2>/dev/null | jq -r '
.items[] |
select(.status.currentReplicas >= .spec.maxReplicas) |
"\(.metadata.namespace)/\(.metadata.name)\tCurrent:\(.status.currentReplicas)\tMax:\(.spec.maxReplicas)\tWARN:AT_MAX"
' | head -10
echo ""
echo "=== HPAs Unable to Scale ==="
kubectl get hpa --all-namespaces -o json 2>/dev/null | jq -r '
.items[] |
select(.status.conditions[]? | select(.type == "ScalingActive" and .status != "True")) |
"\(.metadata.namespace)/\(.metadata.name)\t\(.status.conditions[] | select(.type == "ScalingActive") | .reason): \(.message // "")[0:80]"
' | head -10
echo ""
echo "=== HPA Metric Sources ==="
kubectl get hpa --all-namespaces -o json 2>/dev/null | jq -r '
.items[] |
"\(.metadata.namespace)/\(.metadata.name)\tMetrics:" +
([.spec.metrics[]? |
if .type == "Resource" then "\(.resource.name):\(.resource.target.averageUtilization // .resource.target.averageValue)%"
elif .type == "Pods" then "pods:\(.pods.metric.name)"
elif .type == "Object" then "object:\(.object.metric.name)"
elif .type == "External" then "external:\(.external.metric.name)"
else .type end
] | join(","))
' | head -15
echo ""
echo "=== VPA Recommendations ==="
kubectl get vpa --all-namespaces -o json 2>/dev/null | jq -r '
.items[] |
"\(.metadata.namespace)/\(.metadata.name)\tMode:\(.spec.updatePolicy.updateMode // "Auto")" as $header |
.status.recommendation.containerRecommendations[]? |
"\($header)\tContainer:\(.containerName)\tTarget-CPU:\(.target.cpu)\tTarget-Mem:\(.target.memory)\tUpperBound-CPU:\(.upperBound.cpu)\tUpperBound-Mem:\(.upperBound.memory)"
' | -15
kubectl get hpa --all-namespaces -o json 2>/dev/null | jq -r | -10
kubectl get events --all-namespaces --field-selector reason=SuccessfulRescale -o custom-columns= 2>/dev/null | -10
Output Format
- Target ≤50 lines per output
- Show current/desired/min/max replicas in a single line per HPA
- Display VPA recommendations with target and upper-bound values
- Flag HPAs at max replicas or unable to scale
- Never modify autoscaler settings in analysis -- read-only inspection
Anti-Hallucination Rules
- NEVER assume resource names — always discover via CLI/API in Phase 1 before referencing in Phase 2.
- NEVER fabricate metric names or dimensions — verify against the service documentation or
--help output.
- NEVER mix CLI commands between service versions — confirm which version/API you are targeting.
- ALWAYS use the discovery → verify → analyze chain — every resource referenced must have been discovered first.
- ALWAYS handle empty results gracefully — an empty response is valid data, not an error to retry.
Counter-Rationalizations
| Shortcut | Counter | Why |
|---|
| "I'll skip discovery and check known resources" | Always run Phase 1 discovery first | Resource names change, new resources appear — assumed names cause errors |
| "The user only asked for a quick check" | Follow the full discovery → analysis flow | Quick checks miss critical issues; structured analysis catches silent failures |
| "Default configuration is probably fine" | Audit configuration explicitly | Defaults often leave logging, security, and optimization features disabled |
| "Metrics aren't needed for this" | Always check relevant metrics when available | API/CLI responses show current state; metrics reveal trends and intermittent issues |
| "I don't have access to that" | Try the command and report the actual error | Assumed permission failures prevent useful investigation; actual errors are informative |
Common Pitfalls
- Metrics Server required: HPA needs Metrics Server for CPU/memory metrics -- check metrics-server health
- HPA + VPA conflict: Using HPA and VPA on the same metric (CPU/memory) causes conflicts -- use VPA in recommendation-only mode with HPA
- Stabilization window: Default scale-down stabilization is 300s -- prevents flapping but delays scale-down
- Custom metrics: External/custom metrics require metrics adapters (Prometheus Adapter, Datadog, etc.)
- MinReplicas: Default minReplicas is 1 -- set higher for HA workloads
- Target utilization: HPA targets averageUtilization across all pods -- one hot pod can trigger scaling
- ScaleDown policies: v2 API supports rate-limiting scale-down with
behavior.scaleDown.policies
- VPA update modes: Off (recommendations only), Initial (set on pod creation), Auto (evict and recreate) -- Auto causes pod restarts
- Resource requests required: HPA percentage-based scaling requires resource requests on containers
- Cooldown: After scaling, HPA waits before re-evaluating -- rapid traffic changes may not scale fast enough