fix: per-metric z-score thresholds, min stddev guard, nodeSelector

- Disk/NAS: z-score threshold 2 (was 5, too insensitive for storage)
- CPU/Memory/Ingress/Istio: z-score threshold 5 (unchanged)
- Add MIN_STDDEV=0.5 guard: skip peak detection when data has
  near-zero variance (prevents false peaks on stable disk data)
- Add nodeSelector: nodegroup=nd to deployment
- Apply thresholds in both /api/report and /api/metrics/[type]

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

charts/infra-report/templates/deployment.yaml

@@ -20,6 +20,10 @@ spec:
       imagePullSecrets:
         {{- toYaml . | nindent 8 }}
       {{- end }}
+      {{- with .Values.nodeSelector }}
+      nodeSelector:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
       containers:
         - name: {{ .Chart.Name }}
           image: "{{ .Values.image.repository }}:{{ .Values.image.tag | default .Chart.AppVersion }}"

charts/infra-report/values.yaml

@@ -40,6 +40,9 @@ config:
   QUERY_ISTIO_RX: 'sum(rate(container_network_receive_bytes_total{namespace="istio-system", pod=~"istio-ingressgateway.*"}[5m]))'
   QUERY_ISTIO_TX: 'sum(rate(container_network_transmit_bytes_total{namespace="istio-system", pod=~"istio-ingressgateway.*"}[5m]))'
 
+nodeSelector:
+  nodegroup: nd
+
 resources:
   requests:
     cpu: "100m"

src/app/api/metrics/[type]/route.ts

@@ -14,6 +14,15 @@ const NETWORK_MAP: Record<string, { rx: string; tx: string }> = {
   istio: { rx: QUERIES.istioReceive, tx: QUERIES.istioTransmit },
 };
 
+const PEAK_THRESHOLD: Record<string, number> = {
+  cpu: 5,
+  memory: 5,
+  disk: 2,
+  nas: 2,
+  ingress: 5,
+  istio: 5,
+};
+
 function parsePrometheusValues(values: [number, string][]): DataPoint[] {
   return values.map(([ts, val]) => ({ timestamp: ts, value: parseFloat(val) }));
 }
@@ -38,13 +47,14 @@ export async function GET(
         queryRange(netQuery.tx, start, end, step),
       ]);
 
+      const threshold = PEAK_THRESHOLD[type] || 5;
       const process = (results: typeof rxResults) => results.map((r) => {
         const dataPoints = parsePrometheusValues(r.values);
         return {
           instance: r.metric.node || r.metric.instance || '',
           dataPoints,
           stats: calculateStats(dataPoints),
-          peaks: detectPeaks(dataPoints),
+          peaks: detectPeaks(dataPoints, threshold),
         };
       });
 
@@ -57,6 +67,7 @@ export async function GET(
     }
 
     const results = await queryRange(query, start, end, step);
+    const threshold = PEAK_THRESHOLD[type] || 5;
     const instances = results.map((r) => {
       const dataPoints = parsePrometheusValues(r.values);
       return {
@@ -64,7 +75,7 @@ export async function GET(
         metric: r.metric,
         dataPoints,
         stats: calculateStats(dataPoints),
-        peaks: detectPeaks(dataPoints),
+        peaks: detectPeaks(dataPoints, threshold),
       };
     });
 

src/app/api/report/route.ts

@@ -7,7 +7,7 @@ function parseValues(values: [number, string][]): DataPoint[] {
   return values.map(([ts, val]) => ({ timestamp: ts, value: parseFloat(val) }));
 }
 
-function processResults(results: { metric: Record<string, string>; values: [number, string][] }[], labelKey?: string) {
+function processResults(results: { metric: Record<string, string>; values: [number, string][] }[], labelKey?: string, peakThreshold?: number) {
   return results.map((r) => {
     const dataPoints = parseValues(r.values);
     return {
@@ -15,7 +15,7 @@ function processResults(results: { metric: Record<string, string>; values: [numb
       metric: r.metric,
       dataPoints,
       stats: calculateStats(dataPoints),
-      peaks: detectPeaks(dataPoints),
+      peaks: detectPeaks(dataPoints, peakThreshold),
     };
   });
 }
@@ -60,10 +60,10 @@ export async function GET(request: NextRequest) {
       period: { start, end, step },
       nodes,
       metrics: {
-        cpu: processResults(cpuData),
-        memory: processResults(memData),
-        disk: processResults(diskData),
-        nas: processResults(nasData, 'mountpoint'),
+        cpu: processResults(cpuData, undefined, 5),
+        memory: processResults(memData, undefined, 5),
+        disk: processResults(diskData, undefined, 2),
+        nas: processResults(nasData, 'mountpoint', 2),
         ingress: {
           receive: processResults(ingressRxData),
           transmit: processResults(ingressTxData),

src/lib/analytics.ts

@@ -31,6 +31,10 @@ export function calculateStats(dataPoints: DataPoint[]): Stats {
   };
 }
 
+// minStddev: stddev가 이 값 미만이면 데이터 변동이 무의미하므로 peak 감지 스킵
+// Disk/NAS 같은 안정적 메트릭에서 미세 변동이 peak으로 잡히는 문제 방지
+const MIN_STDDEV = 0.5;
+
 export function detectPeaks(
   dataPoints: DataPoint[],
   threshold: number = 5
@@ -41,8 +45,11 @@ export function detectPeaks(
     values.reduce((sum, v) => sum + Math.pow(v - mean, 2), 0) / values.length
   );
 
+  // stddev가 너무 작으면 데이터가 거의 일정한 것이므로 peak 없음
+  const effectiveStddev = stddev < MIN_STDDEV ? 0 : stddev;
+
   const points = dataPoints.map((d) => {
-    const zScore = stddev === 0 ? 0 : (d.value - mean) / stddev;
+    const zScore = effectiveStddev === 0 ? 0 : (d.value - mean) / effectiveStddev;
     return { ...d, zScore, isPeak: zScore > threshold };
   });