PERF(resources): remove CPU limits - keep memory limits only

- CPU throttling prevents app startup, not crashes
- Memory OOM is the real cascading failure cause
- CPU request ensures fair scheduling

cert-manager/helm-values.yaml

@@ -8,22 +8,28 @@ replicaCount: 1
 
 resources:
   requests:
-    cpu: 3m  # Reduced to 30% of original (10m -> 3m)
+    cpu: 15m
-    memory: 32Mi
+    memory: 64Mi
+  limits:
+    memory: 96Mi
 
 webhook:
   replicaCount: 1
   resources:
     requests:
-      cpu: 3m  # Reduced to 30% of original (10m -> 3m)
+      cpu: 15m
-      memory: 32Mi
+      memory: 64Mi
+    limits:
+      memory: 96Mi
 
 cainjector:
   replicaCount: 1
   resources:
     requests:
-      cpu: 3m  # Reduced to 30% of original (10m -> 3m)
+      cpu: 15m
-      memory: 32Mi
+      memory: 64Mi
+    limits:
+      memory: 96Mi
 
 # Prometheus metrics
 prometheus:

traefik/helm-values.yaml

@@ -9,6 +9,14 @@ spec:
     deployment:
       replicas: 3
 
+    # Resource limits
+    resources:
+      requests:
+        cpu: 50m
+        memory: 128Mi
+      limits:
+        memory: 192Mi
+
     # Pod Anti-Affinity - 가능하면 각 노드에 분산 배치 (soft)
     affinity:
       podAntiAffinity: