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

alertmanager/helm-values.yaml

@@ -10,8 +10,10 @@ persistence:
 
 resources:
   requests:
-    cpu: 10m
+    cpu: 15m
-    memory: 32Mi
+    memory: 100Mi
+  limits:
+    memory: 150Mi
 
 # Disable default config - use secret instead
 config:

goldilocks/helm-values.yaml

@@ -7,12 +7,11 @@ dashboard:
   replicaCount: 1
 
   resources:
-    limits:
-      cpu: null
-      memory: 256Mi
     requests:
-      cpu: 25m
+      cpu: 15m
-      memory: 128Mi
+      memory: 100Mi
+    limits:
+      memory: 150Mi
 
   service:
     type: ClusterIP
@@ -39,12 +38,11 @@ controller:
   enabled: true
 
   resources:
-    limits:
-      cpu: null
-      memory: 256Mi
     requests:
-      cpu: 25m
+      cpu: 15m
-      memory: 128Mi
+      memory: 100Mi
+    limits:
+      memory: 150Mi
 
   # Enable VPA recommendations for all namespaces
   # Set to false to only monitor namespaces with the label: goldilocks.fairwinds.com/enabled=true

grafana/helm-values.yaml

@@ -24,8 +24,10 @@ podSecurityContext:
 
 resources:
   requests:
-    cpu: 25m
+    cpu: 11m
-    memory: 128Mi
+    memory: 425Mi
+  limits:
+    memory: 425Mi
 
 service:
   type: ClusterIP

kube-state-metrics/helm-values.yaml

@@ -5,8 +5,10 @@ fullnameOverride: kube-state-metrics
 
 resources:
   requests:
-    cpu: 10m
+    cpu: 15m
-    memory: 64Mi
+    memory: 100Mi
+  limits:
+    memory: 150Mi
 
 service:
   type: ClusterIP

loki/helm-values.yaml

@@ -43,8 +43,10 @@ singleBinary:
     storageClass: local-path
   resources:
     requests:
-      cpu: 15m  # Reduced from 50m based on actual usage (10m)
+      cpu: 23m
-      memory: 128Mi
+      memory: 462Mi
+    limits:
+      memory: 462Mi
 
 # Disable components not needed in single binary mode
 backend:

node-exporter/helm-values.yaml

@@ -8,8 +8,10 @@ hostPID: true
 
 resources:
   requests:
-    cpu: 10m
+    cpu: 15m
-    memory: 50Mi
+    memory: 64Mi
+  limits:
+    memory: 96Mi
 
 service:
   type: ClusterIP

prometheus/helm-values.yaml

@@ -53,11 +53,11 @@ prometheus:
               storage: 5Gi
 
     resources:
-      limits:
-        memory: 1Gi           # limit 추가 (무제한 증가 방지)
       requests:
         cpu: 200m
-        memory: 512Mi         # 256Mi → 512Mi (실제 사용량 반영)
+        memory: 512Mi
+      limits:
+        memory: 768Mi
     
     # ServiceMonitor 자동 발견 - 모든 ServiceMonitor 선택
     serviceMonitorSelectorNilUsesHelmValues: false

promtail/helm-values.yaml

@@ -32,8 +32,10 @@ defaultVolumeMounts:
 # Resources
 resources:
   requests:
-    cpu: 25m  # Reduced from 50m based on actual usage (8-17m)
+    cpu: 23m
-    memory: 64Mi
+    memory: 182Mi
+  limits:
+    memory: 182Mi
 
 # Tolerations to run on all nodes including master
 tolerations:

uptime-kuma/helm-values.yaml

@@ -13,10 +13,10 @@ persistence:
 
 resources:
   requests:
-    cpu: 50m
+    cpu: 15m
-    memory: 128Mi
+    memory: 200Mi
   limits:
-    memory: 256Mi
+    memory: 300Mi
 
 livenessProbe:
   enabled: true

vpa/helm-values.yaml

@@ -7,12 +7,11 @@ recommender:
   replicaCount: 1
 
   resources:
-    limits:
-      cpu: null
-      memory: 1Gi
     requests:
-      cpu: 50m
+      cpu: 15m
-      memory: 512Mi
+      memory: 128Mi
+    limits:
+      memory: 192Mi
 
 # Updater - applies recommended resource requests to pods
 # Disabled because we're using updateMode: Off (recommendations only)