fix(observability): tune cpu and mem quotas

jfroy · Aug 22, 2024 · aaab8bd · aaab8bd
1 parent 3b0b245
commit aaab8bd
Show file tree

Hide file tree

Showing 11 changed files with 26 additions and 37 deletions.
diff --git a/kubernetes/apps/observability/gatus/app/helmrelease.yaml b/kubernetes/apps/observability/gatus/app/helmrelease.yaml
@@ -60,11 +60,6 @@ spec:
               UNIQUE_FILENAMES: true
               METHOD: WATCH
             securityContext: *securityContext
-            resources: &resources
-              requests:
-                cpu: 10m
-              limits:
-                memory: 128Mi
             restartPolicy: Always
         containers:
           app:
@@ -93,8 +88,12 @@ spec:
                   timeoutSeconds: 1
                   failureThreshold: 3
               readiness: *probes
+            resources:
+              requests:
+                cpu: 1m
+              limits:
+                memory: 128Mi
             securityContext: *securityContext
-            resources: *resources
     defaultPodOptions:
       dnsConfig:
         options:

diff --git a/kubernetes/apps/observability/kube-prometheus-stack/app/helmrelease.yaml b/kubernetes/apps/observability/kube-prometheus-stack/app/helmrelease.yaml
@@ -67,6 +67,11 @@ spec:
               resources:
                 requests:
                   storage: 1Gi
+        resources:
+          requests:
+            cpu: 1m
+          limits:
+            memory: 50Mi
 
     grafana:
       enabled: false

diff --git a/kubernetes/apps/observability/silence-operator/app/helmrelease.yaml b/kubernetes/apps/observability/silence-operator/app/helmrelease.yaml
@@ -82,15 +82,12 @@ spec:
                 enabled: false
             resources:
               requests:
-                cpu: 10m
-                memory: 128Mi
+                cpu: 1m
               limits:
-                memory: 256Mi
+                memory: 30Mi
             securityContext:
               allowPrivilegeEscalation: false
-              capabilities:
-                drop:
-                  - "ALL"
+              capabilities: { drop: ["ALL"] }
               seccompProfile:
                 type: RuntimeDefault
     persistence:

diff --git a/kubernetes/apps/observability/unpoller/app/helmrelease.yaml b/kubernetes/apps/observability/unpoller/app/helmrelease.yaml
@@ -59,7 +59,7 @@ spec:
               capabilities: { drop: ["ALL"] }
             resources:
               requests:
-                cpu: 10m
+                cpu: 5m
               limits:
                 memory: 64Mi
     defaultPodOptions:

diff --git a/kubernetes/apps/observability/vector-aggregator/app/helmrelease.yaml b/kubernetes/apps/observability/vector-aggregator/app/helmrelease.yaml
@@ -34,6 +34,11 @@ spec:
         containerPort: 6000
       - name: prom-exporter
         containerPort: 9090
+    resources:
+      requests:
+        cpu: 3m
+      limits:
+        memory: 512Mi
     service:
       ports:
       - name: vector

diff --git a/kubernetes/apps/observability/victorialogs/app/helmrelease.yaml b/kubernetes/apps/observability/victorialogs/app/helmrelease.yaml
@@ -39,9 +39,7 @@ spec:
       resources:
         requests:
           cpu: 5m
-          memory: 50Mi
         limits:
-          cpu: 200m
           memory: 300Mi
       ingress:
         enabled: true

diff --git a/kubernetes/apps/observability/victoriametrics/operator/helmrelease.yaml b/kubernetes/apps/observability/victoriametrics/operator/helmrelease.yaml
@@ -32,13 +32,6 @@ spec:
       pspEnabled: false
     operator:
       enable_converter_ownership: true
-    resources:
-      limits:
-        cpu: 200m
-        memory: 128Mi
-      requests:
-        cpu: 5m
-        memory: 50M
     admissionWebhooks:
       enabled: true
       certManager:

diff --git a/kubernetes/apps/observability/victoriametrics/stack/vmagent.yaml b/kubernetes/apps/observability/victoriametrics/stack/vmagent.yaml
@@ -16,11 +16,9 @@ spec:
   replicaCount: 1
   resources:
     requests:
-      cpu: 300m
-      memory: 128Mi
+      cpu: 3m
     limits:
-      cpu: 1
-      memory: 1Gi
+      memory: 128Mi
   scrapeInterval: 60s
   selectAllByDefault: true
   statefulMode: true

diff --git a/kubernetes/apps/observability/victoriametrics/stack/vmalert.yaml b/kubernetes/apps/observability/victoriametrics/stack/vmalert.yaml
@@ -20,11 +20,9 @@ spec:
   replicaCount: 1
   resources:
     requests:
-      cpu: 20m
-      memory: 64Mi
+      cpu: 3m
     limits:
-      cpu: 1
-      memory: 1Gi
+      memory: 128Mi
   selectAllByDefault: true
   topologySpreadConstraints:
     - maxSkew: 1

diff --git a/kubernetes/apps/observability/victoriametrics/stack/vmauth.yaml b/kubernetes/apps/observability/victoriametrics/stack/vmauth.yaml
@@ -15,11 +15,9 @@ spec:
   replicaCount: 1
   resources:
     requests:
-      memory: 64Mi
-      cpu: 250m
+      cpu: 1m
     limits:
-      memory: 128Mi
-      cpu: 500m
+      memory: 64Mi
   topologySpreadConstraints:
     - maxSkew: 1
       topologyKey: kubernetes.io/hostname

diff --git a/kubernetes/apps/observability/victoriametrics/stack/vmsingle.yaml b/kubernetes/apps/observability/victoriametrics/stack/vmsingle.yaml
@@ -12,10 +12,8 @@ spec:
   replicaCount: 1
   resources:
     requests:
-      cpu: 200m
-      memory: 1Gi
+      cpu: 5m
     limits:
-      cpu: 1
       memory: 2Gi
   # -- Data retention period. Possible units character: h(ours), d(ays), w(eeks), y(ears), if no unit character specified - month. The minimum retention period is 24h. See these [docs](https://docs.victoriametrics.com/single-server-victoriametrics/#retention)
   retentionPeriod: "6"