Merge pull request #1381 from porter-dev/ym/gpu_autoscaling

feat: umbrella templating for gpu autoscaling
porter-dev · Aug 19, 2024 · 129fcca · 129fcca
2 parents 6f66e53 + 53054ec
commit 129fcca
Show file tree

Hide file tree

Showing 5 changed files with 78 additions and 1 deletion.
diff --git a/applications/testdata/web.yaml b/applications/testdata/web.yaml
@@ -1,5 +1,17 @@
 autoscaling:
-  enabled: false
+  enabled: true
+  behavior:
+    enabled: true
+    scaleDown:
+      stabilizationWindowSeconds: 3600
+      type: Pods
+      value: 2
+      periodSeconds: 60
+    scaleUp:
+      stabilizationWindowSeconds: 3600
+      type: Pods
+      value: 2
+      periodSeconds: 60
   gpu: false
   maxReplicas: 10
   minReplicas: 1

diff --git a/applications/web/templates/hpa-blue-green.yaml b/applications/web/templates/hpa-blue-green.yaml
@@ -33,6 +33,26 @@ spec:
           type: Utilization
           averageUtilization: {{ . }}
 {{- end }}
+{{- if .Values.autoscaling.gpu }}
+  {{- with .Values.autoscaling.targetGPUUtilizationPercentage }}
+    - type: Pods
+      pods:
+        metric:
+          name: gpu_utilization
+        target:
+          type: AverageValue
+          averageValue: {{ . }}
+    {{- end }}
+    {{- with .Values.autoscaling.targetGPUMemoryUtilizationRatio }}
+    - type: Pods
+      pods:
+        metric:
+          name: gpu_memory_utilization_ratio
+        target:
+          type: AverageValue
+          averageValue: {{ . | div 100.0 }}
+    {{- end }}
+{{- end }}
 {{- end }}
 {{- end }}
 {{- end }}
diff --git a/applications/web/templates/hpa.yaml b/applications/web/templates/hpa.yaml
@@ -12,11 +12,13 @@ spec:
   {{- if .Values.autoscaling.behavior.enabled }}
   behavior:
     scaleDown:
+      stabilizationWindowSeconds: {{ .Values.autoscaling.behavior.scaleDown.stabilizationWindowSeconds | default 60 }}
       policies:
       - type: {{ .Values.autoscaling.behavior.scaleDown.type }}
         value: {{ .Values.autoscaling.behavior.scaleDown.value }}
         periodSeconds: {{ .Values.autoscaling.behavior.scaleDown.periodSeconds }}
     scaleUp:
+      stabilizationWindowSeconds: {{ .Values.autoscaling.behavior.scaleUp.stabilizationWindowSeconds | default 60 }}
       policies:
       - type: {{ .Values.autoscaling.behavior.scaleUp.type }}
         value: {{ .Values.autoscaling.behavior.scaleUp.value }}
@@ -50,5 +52,25 @@ spec:
           type: Utilization
           averageUtilization: {{ . }}
 {{- end }}
+{{- if .Values.autoscaling.gpu }}
+  {{- with .Values.autoscaling.targetGPUUtilizationPercentage }}
+    - type: Pods
+      pods:
+        metric:
+          name: gpu_utilization
+        target:
+          type: AverageValue
+          averageValue: {{ . }}
+    {{- end }}
+    {{- with .Values.autoscaling.targetGPUMemoryUtilizationRatio }}
+    - type: Pods
+      pods:
+        metric:
+          name: gpu_memory_utilization_ratio
+        target:
+          type: AverageValue
+          averageValue: {{ . | div 100.0 }}
+    {{- end }}
+{{- end }}
 {{- end }}
 {{- end }}
diff --git a/applications/web/values.yaml b/applications/web/values.yaml
@@ -124,10 +124,13 @@ resources:
 
 autoscaling:
   enabled: false
+  gpu: false
   minReplicas: 1
   maxReplicas: 10
   targetCPUUtilizationPercentage: 50
   targetMemoryUtilizationPercentage: 50
+  targetGPUUtilizationPercentage: 50
+  targetGPUMemoryUtilizationRatio: 50
   behavior:
     enabled: false
     scaleDown:

diff --git a/applications/worker/templates/hpa.yaml b/applications/worker/templates/hpa.yaml
@@ -49,4 +49,24 @@ spec:
           type: Utilization
           averageUtilization: {{ . }}
 {{- end }}
+{{- if .Values.autoscaling.gpu }}
+  {{- with .Values.autoscaling.targetGPUUtilizationPercentage }}
+    - type: Pods
+      pods:
+        metric:
+          name: gpu_utilization
+        target:
+          type: AverageValue
+          averageValue: {{ . }}
+    {{- end }}
+    {{- with .Values.autoscaling.targetGPUMemoryUtilizationRatio }}
+    - type: Pods
+      pods:
+        metric:
+          name: gpu_memory_utilization_ratio
+        target:
+          type: AverageValue
+          averageValue: {{ . | div 100.0 }}
+    {{- end }}
+{{- end }}
 {{- end }}