fix(helm): leave spec.replicas unset unless replicaCount is explicitly configured

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

helm/litellm/templates/backend/deployment.yaml

@@ -7,7 +7,7 @@ metadata:
     {{- include "litellm.commonLabels" . | nindent 4 }}
     app.kubernetes.io/component: backend
 spec:
-  {{- if not .Values.backend.hpa.enabled }}
+  {{- if and (not .Values.backend.hpa.enabled) (not (kindIs "invalid" .Values.backend.replicaCount)) }}
   replicas: {{ .Values.backend.replicaCount }}
   {{- end }}
   {{- with .Values.backend.strategy }}

helm/litellm/templates/gateway/deployment.yaml

@@ -7,7 +7,7 @@ metadata:
     {{- include "litellm.commonLabels" . | nindent 4 }}
     app.kubernetes.io/component: gateway
 spec:
-  {{- if not .Values.gateway.hpa.enabled }}
+  {{- if and (not .Values.gateway.hpa.enabled) (not (kindIs "invalid" .Values.gateway.replicaCount)) }}
   replicas: {{ .Values.gateway.replicaCount }}
   {{- end }}
   {{- with .Values.gateway.strategy }}

helm/litellm/templates/ui/deployment.yaml

@@ -7,7 +7,7 @@ metadata:
     {{- include "litellm.commonLabels" . | nindent 4 }}
     app.kubernetes.io/component: ui
 spec:
-  {{- if not .Values.ui.hpa.enabled }}
+  {{- if and (not .Values.ui.hpa.enabled) (not (kindIs "invalid" .Values.ui.replicaCount)) }}
   replicas: {{ .Values.ui.replicaCount }}
   {{- end }}
   {{- with .Values.ui.strategy }}

helm/litellm/tests/replica_count_tests.yaml

@@ -49,6 +49,22 @@ tests:
           path: spec.replicas
           value: 0
 
+  - it: a component with HPA disabled but no replicaCount set keeps omitting spec.replicas, so upgrades do not reset a hand-scaled Deployment
+    set:
+      gateway.hpa.enabled: false
+      backend.hpa.enabled: false
+      ui.hpa.enabled: false
+    asserts:
+      - notExists:
+          path: spec.replicas
+        template: gateway/deployment.yaml
+      - notExists:
+          path: spec.replicas
+        template: backend/deployment.yaml
+      - notExists:
+          path: spec.replicas
+        template: ui/deployment.yaml
+
   - it: every component omits spec.replicas when its HPA is enabled, so the autoscaler owns the count
     set:
       gateway.hpa.enabled: true

helm/litellm/values.yaml

@@ -397,10 +397,11 @@ gateway:
   #   failureThreshold: 30
   #   periodSeconds: 10
   startupProbe: {}
-  # Fixed pod count, rendered into the Deployment's spec.replicas only when
-  # hpa.enabled is false. With the HPA on, the autoscaler owns the count and
-  # this value is ignored.
-  replicaCount: 1
+  # Optional fixed pod count, rendered into the Deployment's spec.replicas only
+  # when hpa.enabled is false. Unset by default so an existing Deployment keeps
+  # its current count; with the HPA on, the autoscaler owns the count, e.g.:
+  #   replicaCount: 3
+  replicaCount:
   hpa:
     enabled: true
     minReplicas: 1
@@ -529,7 +530,7 @@ backend:
   # Optional startupProbe; same shape as gateway.startupProbe. Empty by default.
   startupProbe: {}
   # Same semantics as gateway.replicaCount.
-  replicaCount: 1
+  replicaCount:
   hpa:
     enabled: true
     minReplicas: 1
@@ -597,7 +598,7 @@ ui:
   # Optional startupProbe; same shape as gateway.startupProbe. Empty by default.
   startupProbe: {}
   # Same semantics as gateway.replicaCount.
-  replicaCount: 1
+  replicaCount:
   hpa:
     enabled: false
     minReplicas: 1