nelm-ready: public images overlays, nodeSelector fix, secrets

.helm/Chart.yaml

@@ -0,0 +1,6 @@
+apiVersion: v2
+name: vllm
+description: vLLM for nelm deploy
+type: application
+version: 1.0.0
+appVersion: "0.14.0"

.helm/templates/dashboard.yaml

@@ -1,4 +1,4 @@
-{{- if eq .Values.werf.env "prod" }}
+{{- if eq .Values.env "prod" }}
 {{- range $path, $_ := .Files.Glob "dashboards/*.json" }}
   {{- $raw := $.Files.Get $path }}
   {{- $json := $raw | fromJson }}

.helm/templates/ingress.yaml

@@ -13,13 +13,15 @@ metadata:
     nginx.ingress.kubernetes.io/proxy-read-timeout: "10000"
     nginx.ingress.kubernetes.io/upstream-keepalive-timeout: "10000"
 spec:
-  ingressClassName: {{ dig .Values.werf.env .Values.ingress.className._default .Values.ingress.className | quote }}
+  ingressClassName: {{ dig .Values.env .Values.ingress.className._default .Values.ingress.className | quote }}
+  {{- if .Values.ingress.tls.enabled }}
   tls:
     - hosts:
-        - {{ dig .Values.werf.env .Values.ingress.host._default .Values.ingress.host | quote }}
+        - {{ dig .Values.env .Values.ingress.host._default .Values.ingress.host | quote }}
-      secretName: {{ dig .Values.werf.env .Values.ingress.secretName._default .Values.ingress.secretName | quote }}
+      secretName: {{ dig .Values.env .Values.ingress.secretName._default .Values.ingress.secretName | quote }}
+  {{- end }}
   rules:
-    - host: {{ dig .Values.werf.env .Values.ingress.host._default .Values.ingress.host | quote }}
+    - host: {{ dig .Values.env .Values.ingress.host._default .Values.ingress.host | quote }}
       http:
         paths:
           - path: /
@@ -29,16 +31,16 @@ spec:
                 name: {{ .Chart.Name }}
                 port:
                   name: http
+{{- if .Values.ingress.tls.enabled }}
 ---
 apiVersion: v1
 kind: Secret
 metadata:
-  name: {{ dig .Values.werf.env .Values.ingress.secretName._default .Values.ingress.secretName }}
+  name: {{ dig .Values.env .Values.ingress.secretName._default .Values.ingress.secretName }}
-  annotations:
-    werf.io/sensitive: "true"
 type: kubernetes.io/tls
 stringData:
   tls.crt: |
-    {{- werf_secret_file (printf "%s/%s" .Values.werf.env "tls.crt") | nindent 4 }}
+{{ dig .Values.env .Values.tls.crt._default .Values.tls.crt | nindent 4 }}
   tls.key: |
-    {{- werf_secret_file (printf "%s/%s" .Values.werf.env "tls.key") | nindent 4 }}
+{{ dig .Values.env .Values.tls.key._default .Values.tls.key | nindent 4 }}
+{{- end }}

.helm/templates/sts.yaml

@@ -34,11 +34,13 @@ spec:
       {{- if and .Values.gpu.enabled .Values.gpu.tolerations }}
       tolerations: {{ toYaml .Values.gpu.tolerations | nindent 6 }}
       {{- end }}
+      {{- with .Values.imagePullSecrets }}
       imagePullSecrets:
-      - name: registrysecret
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
       containers:
         - name: {{ .Chart.Name }}
-          image: {{ .Values.werf.image.vllm }}
+          image: {{ .Values.image.vllm }}
           imagePullPolicy: IfNotPresent
           volumeMounts:
             - name: data
@@ -48,7 +50,7 @@ spec:
               containerPort: {{ .Values.port }}
               protocol: TCP
           args:
-          - {{ .Values.env.VLLM_MODEL }}
+          - {{ .Values.envVars.VLLM_MODEL }}
           - --async-scheduling
           - --host
           - 0.0.0.0
@@ -57,7 +59,7 @@ spec:
           - --gpu-memory-utilization
           - "0.2"
           - --max-model-len
-          - "{{ .Values.env.VLLM_MAX_MODEL_LEN }}"
+          - "{{ .Values.envVars.VLLM_MAX_MODEL_LEN }}"
           - --enforce-eager
           - --tensor-parallel-size
           - "1"
@@ -65,7 +67,7 @@ spec:
           - --chat-template-content-format
           - "string"
           env:
-          {{- range $n, $v := .Values.env }}
+          {{- range $n, $v := .Values.envVars }}
             - name: {{ $n }}
               value: {{ $v | quote }}
           {{- end }}
@@ -85,16 +87,16 @@ spec:
             periodSeconds: 5
             timeoutSeconds: 6
             failureThreshold: 3
-          resources: {{ dig .Values.werf.env .Values.resources._default .Values.resources | toYaml | nindent 12 }}
+          resources: {{ dig .Values.env .Values.resources._default .Values.resources | toYaml | nindent 12 }}
   volumeClaimTemplates:
   - metadata:
       name: data
     spec:
       accessModes: [ "ReadWriteOnce" ]
-      storageClassName: {{ dig .Values.werf.env .Values.storage.class._default .Values.storage.class | quote }}
+      storageClassName: {{ dig .Values.env .Values.storage.class._default .Values.storage.class | quote }}
       resources:
         requests:
-          storage: {{ dig .Values.werf.env .Values.storage.size._default .Values.storage.size | quote }}
+          storage: {{ dig .Values.env .Values.storage.size._default .Values.storage.size | quote }}
 ---
 apiVersion: v1
 kind: Service

.helm/values-public.yaml

@@ -0,0 +1,22 @@
+image:
+  vllm: docker.io/vllm/vllm-openai:v0.14.0
+imagePullSecrets: []
+gpu:
+  enabled: false
+ingress:
+  tls:
+    enabled: false
+resources:
+  _default:
+    requests:
+      cpu: "100m"
+      memory: "512Mi"
+    limits:
+      memory: "1Gi"
+
+nodeSelector: {}
+tolerations: []
+gpu:
+  enabled: false
+  nodeSelector: {}
+  tolerations: []

.helm/values.yaml

@@ -1,5 +1,17 @@
+global:
+  cluster: ""
+
+env: dev
+
+imagePullSecrets:
+  - name: registrysecret
+
+image:
+  vllm: 192.168.10.68:5000/vllm/vllm-openai:v0.14.0
+
 port: 8004
-env:
+
+envVars:
   VLLM_MODEL: kristaller486/dots.ocr-1.5
   VLLM_MAX_TOKENS: 24000
   VLLM_TEMPERATURE: 0.2
@@ -39,14 +51,16 @@ storage:
   size:
     _default: 10Gi
   class:
-    _default: local-storage-class-cuda-worker
+    _default: localpath
 
 ingress:
+  tls:
+    enabled: false
   className:
     _default: nginx
   host:
-    _default: vllm.pro.lukoil.com
+    _default: vllm.valdemorte.online
-    preprod: vllm.preprod.pro.lukoil.com
+    preprod: vllm.preprod.valdemorte.online
-    prod: vllm.prod.pro.lukoil.com
+    prod: vllm.prod.valdemorte.online
   secretName:
     _default: vllm-tls

Dockerfile

@@ -0,0 +1 @@
+FROM docker.io/vllm/vllm-openai:v0.14.0