helmchart: add image pull policy for serving runtimes (kserve#4814)

Signed-off-by: Stergios Gemelas <sgemelas8@gmail.com>
Co-authored-by: Sivanantham <90966311+sivanantha321@users.noreply.github.com>

Author: stergem

charts/kserve-resources/README.md

@@ -109,6 +109,7 @@ $ helm install kserve oci://ghcr.io/kserve/charts/kserve --version v0.16.0
 | kserve.servingruntime.huggingfaceserver.disabled | bool | `false` |  |
 | kserve.servingruntime.huggingfaceserver.hostIPC.enabled | bool | `false` |  |
 | kserve.servingruntime.huggingfaceserver.image | string | `"kserve/huggingfaceserver"` |  |
+| kserve.servingruntime.huggingfaceserver.imagePullPolicy | string | `"IfNotPresent"` |  |
 | kserve.servingruntime.huggingfaceserver.imagePullSecrets | list | `[]` |  |
 | kserve.servingruntime.huggingfaceserver.lmcacheUseExperimental | string | `"True"` |  |
 | kserve.servingruntime.huggingfaceserver.securityContext.allowPrivilegeEscalation | bool | `false` |  |
@@ -126,6 +127,7 @@ $ helm install kserve oci://ghcr.io/kserve/charts/kserve --version v0.16.0
 | kserve.servingruntime.huggingfaceserver_multinode.shm.sizeLimit | string | `"3Gi"` |  |
 | kserve.servingruntime.lgbserver.disabled | bool | `false` |  |
 | kserve.servingruntime.lgbserver.image | string | `"kserve/lgbserver"` |  |
+| kserve.servingruntime.lgbserver.imagePullPolicy | string | `"IfNotPresent"` |  |
 | kserve.servingruntime.lgbserver.imagePullSecrets | list | `[]` |  |
 | kserve.servingruntime.lgbserver.securityContext.allowPrivilegeEscalation | bool | `false` |  |
 | kserve.servingruntime.lgbserver.securityContext.capabilities.drop[0] | string | `"ALL"` |  |
@@ -134,6 +136,7 @@ $ helm install kserve oci://ghcr.io/kserve/charts/kserve --version v0.16.0
 | kserve.servingruntime.lgbserver.tag | string | `"v0.16.0"` |  |
 | kserve.servingruntime.mlserver.disabled | bool | `false` |  |
 | kserve.servingruntime.mlserver.image | string | `"docker.io/seldonio/mlserver"` |  |
+| kserve.servingruntime.mlserver.imagePullPolicy | string | `"IfNotPresent"` |  |
 | kserve.servingruntime.mlserver.imagePullSecrets | list | `[]` |  |
 | kserve.servingruntime.mlserver.modelClassPlaceholder | string | `"{{.Labels.modelClass}}"` |  |
 | kserve.servingruntime.mlserver.securityContext.allowPrivilegeEscalation | bool | `false` |  |
@@ -144,6 +147,7 @@ $ helm install kserve oci://ghcr.io/kserve/charts/kserve --version v0.16.0
 | kserve.servingruntime.modelNamePlaceholder | string | `"{{.Name}}"` |  |
 | kserve.servingruntime.paddleserver.disabled | bool | `false` |  |
 | kserve.servingruntime.paddleserver.image | string | `"kserve/paddleserver"` |  |
+| kserve.servingruntime.paddleserver.imagePullPolicy | string | `"IfNotPresent"` |  |
 | kserve.servingruntime.paddleserver.imagePullSecrets | list | `[]` |  |
 | kserve.servingruntime.paddleserver.securityContext.allowPrivilegeEscalation | bool | `false` |  |
 | kserve.servingruntime.paddleserver.securityContext.capabilities.drop[0] | string | `"ALL"` |  |
@@ -152,6 +156,7 @@ $ helm install kserve oci://ghcr.io/kserve/charts/kserve --version v0.16.0
 | kserve.servingruntime.paddleserver.tag | string | `"v0.16.0"` |  |
 | kserve.servingruntime.pmmlserver.disabled | bool | `false` |  |
 | kserve.servingruntime.pmmlserver.image | string | `"kserve/pmmlserver"` |  |
+| kserve.servingruntime.pmmlserver.imagePullPolicy | string | `"IfNotPresent"` |  |
 | kserve.servingruntime.pmmlserver.imagePullSecrets | list | `[]` |  |
 | kserve.servingruntime.pmmlserver.securityContext.allowPrivilegeEscalation | bool | `false` |  |
 | kserve.servingruntime.pmmlserver.securityContext.capabilities.drop[0] | string | `"ALL"` |  |
@@ -160,6 +165,7 @@ $ helm install kserve oci://ghcr.io/kserve/charts/kserve --version v0.16.0
 | kserve.servingruntime.pmmlserver.tag | string | `"v0.16.0"` |  |
 | kserve.servingruntime.sklearnserver.disabled | bool | `false` |  |
 | kserve.servingruntime.sklearnserver.image | string | `"kserve/sklearnserver"` |  |
+| kserve.servingruntime.sklearnserver.imagePullPolicy | string | `"IfNotPresent"` |  |
 | kserve.servingruntime.sklearnserver.imagePullSecrets | list | `[]` |  |
 | kserve.servingruntime.sklearnserver.securityContext.allowPrivilegeEscalation | bool | `false` |  |
 | kserve.servingruntime.sklearnserver.securityContext.capabilities.drop[0] | string | `"ALL"` |  |
@@ -168,6 +174,7 @@ $ helm install kserve oci://ghcr.io/kserve/charts/kserve --version v0.16.0
 | kserve.servingruntime.sklearnserver.tag | string | `"v0.16.0"` |  |
 | kserve.servingruntime.tensorflow.disabled | bool | `false` |  |
 | kserve.servingruntime.tensorflow.image | string | `"tensorflow/serving"` |  |
+| kserve.servingruntime.tensorflow.imagePullPolicy | string | `"IfNotPresent"` |  |
 | kserve.servingruntime.tensorflow.imagePullSecrets | list | `[]` |  |
 | kserve.servingruntime.tensorflow.securityContext.allowPrivilegeEscalation | bool | `false` |  |
 | kserve.servingruntime.tensorflow.securityContext.capabilities.drop[0] | string | `"ALL"` |  |
@@ -177,6 +184,7 @@ $ helm install kserve oci://ghcr.io/kserve/charts/kserve --version v0.16.0
 | kserve.servingruntime.tensorflow.tag | string | `"2.6.2"` |  |
 | kserve.servingruntime.torchserve.disabled | bool | `false` |  |
 | kserve.servingruntime.torchserve.image | string | `"pytorch/torchserve-kfs"` |  |
+| kserve.servingruntime.torchserve.imagePullPolicy | string | `"IfNotPresent"` |  |
 | kserve.servingruntime.torchserve.imagePullSecrets | list | `[]` |  |
 | kserve.servingruntime.torchserve.securityContext.allowPrivilegeEscalation | bool | `false` |  |
 | kserve.servingruntime.torchserve.securityContext.capabilities.drop[0] | string | `"ALL"` |  |
@@ -187,6 +195,7 @@ $ helm install kserve oci://ghcr.io/kserve/charts/kserve --version v0.16.0
 | kserve.servingruntime.torchserve.tag | string | `"0.9.0"` |  |
 | kserve.servingruntime.tritonserver.disabled | bool | `false` |  |
 | kserve.servingruntime.tritonserver.image | string | `"nvcr.io/nvidia/tritonserver"` |  |
+| kserve.servingruntime.tritonserver.imagePullPolicy | string | `"IfNotPresent"` |  |
 | kserve.servingruntime.tritonserver.imagePullSecrets | list | `[]` |  |
 | kserve.servingruntime.tritonserver.securityContext.allowPrivilegeEscalation | bool | `false` |  |
 | kserve.servingruntime.tritonserver.securityContext.capabilities.drop[0] | string | `"ALL"` |  |
@@ -196,6 +205,7 @@ $ helm install kserve oci://ghcr.io/kserve/charts/kserve --version v0.16.0
 | kserve.servingruntime.tritonserver.tag | string | `"23.05-py3"` |  |
 | kserve.servingruntime.xgbserver.disabled | bool | `false` |  |
 | kserve.servingruntime.xgbserver.image | string | `"kserve/xgbserver"` |  |
+| kserve.servingruntime.xgbserver.imagePullPolicy | string | `"IfNotPresent"` |  |
 | kserve.servingruntime.xgbserver.imagePullSecrets | list | `[]` |  |
 | kserve.servingruntime.xgbserver.securityContext.allowPrivilegeEscalation | bool | `false` |  |
 | kserve.servingruntime.xgbserver.securityContext.capabilities.drop[0] | string | `"ALL"` |  |

charts/kserve-resources/templates/clusterservingruntimes.yaml

@@ -18,6 +18,7 @@ spec:
   containers:
     - name: kserve-container
       image: "{{ .Values.kserve.servingruntime.lgbserver.image }}:{{ .Values.kserve.servingruntime.lgbserver.tag }}"
+      imagePullPolicy: {{ .Values.kserve.servingruntime.lgbserver.imagePullPolicy }}
       args:
         - --model_name={{ .Values.kserve.servingruntime.modelNamePlaceholder }}
         - --model_dir=/mnt/models
@@ -88,6 +89,7 @@ spec:
   containers:
     - name: kserve-container
       image: "{{ .Values.kserve.servingruntime.mlserver.image }}:{{ .Values.kserve.servingruntime.mlserver.tag }}"
+      imagePullPolicy: {{ .Values.kserve.servingruntime.mlserver.imagePullPolicy }}
       env:
         - name: "MLSERVER_MODEL_IMPLEMENTATION"
           value: "{{ .Values.kserve.servingruntime.mlserver.modelClassPlaceholder }}"
@@ -134,6 +136,7 @@ spec:
   containers:
     - name: kserve-container
       image: "{{ .Values.kserve.servingruntime.paddleserver.image }}:{{ .Values.kserve.servingruntime.paddleserver.tag }}"
+      imagePullPolicy: {{ .Values.kserve.servingruntime.paddleserver.imagePullPolicy }}
       args:
         - --model_name={{ .Values.kserve.servingruntime.modelNamePlaceholder }}
         - --model_dir=/mnt/models
@@ -179,6 +182,7 @@ spec:
   containers:
     - name: kserve-container
       image: "{{ .Values.kserve.servingruntime.pmmlserver.image }}:{{ .Values.kserve.servingruntime.pmmlserver.tag }}"
+      imagePullPolicy: {{ .Values.kserve.servingruntime.pmmlserver.imagePullPolicy }}
       args:
         - --model_name={{ .Values.kserve.servingruntime.modelNamePlaceholder }}
         - --model_dir=/mnt/models
@@ -220,6 +224,7 @@ spec:
   containers:
     - name: kserve-container
       image: "{{ .Values.kserve.servingruntime.sklearnserver.image }}:{{ .Values.kserve.servingruntime.sklearnserver.tag }}"
+      imagePullPolicy: {{ .Values.kserve.servingruntime.sklearnserver.imagePullPolicy }}
       args:
         - --model_name={{ .Values.kserve.servingruntime.modelNamePlaceholder }}
         - --model_dir=/mnt/models
@@ -265,6 +270,7 @@ spec:
   containers:
     - name: kserve-container
       image: "{{ .Values.kserve.servingruntime.tensorflow.image }}:{{ .Values.kserve.servingruntime.tensorflow.tag }}"
+      imagePullPolicy: {{ .Values.kserve.servingruntime.tensorflow.imagePullPolicy }}
       command: [/usr/bin/tensorflow_model_server]
       args:
         - --model_name={{ .Values.kserve.servingruntime.modelNamePlaceholder }}
@@ -310,6 +316,7 @@ spec:
   containers:
     - name: kserve-container
       image: "{{ .Values.kserve.servingruntime.torchserve.image }}:{{ .Values.kserve.servingruntime.torchserve.tag }}"
+      imagePullPolicy: {{ .Values.kserve.servingruntime.torchserve.imagePullPolicy }}
       args:
         - torchserve
         - --start
@@ -373,6 +380,7 @@ spec:
   containers:
     - name: kserve-container
       image: "{{ .Values.kserve.servingruntime.tritonserver.image }}:{{ .Values.kserve.servingruntime.tritonserver.tag }}"
+      imagePullPolicy: {{ .Values.kserve.servingruntime.tritonserver.imagePullPolicy }}
       args:
         - tritonserver
         - --model-store=/mnt/models
@@ -417,6 +425,7 @@ spec:
   containers:
     - name: kserve-container
       image: "{{ .Values.kserve.servingruntime.xgbserver.image }}:{{ .Values.kserve.servingruntime.xgbserver.tag }}"
+      imagePullPolicy: {{ .Values.kserve.servingruntime.xgbserver.imagePullPolicy }}
       args:
         - --model_name={{ .Values.kserve.servingruntime.modelNamePlaceholder }}
         - --model_dir=/mnt/models
@@ -458,6 +467,7 @@ spec:
   containers:
     - name: kserve-container
       image: "{{ .Values.kserve.servingruntime.huggingfaceserver.image }}:{{ .Values.kserve.servingruntime.huggingfaceserver.tag }}"
+      imagePullPolicy: {{ .Values.kserve.servingruntime.huggingfaceserver.imagePullPolicy }}
       args:
         - --model_name={{ .Values.kserve.servingruntime.modelNamePlaceholder }}
       env:
@@ -516,6 +526,7 @@ spec:
   containers:
     - name: kserve-container
       image: "{{ .Values.kserve.servingruntime.huggingfaceserver.image }}:{{ .Values.kserve.servingruntime.huggingfaceserver.tag }}-gpu"
+      imagePullPolicy: {{ .Values.kserve.servingruntime.huggingfaceserver.imagePullPolicy }}
       args:
         - --model_name={{ .Values.kserve.servingruntime.modelNamePlaceholder }}
       command:
@@ -617,6 +628,7 @@ spec:
     containers:
       - name: worker-container
         image: "{{ .Values.kserve.servingruntime.huggingfaceserver.image }}:{{ .Values.kserve.servingruntime.huggingfaceserver.tag }}-gpu"
+        imagePullPolicy: {{ .Values.kserve.servingruntime.huggingfaceserver.imagePullPolicy }}
         command:
         - "bash"
         - "-c"

charts/kserve-resources/values.yaml

@@ -288,6 +288,7 @@ kserve:
       image: tensorflow/serving
       tag: 2.6.2
       imagePullSecrets: []
+      imagePullPolicy: IfNotPresent
       securityContext:
         runAsUser: 1000 # User is not defined in the Dockerfile, so we need to set it here to run as non-root
         allowPrivilegeEscalation: false
@@ -302,6 +303,7 @@ kserve:
       tag: 1.5.0
       modelClassPlaceholder: "{{.Labels.modelClass}}"
       imagePullSecrets: []
+      imagePullPolicy: IfNotPresent
       securityContext:
         allowPrivilegeEscalation: false
         privileged: false
@@ -314,6 +316,7 @@ kserve:
       image: kserve/sklearnserver
       tag: *defaultVersion
       imagePullSecrets: []
+      imagePullPolicy: IfNotPresent
       securityContext:
         allowPrivilegeEscalation: false
         privileged: false
@@ -326,6 +329,7 @@ kserve:
       image: kserve/xgbserver
       tag: *defaultVersion
       imagePullSecrets: []
+      imagePullPolicy: IfNotPresent
       securityContext:
         allowPrivilegeEscalation: false
         privileged: false
@@ -338,6 +342,7 @@ kserve:
       image: kserve/huggingfaceserver
       tag: *defaultVersion
       imagePullSecrets: []
+      imagePullPolicy: IfNotPresent
       securityContext:
         allowPrivilegeEscalation: false
         privileged: false
@@ -369,6 +374,7 @@ kserve:
       image: nvcr.io/nvidia/tritonserver
       tag: 23.05-py3
       imagePullSecrets: []
+      imagePullPolicy: IfNotPresent
       securityContext:
         runAsUser: 1000 # https://docs.nvidia.com/deeplearning/triton-inference-server/user-guide/docs/customization_guide/deploy.html#run-as-a-non-root-user
         allowPrivilegeEscalation: false
@@ -382,6 +388,7 @@ kserve:
       image: kserve/pmmlserver
       tag: *defaultVersion
       imagePullSecrets: []
+      imagePullPolicy: IfNotPresent
       securityContext:
         allowPrivilegeEscalation: false
         privileged: false
@@ -394,6 +401,7 @@ kserve:
       image: kserve/paddleserver
       tag: *defaultVersion
       imagePullSecrets: []
+      imagePullPolicy: IfNotPresent
       securityContext:
         allowPrivilegeEscalation: false
         privileged: false
@@ -406,6 +414,7 @@ kserve:
       image: kserve/lgbserver
       tag: *defaultVersion
       imagePullSecrets: []
+      imagePullPolicy: IfNotPresent
       securityContext:
         allowPrivilegeEscalation: false
         privileged: false
@@ -419,6 +428,7 @@ kserve:
       tag: 0.9.0
       serviceEnvelopePlaceholder: "{{.Labels.serviceEnvelope}}"
       imagePullSecrets: []
+      imagePullPolicy: IfNotPresent
       securityContext:
         runAsUser: 1000 # User ID is not defined in the Dockerfile, so we need to set it here to run as non-root
         allowPrivilegeEscalation: false