feat: GKE examples (#2721)

Signed-off-by: Biswa Panda <biswa.panda@gmail.com>
Signed-off-by: FortunaZhang <175659032+FortunaZhang@users.noreply.github.com>
Co-authored-by: FortunaZhang <175659032+FortunaZhang@users.noreply.github.com>

examples/deployments/GKE/README.md

+# Dynamo Deployment on GKE
+
+## Pre-requisites
+
+### Install gcloud CLI
+https://cloud.google.com/sdk/docs/install
+
+### Create GKE cluster
+
+```bash
+export PROJECT_ID=<>
+export REGION=<>
+export ZONE=<>
+export CLUSTER_NAME=<>
+export CLUSTER_MACHINE_TYPE=n2-standard-4
+export NODE_POOL_MACHINE_TYPE=g2-standard-24
+export GPU_TYPE=nvidia-l4
+export GPU_COUNT=2
+export CPU_NODE=2
+export GPU_NODE=2
+export DISK_SIZE=200
+
+gcloud container clusters create ${CLUSTER_NAME} \
+ 	--project=${PROJECT_ID} \
+ 	--location=${ZONE} \
+	--subnetwork=default \
+    --disk-size=${DISK_SIZE} \
+	--machine-type=${CLUSTER_MACHINE_TYPE} \
+ 	--num-nodes=${CPU_NODE}
+```
+
+#### Create GPU pool
+
+```bash
+gcloud container node-pools create gpu-pool \
+ 	--accelerator type=${GPU_TYPE},count=${GPU_COUNT},gpu-driver-version=latest \
+ 	--project=${PROJECT_ID} \
+ 	--location=${ZONE} \
+ 	--cluster=${CLUSTER_NAME} \
+	--machine-type=${NODE_POOL_MACHINE_TYPE} \
+    --disk-size=${DISK_SIZE} \
+    --num-nodes=${GPU_NODE} \
+    --enable-autoscaling \
+    --min-nodes=1 \
+    --max-nodes=3
+```
+
+###  Clone Dynamo GitHub repository
+
+**Note:** Please make sure GitHub branch/commit version matches with Dynamo platform and VLLM container.
+
+```bash
+git clone https://github.com/ai-dynamo/dynamo.git
+
+# Checkout to the desired branch
+git checkout release/0.6.0
+```
+
+###  Set environment variables for GKE
+
+```bash
+export NAMESPACE=dynamo-system
+kubectl create namespace $NAMESPACE
+kubectl config set-context --current --namespace=$NAMESPACE
+
+export HF_TOKEN=<HF_TOKEN>
+kubectl create secret generic hf-token-secret \
+  --from-literal=HF_TOKEN=${HF_TOKEN} \
+  -n ${NAMESPACE}
+```
+
+## Install Dynamo Kubernetes Platform
+
+[See installation steps](/docs/kubernetes/installation_guide.md#overview)
+
+After installation, verify the installation:
+
+**Expected output**
+
+```bash
+kubectl get pods
+NAME                                                              READY   STATUS             RESTARTS   AGE
+dynamo-platform-dynamo-operator-controller-manager-69b9794fpgv9   2/2     Running            0          4m27s
+dynamo-platform-etcd-0                                            1/1     Running            0          4m27s
+dynamo-platform-nats-0                                            2/2     Running            0          4m27s
+```
+
+## Deploy Inference Graph
+
+We will deploy a LLM model to the Dynamo platform. Here we use `Qwen/Qwen3-0.6B` model with VLLM and disaggregated deployment as an example.
+
+In the deployment yaml file, some adjustments have to/ could be made:
+
+- **(Required)** Add args to change `LD_LIBRARY_PATH` and `PATH` of decoder container, to enable GKE find the correct GPU driver
+- Change VLLM  image to the desired one on NGC
+- Add namespace to metadata
+- Adjust GPU/CPU request and limits
+- Change model to deploy
+
+More configurations please refer to https://github.com/ai-dynamo/dynamo/tree/main/examples/deployments/GKE/vllm
+
+### Highlighted configurations in yaml file
+Please note that `LD_LIBRARY_PATH` needs to be set properly in GKE as per [Run GPUs in GKE](https://cloud.google.com/kubernetes-engine/docs/how-to/gpus)
+
+The following snippet needs to be present in the `args` field of the deployment `yaml` file:
+
+```bash
+export LD_LIBRARY_PATH=/usr/local/nvidia/lib64:$LD_LIBRARY_PATH
+export PATH=$PATH:/usr/local/nvidia/bin:/usr/local/nvidia/lib64
+/sbin/ldconfig
+```
+
+For example, refer to the following from [`examples/deployments/GKE/vllm/disagg_gke.yaml`](./vllm/disagg_gke.yaml)
+
+```yaml
+metadata:
+  name: vllm-disagg
+  namespace: dynamo-system
+spec:
+  services:
+    Frontend:
+          image: nvcr.io/nvidia/ai-dynamo/vllm-runtime:0.6.0
+    VllmDecodeWorker:
+​​      resources:
+        limits:
+          gpu: "3"
+          image: nvcr.io/nvidia/ai-dynamo/vllm-runtime:0.6.0
+          args:
+            - |
+            export LD_LIBRARY_PATH=/usr/local/nvidia/lib64:$LD_LIBRARY_PATH
+            export PATH=$PATH:/usr/local/nvidia/bin:/usr/local/nvidia/lib64
+            /sbin/ldconfig
+            python3 -m dynamo.vllm --model Qwen/Qwen3-0.6B
+```
+
+## Deploy the model
+
+```bash
+cd dynamo/examples/deployments/GKE/vllm
+
+kubectl apply -f disagg_gke.yaml -n ${NAMESPACE}
+```
+
+**Expected output after successful deployment**
+
+```bash
+kubectl get pods
+NAME                                                              READY   STATUS    RESTARTS   AGE
+dynamo-platform-dynamo-operator-controller-manager-c665684ssqkx   2/2     Running   0          65m
+dynamo-platform-etcd-0                                            1/1     Running   0          65m
+dynamo-platform-nats-0                                            2/2     Running   0          65m
+vllm-disagg-frontend-5954ddc4dd-4w2cb                             1/1     Running   0          11m
+vllm-disagg-vllmdecodeworker-77844cfcff-ddn4v                     1/1     Running   0          11m
+vllm-disagg-vllmprefillworker-55d5b74b4f-zrskh                    1/1     Running   0          11m
+```
+
+## Test the Deployment
+
+```bash
+export DEPLOYMENT_NAME=vllm-disagg
+
+# Find the frontend pod
+export FRONTEND_POD=$(kubectl get pods -n ${NAMESPACE} | grep "${DEPLOYMENT_NAME}-frontend" | sort -k1 | tail -n1 | awk '{print $1}')
+
+# Forward the pod's port to localhost
+kubectl port-forward deployment/vllm-disagg-frontend  8000:8000 -n ${NAMESPACE}
+
+# disagg
+curl localhost:8000/v1/chat/completions \
+  -H "Content-Type: application/json" \
+  -d '{
+    "model": "Qwen/Qwen3-0.6B",
+    "messages": [
+    {
+        "role": "user",
+        "content": "In the heart of Eldoria, an ancient land of boundless magic and mysterious creatures, lies the long-forgotten city of Aeloria. Once a beacon of knowledge and power, Aeloria was buried beneath the shifting sands of time, lost to the world for centuries. You are an intrepid explorer, known for your unparalleled curiosity and courage, who has stumbled upon an ancient map hinting at ests that Aeloria holds a secret so profound that it has the potential to reshape the very fabric of reality. Your journey will take you through treacherous deserts, enchanted forests, and across perilous mountain ranges. Your Task: Character Background: Develop a detailed background for your character. Describe their motivations for seeking out Aeloria, their skills and weaknesses, and any personal connections to the ancient city or its legends. Are they driven by a quest for knowledge, a search for lost familt clue is hidden."
+    }
+    ],
+    "stream":false,
+    "max_tokens": 30
+  }'
+```
+
+### Response
+
+```json
+{"id":"chatcmpl-bd0670d9-0342-4eea-97c1-99b69f1f931f","choices":[{"index":0,"message":{"content":"Okay, here’s a detailed character background for your intrepid explorer, tailored to fit the premise of Aeloria, with a focus on a","refusal":null,"tool_calls":null,"role":"assistant","function_call":null,"audio":null},"finish_reason":"stop","logprobs":null}],"created":1756336263,"model":"Qwen/Qwen3-0.6B","service_tier":null,"system_fingerprint":null,"object":"chat.completion","usage":{"prompt_tokens":190,"completion_tokens":29,"total_tokens":219,"prompt_tokens_details":null,"completion_tokens_details":null}}
+```

examples/deployments/GKE/sglang/disagg.yaml

+# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+apiVersion: nvidia.com/v1alpha1
+kind: DynamoGraphDeployment
+metadata:
+  name: sglang-disagg
+spec:
+  services:
+    Frontend:
+      dynamoNamespace: sglang-disagg
+      componentType: frontend
+      replicas: 1
+      extraPodSpec:
+        mainContainer:
+          image: my-registry/sglang-runtime:my-tag
+    decode:
+      envFromSecret: hf-token-secret
+      dynamoNamespace: sglang-disagg
+      componentType: worker
+      subComponentType: decode
+      replicas: 1
+      resources:
+        limits:
+          gpu: "1"
+      extraPodSpec:
+        mainContainer:
+          image: my-registry/sglang-runtime:my-tag
+          workingDir: /workspace/components/backends/sglang
+          command:
+            - /bin/sh
+            - -c
+          args:
+          - |
+            export LD_LIBRARY_PATH=/usr/local/nvidia/lib64:$LD_LIBRARY_PATH
+            export PATH=$PATH:/usr/local/nvidia/bin:/usr/local/nvidia/lib64
+            /sbin/ldconfig
+            nvidia-smi
+            exec python3 -m dynamo.sglang --model-path Qwen/Qwen3-0.6B --served-model-name Qwen/Qwen3-0.6B --page-size 16 --tp 1 --trust-remote-code --skip-tokenizer-init --disaggregation-mode decode --disaggregation-transfer-backend nixl --disaggregation-bootstrap-port --disaggregation-bootstrap-port "12345" --host "0.0.0.0"
+    prefill:
+      envFromSecret: hf-token-secret
+      dynamoNamespace: sglang-disagg
+      componentType: worker
+      subComponentType: prefill
+      replicas: 1
+      resources:
+        limits:
+          gpu: "1"
+      extraPodSpec:
+        mainContainer:
+          image: my-registry/sglang-runtime:my-tag
+          workingDir: /workspace/components/backends/sglang
+          command:
+            - /bin/sh
+            - -c
+          args:
+          - |
+            export LD_LIBRARY_PATH=/usr/local/nvidia/lib64:$LD_LIBRARY_PATH
+            export PATH=$PATH:/usr/local/nvidia/bin:/usr/local/nvidia/lib64
+            /sbin/ldconfig
+            nvidia-smi
+            exec python3 -m dynamo.sglang --model-path Qwen/Qwen3-0.6B --served-model-name Qwen/Qwen3-0.6B --page-size 16 --tp 1 --trust-remote-code --skip-tokenizer-init --disaggregation-mode prefill --disaggregation-transfer-backend nixl --disaggregation-bootstrap-port "12345" --host "0.0.0.0"
\ No newline at end of file

examples/deployments/GKE/vllm/disagg.yaml

+# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+
+apiVersion: nvidia.com/v1alpha1
+kind: DynamoGraphDeployment
+metadata:
+  name: vllm-disagg
+spec:
+  services:
+    Frontend:
+      dynamoNamespace: vllm-disagg
+      componentType: frontend
+      replicas: 1
+      extraPodSpec:
+        mainContainer:
+          image: my-registry/vllm-runtime:my-tag
+    VllmDecodeWorker:
+      dynamoNamespace: vllm-disagg
+      envFromSecret: hf-token-secret
+      componentType: worker
+      subComponentType: decode
+      replicas: 1
+      resources:
+        limits:
+          gpu: "1"
+      extraPodSpec:
+        mainContainer:
+          startupProbe:
+            initialDelaySeconds: 180
+          image: my-registry/vllm-runtime:my-tag
+          workingDir: /workspace/components/backends/vllm
+          command:
+            - /bin/sh
+            - -c
+          args:
+          - |
+            export LD_LIBRARY_PATH=/usr/local/nvidia/lib64:$LD_LIBRARY_PATH
+            export PATH=$PATH:/usr/local/nvidia/bin:/usr/local/nvidia/lib64
+            /sbin/ldconfig
+            python3 -m dynamo.vllm --model Qwen/Qwen3-0.6B
+    VllmPrefillWorker:
+      dynamoNamespace: vllm-disagg
+      envFromSecret: hf-token-secret
+      componentType: worker
+      subComponentType: prefill
+      replicas: 1
+      resources:
+        limits:
+          gpu: "1"
+      extraPodSpec:
+        mainContainer:
+          image: my-registry/vllm-runtime:my-tag
+          workingDir: /workspace/components/backends/vllm
+          command:
+            - /bin/sh
+            - -c
+          args:
+          - |
+            export LD_LIBRARY_PATH=/usr/local/nvidia/lib64:$LD_LIBRARY_PATH
+            export PATH=$PATH:/usr/local/nvidia/bin:/usr/local/nvidia/lib64
+            /sbin/ldconfig
+            python3 -m dynamo.vllm --model Qwen/Qwen3-0.6B --is-prefill-worker