disagg_planner.yaml

# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
# SPDX-License-Identifier: Apache-2.0
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
apiVersion: nvidia.com/v1alpha1
kind: DynamoGraphDeployment
metadata:
  name: disagg-planner
spec:
  envs:
    - name: DYN_DEPLOYMENT_CONFIG
      value: '{"Common":{"model":"deepseek-ai/DeepSeek-R1-Distill-Llama-8B","block-size":64,"max-model-len":16384,"kv-transfer-config":"{\"kv_connector\":\"DynamoNixlConnector\"}"},"Frontend":{"served_model_name":"deepseek-ai/DeepSeek-R1-Distill-Llama-8B","endpoint":"dynamo.VllmWorker.generate","port":8000,"router":"round-robin","common-configs":["block-size"]},"VllmWorker":{"remote-prefill":true,"conditional-disagg":true,"max-local-prefill-length":10,"max-prefill-queue-size":2,"enable-prefix-caching":true,"common-configs":["model","block-size","max-model-len","kv-transfer-config"]},"PrefillWorker":{"max-num-batched-tokens":16384,"common-configs":["model","block-size","max-model-len","kv-transfer-config"]},"Prometheus":{"global":{"scrape_interval":"5s"},"scrape_configs":[{"job_name":"prometheus","static_configs":[{"targets":["localhost:9090"]}]},{"job_name":"frontend","static_configs":[{"targets":["localhost:8000"]}]}]},"Planner":{"adjustment-interval":180,"profile-results-dir":"/workspace/examples/profiling_results","isl":3000,"osl":150,"ttft":0.5,"itl":0.05,"load-predictor":"arima"}}'
  services:
    Frontend:
      dynamoNamespace: vllm-v0-disagg-planner
      componentType: main
      replicas: 1
      resources:
        requests:
          cpu: "2"
          memory: "4Gi"
        limits:
          cpu: "2"
          memory: "4Gi"
      extraPodSpec:
        mainContainer:
          image: nvcr.io/nvidia/ai-dynamo/vllm-runtime:latest
          workingDir: /workspace/examples/vllm_v0
          args:
            - dynamo
            - serve
            - graphs.disagg_planner:Frontend
            - --system-app-port
            - "5000"
            - --enable-system-app
            - --use-default-health-checks
            - --service-name
            - Frontend

    VllmWorker:
      dynamoNamespace: vllm-v0-disagg-planner
      envFromSecret: hf-token-secret
      replicas: 1
      resources:
        requests:
          cpu: "20"
          memory: "40Gi"
          gpu: "2"
        limits:
          cpu: "20"
          memory: "40Gi"
          gpu: "2"
      extraPodSpec:
        mainContainer:
          image: nvcr.io/nvidia/ai-dynamo/vllm-runtime:latest
          workingDir: /workspace/examples/vllm_v0
          args:
            - dynamo
            - serve
            - graphs.disagg_planner:VllmWorker
            - --system-app-port
            - "5000"
            - --enable-system-app
            - --use-default-health-checks
            - --service-name
            - VllmWorker

    PrefillWorker:
      dynamoNamespace: vllm-v0-disagg-planner
      envFromSecret: hf-token-secret
      replicas: 1
      resources:
        requests:
          cpu: "20"
          memory: "40Gi"
          gpu: "2"
        limits:
          cpu: "20"
          memory: "40Gi"
          gpu: "2"
      extraPodSpec:
        mainContainer:
          image: nvcr.io/nvidia/ai-dynamo/vllm-runtime:latest
          workingDir: /workspace/examples/vllm_v0
          args:
            - dynamo
            - serve
            - graphs.disagg_planner:PrefillWorker
            - --system-app-port
            - "5000"
            - --enable-system-app
            - --use-default-health-checks
            - --service-name
            - PrefillWorker

    Planner:
      dynamoNamespace: vllm-v0-disagg-planner
      replicas: 1
      componentType: planner
      resources:
        requests:
          cpu: "2"
          memory: "2Gi"
        limits:
          cpu: "2"
          memory: "2Gi"
      extraPodSpec:
        mainContainer:
          image: nvcr.io/nvidia/ai-dynamo/vllm-runtime:latest
          workingDir: /workspace/examples/vllm_v0
          args:
            - dynamo
            - serve
            - graphs.disagg_planner:Planner
            - --system-app-port
            - "5000"
            - --enable-system-app
            - --use-default-health-checks
            - --service-name
            - Planner
            - --Planner.environment=kubernetes

    Prometheus:
      dynamoNamespace: vllm-v0-disagg-planner
      replicas: 1
      resources:
        requests:
          cpu: "1000m"
          memory: "1000Mi"
        limits:
          cpu: "1000m"
          memory: "1000Mi"
      extraPodSpec:
        mainContainer:
          image: nvcr.io/nvidia/ai-dynamo/vllm-runtime:latest
          workingDir: /workspace/examples/vllm_v0
          args:
            - dynamo
            - serve
            - graphs.disagg_planner:Prometheus
            - --system-app-port
            - "5000"
            - --enable-system-app
            - --use-default-health-checks
            - --service-name
            - Prometheus