chore: many bug fixes and improvements when testing planner (#2776)

Signed-off-by: hongkuanz <hongkuanz@nvidia.com>
Signed-off-by: hongkuan <hongkuanz@nvidia.com>

tests/planner/perf_test_configs/disagg_8b_2p2d.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: main
+      replicas: 1
+      livenessProbe:
+        httpGet:
+          path: /health
+          port: 8000
+        initialDelaySeconds: 20
+        periodSeconds: 5
+        timeoutSeconds: 5
+        failureThreshold: 3
+      readinessProbe:
+        exec:
+          command:
+            - /bin/sh
+            - -c
+            - 'curl -s http://localhost:8000/health | jq -e ".status == \"healthy\""'
+        initialDelaySeconds: 60
+        periodSeconds: 60
+        timeoutSeconds: 30
+        failureThreshold: 10
+      resources:
+        requests:
+          cpu: "16"
+          memory: "10Gi"
+        limits:
+          cpu: "128"
+          memory: "100Gi"
+      extraPodSpec:
+        mainContainer:
+          image: nvcr.io/nvidian/nim-llm-dev/vllm-runtime:hzhou-0902-01
+          workingDir: /workspace/components/backends/vllm
+          command:
+            - /bin/sh
+            - -c
+          args:
+            - "python3 -m dynamo.frontend --http-port 8000 --kv-cache-block-size 128"
+    VllmDecodeWorker:
+      dynamoNamespace: vllm-disagg
+      envFromSecret: hf-token-secret
+      componentType: worker
+      replicas: 2
+      livenessProbe:
+        httpGet:
+          path: /live
+          port: 9090
+        periodSeconds: 5
+        timeoutSeconds: 30
+        failureThreshold: 1
+      readinessProbe:
+        httpGet:
+          path: /health
+          port: 9090
+        periodSeconds: 10
+        timeoutSeconds: 30
+        failureThreshold: 60
+      resources:
+        requests:
+          cpu: "16"
+          memory: "20Gi"
+          gpu: "1"
+        limits:
+          cpu: "128"
+          memory: "100Gi"
+          gpu: "1"
+      envs:
+        - name: DYN_SYSTEM_ENABLED
+          value: "true"
+        - name: DYN_SYSTEM_USE_ENDPOINT_HEALTH_STATUS
+          value: "[\"generate\"]"
+        - name: DYN_SYSTEM_PORT
+          value: "9090"
+      extraPodSpec:
+        mainContainer:
+          startupProbe:
+            httpGet:
+              path: /health
+              port: 9090
+            periodSeconds: 10
+            failureThreshold: 60
+          image: nvcr.io/nvidian/nim-llm-dev/vllm-runtime:hzhou-0902-01
+          workingDir: /workspace/components/backends/vllm
+          command:
+            - /bin/sh
+            - -c
+          args:
+            - "python3 -m dynamo.vllm --model nvidia/Llama-3.1-8B-Instruct-FP8 --no-enable-prefix-caching --block-size 128 2>&1 | tee /tmp/vllm.log"
+    VllmPrefillWorker:
+      dynamoNamespace: vllm-disagg
+      envFromSecret: hf-token-secret
+      componentType: worker
+      replicas: 2
+      livenessProbe:
+        httpGet:
+          path: /live
+          port: 9090
+        periodSeconds: 5
+        timeoutSeconds: 30
+        failureThreshold: 1
+      readinessProbe:
+        httpGet:
+          path: /health
+          port: 9090
+        periodSeconds: 10
+        timeoutSeconds: 30
+        failureThreshold: 60
+      resources:
+        requests:
+          cpu: "16"
+          memory: "20Gi"
+          gpu: "1"
+        limits:
+          cpu: "128"
+          memory: "100Gi"
+          gpu: "1"
+      envs:
+        - name: DYN_SYSTEM_ENABLED
+          value: "true"
+        - name: DYN_SYSTEM_USE_ENDPOINT_HEALTH_STATUS
+          value: "[\"generate\"]"
+        - name: DYN_SYSTEM_PORT
+          value: "9090"
+      extraPodSpec:
+        mainContainer:
+          startupProbe:
+            httpGet:
+              path: /health
+              port: 9090
+            periodSeconds: 10
+            failureThreshold: 60
+          image: nvcr.io/nvidian/nim-llm-dev/vllm-runtime:hzhou-0902-01
+          workingDir: /workspace/components/backends/vllm
+          command:
+            - /bin/sh
+            - -c
+          args:
+            - "python3 -m dynamo.vllm --model nvidia/Llama-3.1-8B-Instruct-FP8 --is-prefill-worker --no-enable-prefix-caching --block-size 128 2>&1 | tee /tmp/vllm.log"

tests/planner/perf_test_configs/disagg_8b_3p1d.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: main
+      replicas: 1
+      livenessProbe:
+        httpGet:
+          path: /health
+          port: 8000
+        initialDelaySeconds: 20
+        periodSeconds: 5
+        timeoutSeconds: 5
+        failureThreshold: 3
+      readinessProbe:
+        exec:
+          command:
+            - /bin/sh
+            - -c
+            - 'curl -s http://localhost:8000/health | jq -e ".status == \"healthy\""'
+        initialDelaySeconds: 60
+        periodSeconds: 60
+        timeoutSeconds: 30
+        failureThreshold: 10
+      resources:
+        requests:
+          cpu: "16"
+          memory: "10Gi"
+        limits:
+          cpu: "128"
+          memory: "100Gi"
+      extraPodSpec:
+        mainContainer:
+          image: nvcr.io/nvidian/nim-llm-dev/vllm-runtime:hzhou-0902-01
+          workingDir: /workspace/components/backends/vllm
+          command:
+            - /bin/sh
+            - -c
+          args:
+            - "python3 -m dynamo.frontend --http-port 8000 --kv-cache-block-size 128"
+    VllmDecodeWorker:
+      dynamoNamespace: vllm-disagg
+      envFromSecret: hf-token-secret
+      componentType: worker
+      replicas: 1
+      livenessProbe:
+        httpGet:
+          path: /live
+          port: 9090
+        periodSeconds: 5
+        timeoutSeconds: 30
+        failureThreshold: 1
+      readinessProbe:
+        httpGet:
+          path: /health
+          port: 9090
+        periodSeconds: 10
+        timeoutSeconds: 30
+        failureThreshold: 60
+      resources:
+        requests:
+          cpu: "16"
+          memory: "20Gi"
+          gpu: "1"
+        limits:
+          cpu: "128"
+          memory: "100Gi"
+          gpu: "1"
+      envs:
+        - name: DYN_SYSTEM_ENABLED
+          value: "true"
+        - name: DYN_SYSTEM_USE_ENDPOINT_HEALTH_STATUS
+          value: "[\"generate\"]"
+        - name: DYN_SYSTEM_PORT
+          value: "9090"
+      extraPodSpec:
+        mainContainer:
+          startupProbe:
+            httpGet:
+              path: /health
+              port: 9090
+            periodSeconds: 10
+            failureThreshold: 60
+          image: nvcr.io/nvidian/nim-llm-dev/vllm-runtime:hzhou-0902-01
+          workingDir: /workspace/components/backends/vllm
+          command:
+            - /bin/sh
+            - -c
+          args:
+            - "python3 -m dynamo.vllm --model nvidia/Llama-3.1-8B-Instruct-FP8 --no-enable-prefix-caching --block-size 128 2>&1 | tee /tmp/vllm.log"
+    VllmPrefillWorker:
+      dynamoNamespace: vllm-disagg
+      envFromSecret: hf-token-secret
+      componentType: worker
+      replicas: 3
+      livenessProbe:
+        httpGet:
+          path: /live
+          port: 9090
+        periodSeconds: 5
+        timeoutSeconds: 30
+        failureThreshold: 1
+      readinessProbe:
+        httpGet:
+          path: /health
+          port: 9090
+        periodSeconds: 10
+        timeoutSeconds: 30
+        failureThreshold: 60
+      resources:
+        requests:
+          cpu: "16"
+          memory: "20Gi"
+          gpu: "1"
+        limits:
+          cpu: "128"
+          memory: "100Gi"
+          gpu: "1"
+      envs:
+        - name: DYN_SYSTEM_ENABLED
+          value: "true"
+        - name: DYN_SYSTEM_USE_ENDPOINT_HEALTH_STATUS
+          value: "[\"generate\"]"
+        - name: DYN_SYSTEM_PORT
+          value: "9090"
+      extraPodSpec:
+        mainContainer:
+          startupProbe:
+            httpGet:
+              path: /health
+              port: 9090
+            periodSeconds: 10
+            failureThreshold: 60
+          image: nvcr.io/nvidian/nim-llm-dev/vllm-runtime:hzhou-0902-01
+          workingDir: /workspace/components/backends/vllm
+          command:
+            - /bin/sh
+            - -c
+          args:
+            - "python3 -m dynamo.vllm --model nvidia/Llama-3.1-8B-Instruct-FP8 --is-prefill-worker --no-enable-prefix-caching --block-size 128 2>&1 | tee /tmp/vllm.log"

tests/planner/perf_test_configs/disagg_8b_planner.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-planner
+  annotations:
+    nvidia.com/enable-grove: "false"
+spec:
+  envs:
+    - name: DYNAMO_SERVICE_CONFIG
+      value: '{"Prometheus":{"global":{"scrape_interval":"5s"},"scrape_configs":[{"job_name":"prometheus","static_configs":[{"targets":["localhost:8000"]}]},{"job_name":"frontend","static_configs":[{"targets":["vllm-disagg-planner-frontend:8000"]}]}]}}'
+    - name: DYNAMO_NAMESPACE
+      value: "vllm-disagg-planner"
+  services:
+    Frontend:
+      dynamoNamespace: vllm-disagg-planner
+      componentType: main
+      replicas: 1
+      livenessProbe:
+        httpGet:
+          path: /health
+          port: 8000
+        initialDelaySeconds: 20
+        periodSeconds: 5
+        timeoutSeconds: 5
+        failureThreshold: 3
+      readinessProbe:
+        exec:
+          command:
+            - /bin/sh
+            - -c
+            - 'curl -s http://localhost:8000/health | jq -e ".status == \"healthy\""'
+        initialDelaySeconds: 60
+        periodSeconds: 60
+        timeoutSeconds: 30
+        failureThreshold: 10
+      resources:
+        requests:
+          cpu: "16"
+          memory: "10Gi"
+        limits:
+          cpu: "128"
+          memory: "100Gi"
+      extraPodSpec:
+        mainContainer:
+          image: nvcr.io/nvidian/nim-llm-dev/vllm-runtime:hzhou-0902-01
+          workingDir: /workspace/components/backends/vllm
+          command:
+            - /bin/sh
+            - -c
+          args:
+            - "python3 -m dynamo.frontend --http-port 8000 --kv-cache-block-size 128 --router-mode kv --kv-overlap-score-weight 0.0 --router-temperature 0.0 --no-kv-events"
+    Planner:
+      dynamoNamespace: vllm-disagg-planner
+      envFromSecret: hf-token-secret
+      componentType: planner
+      replicas: 1
+      envs:
+        - name: PROMETHEUS_PORT
+          value: "8000"
+      livenessProbe:
+        exec:
+          command:
+            - /bin/sh
+            - -c
+            - "exit 0"
+        periodSeconds: 60
+        timeoutSeconds: 30
+        failureThreshold: 10
+      readinessProbe:
+        exec:
+          command:
+            - /bin/sh
+            - -c
+            - "exit 0"
+        initialDelaySeconds: 60
+        periodSeconds: 60
+        timeoutSeconds: 30
+        failureThreshold: 10
+      extraPodSpec:
+        mainContainer:
+          image: nvcr.io/nvidian/nim-llm-dev/vllm-runtime:hzhou-0902-01
+          workingDir: /workspace/components/planner/src/dynamo/planner
+          ports:
+            - name: metrics
+              containerPort: 9085
+          command:
+            - /bin/sh
+            - -c
+          args:
+            - >-
+              python3 -m planner_sla
+              --environment=kubernetes
+              --backend=vllm
+              --ttft 0.2
+              --itl 0.01
+              --profile-results-dir /workspace/tests/planner/profiling_results/H200_TP1P_TP1D/
+              --adjustment-interval=60
+              --prometheus-port=9085
+              --no-correction
+    Prometheus:
+      dynamoNamespace: vllm-disagg-planner
+      componentType: main
+      replicas: 1
+      envs:
+        - name: PYTHONPATH
+          value: "/workspace/components/planner/src"
+        - name: PROMETHEUS_PORT
+          value: "8000"
+      livenessProbe:
+        exec:
+          command:
+            - /bin/sh
+            - -c
+            - "exit 0"
+        periodSeconds: 60
+        timeoutSeconds: 30
+        failureThreshold: 10
+      readinessProbe:
+        exec:
+          command:
+            - /bin/sh
+            - -c
+            - "exit 0"
+        initialDelaySeconds: 30
+        periodSeconds: 60
+        timeoutSeconds: 30
+        failureThreshold: 10
+      extraPodSpec:
+        mainContainer:
+          image: nvcr.io/nvidian/nim-llm-dev/vllm-runtime:hzhou-0902-01
+          workingDir: /workspace/components/backends/vllm
+          command:
+            - /bin/sh
+            - -c
+          args:
+            - "python3 -m dynamo.planner.prometheus"
+    VllmDecodeWorker:
+      dynamoNamespace: vllm-disagg-planner
+      envFromSecret: hf-token-secret
+      componentType: worker
+      replicas: 1
+      livenessProbe:
+        httpGet:
+          path: /live
+          port: 9090
+        periodSeconds: 5
+        timeoutSeconds: 30
+        failureThreshold: 1
+      readinessProbe:
+        httpGet:
+          path: /health
+          port: 9090
+        periodSeconds: 10
+        timeoutSeconds: 30
+        failureThreshold: 60
+      resources:
+        requests:
+          cpu: "16"
+          memory: "20Gi"
+          gpu: "1"
+        limits:
+          cpu: "128"
+          memory: "100Gi"
+          gpu: "1"
+      envs:
+        - name: DYN_SYSTEM_ENABLED
+          value: "true"
+        - name: DYN_SYSTEM_USE_ENDPOINT_HEALTH_STATUS
+          value: "[\"generate\"]"
+        - name: DYN_SYSTEM_PORT
+          value: "9090"
+      extraPodSpec:
+        terminationGracePeriodSeconds: 600
+        mainContainer:
+          startupProbe:
+            httpGet:
+              path: /health
+              port: 9090
+            periodSeconds: 10
+            failureThreshold: 60
+          image: nvcr.io/nvidian/nim-llm-dev/vllm-runtime:hzhou-0902-01
+          workingDir: /workspace/components/backends/vllm
+          command:
+            - python3
+          args:
+            - -m
+            - dynamo.vllm
+            - --model
+            - nvidia/Llama-3.1-8B-Instruct-FP8
+            - --no-enable-prefix-caching
+            - --block-size
+            - "128"
+    VllmPrefillWorker:
+      dynamoNamespace: vllm-disagg-planner
+      envFromSecret: hf-token-secret
+      componentType: worker
+      replicas: 1
+      livenessProbe:
+        httpGet:
+          path: /live
+          port: 9090
+        periodSeconds: 5
+        timeoutSeconds: 30
+        failureThreshold: 1
+      readinessProbe:
+        httpGet:
+          path: /health
+          port: 9090
+        periodSeconds: 10
+        timeoutSeconds: 30
+        failureThreshold: 60
+      resources:
+        requests:
+          cpu: "16"
+          memory: "20Gi"
+          gpu: "1"
+        limits:
+          cpu: "128"
+          memory: "100Gi"
+          gpu: "1"
+      envs:
+        - name: DYN_SYSTEM_ENABLED
+          value: "true"
+        - name: DYN_SYSTEM_USE_ENDPOINT_HEALTH_STATUS
+          value: "[\"generate\"]"
+        - name: DYN_SYSTEM_PORT
+          value: "9090"
+      extraPodSpec:
+        terminationGracePeriodSeconds: 600
+        mainContainer:
+          startupProbe:
+            httpGet:
+              path: /health
+              port: 9090
+            periodSeconds: 10
+            failureThreshold: 60
+          image: nvcr.io/nvidian/nim-llm-dev/vllm-runtime:hzhou-0902-01
+          workingDir: /workspace/components/backends/vllm
+          command:
+            - python3
+          args:
+            - -m
+            - dynamo.vllm
+            - --model
+            - nvidia/Llama-3.1-8B-Instruct-FP8
+            - --is-prefill-worker
+            - --no-enable-prefix-caching
+            - --block-size
+            - "128"

tests/planner/perf_test_configs/disagg_8b_tp2.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: main
+      replicas: 1
+      livenessProbe:
+        httpGet:
+          path: /health
+          port: 8000
+        initialDelaySeconds: 20
+        periodSeconds: 5
+        timeoutSeconds: 5
+        failureThreshold: 3
+      readinessProbe:
+        exec:
+          command:
+            - /bin/sh
+            - -c
+            - 'curl -s http://localhost:8000/health | jq -e ".status == \"healthy\""'
+        initialDelaySeconds: 60
+        periodSeconds: 60
+        timeoutSeconds: 30
+        failureThreshold: 10
+      resources:
+        requests:
+          cpu: "16"
+          memory: "10Gi"
+        limits:
+          cpu: "128"
+          memory: "100Gi"
+      extraPodSpec:
+        mainContainer:
+          image: nvcr.io/nvidian/nim-llm-dev/vllm-runtime:hzhou-0902-01
+          workingDir: /workspace/components/backends/vllm
+          command:
+            - /bin/sh
+            - -c
+          args:
+            - "python3 -m dynamo.frontend --http-port 8000 --kv-cache-block-size 128"
+    VllmDecodeWorker:
+      dynamoNamespace: vllm-disagg
+      envFromSecret: hf-token-secret
+      componentType: worker
+      replicas: 1
+      livenessProbe:
+        httpGet:
+          path: /live
+          port: 9090
+        periodSeconds: 5
+        timeoutSeconds: 30
+        failureThreshold: 1
+      readinessProbe:
+        httpGet:
+          path: /health
+          port: 9090
+        periodSeconds: 10
+        timeoutSeconds: 30
+        failureThreshold: 60
+      resources:
+        requests:
+          cpu: "16"
+          memory: "50Gi"
+          gpu: "2"
+        limits:
+          cpu: "128"
+          memory: "100Gi"
+          gpu: "2"
+      envs:
+        - name: DYN_SYSTEM_ENABLED
+          value: "true"
+        - name: DYN_SYSTEM_USE_ENDPOINT_HEALTH_STATUS
+          value: "[\"generate\"]"
+        - name: DYN_SYSTEM_PORT
+          value: "9090"
+      extraPodSpec:
+        mainContainer:
+          startupProbe:
+            httpGet:
+              path: /health
+              port: 9090
+            periodSeconds: 10
+            failureThreshold: 60
+          image: nvcr.io/nvidian/nim-llm-dev/vllm-runtime:hzhou-0902-01
+          workingDir: /workspace/components/backends/vllm
+          command:
+            - /bin/sh
+            - -c
+          args:
+            - "python3 -m dynamo.vllm --model nvidia/Llama-3.1-8B-Instruct-FP8 --no-enable-prefix-caching --block-size 128 --tensor-parallel-size 2 2>&1 | tee /tmp/vllm.log"
+    VllmPrefillWorker:
+      dynamoNamespace: vllm-disagg
+      envFromSecret: hf-token-secret
+      componentType: worker
+      replicas: 1
+      livenessProbe:
+        httpGet:
+          path: /live
+          port: 9090
+        periodSeconds: 5
+        timeoutSeconds: 30
+        failureThreshold: 1
+      readinessProbe:
+        httpGet:
+          path: /health
+          port: 9090
+        periodSeconds: 10
+        timeoutSeconds: 30
+        failureThreshold: 60
+      resources:
+        requests:
+          cpu: "16"
+          memory: "50Gi"
+          gpu: "2"
+        limits:
+          cpu: "128"
+          memory: "100Gi"
+          gpu: "2"
+      envs:
+        - name: DYN_SYSTEM_ENABLED
+          value: "true"
+        - name: DYN_SYSTEM_USE_ENDPOINT_HEALTH_STATUS
+          value: "[\"generate\"]"
+        - name: DYN_SYSTEM_PORT
+          value: "9090"
+      extraPodSpec:
+        mainContainer:
+          startupProbe:
+            httpGet:
+              path: /health
+              port: 9090
+            periodSeconds: 10
+            failureThreshold: 60
+          image: nvcr.io/nvidian/nim-llm-dev/vllm-runtime:hzhou-0902-01
+          workingDir: /workspace/components/backends/vllm
+          command:
+            - /bin/sh
+            - -c
+          args:
+            - "python3 -m dynamo.vllm --model nvidia/Llama-3.1-8B-Instruct-FP8 --is-prefill-worker --no-enable-prefix-caching --block-size 128 --tensor-parallel-size 2 2>&1 | tee /tmp/vllm.log"

tests/planner/perf_test_configs/image_cache_daemonset.yaml

+# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+
+apiVersion: apps/v1
+kind: DaemonSet
+metadata:
+  name: vllm-runtime-image-cache
+  labels:
+    app: vllm-runtime-image-cache
+spec:
+  selector:
+    matchLabels:
+      app: vllm-runtime-image-cache
+  template:
+    metadata:
+      labels:
+        app: vllm-runtime-image-cache
+    spec:
+      imagePullSecrets:
+      - name: nvcr-imagepullsecret
+      containers:
+      - name: image-cache
+        image: nvcr.io/nvidian/nim-llm-dev/vllm-runtime:hzhou-0902-01
+        command:
+        - /bin/sh
+        - -c
+        - "sleep infinity"
+        resources:
+          requests:
+            cpu: "10m"
+            memory: "64Mi"
+          limits:
+            cpu: "100m"
+            memory: "128Mi"
+      tolerations:
+      - key: node-role.kubernetes.io/master
+        operator: Exists
+        effect: NoSchedule
+      - key: node-role.kubernetes.io/control-plane
+        operator: Exists
+        effect: NoSchedule