# vllm: 0.28.0 -> 0.33.0, for this configuration  (input read as: kubernetes)

severity  option                    what changes
--------  ------------------------  ----------------------------------------
INFO      --disable-prefix-caching  not an option of vllm 0.28.0, but exists in vllm 0.33.0

## Defaults that changed for options this configuration does not set (5)

option                         change
-----------------------------  ----------------------------------------
--allowed-origins              default ['*'] -> []
--cp-kv-cache-interleave-size  default 1 -> unset
--dcp-comm-backend             default 'ag_rs' -> unset
--disable-cascade-attn         default True -> unset; negations added: --enable-cascade-attn
--mamba-cache-mode             default 'none' -> unset; choices removed: all

## Suggested configuration for vllm 0.33.0

apiVersion: apps/v1
kind: Deployment
metadata:
  name: qwen3
spec:
  replicas: 1
  selector:
    matchLabels:
      app: qwen3
  template:
    metadata:
      labels:
        app: qwen3
    spec:
      containers:
      - name: vllm
        image: vllm/vllm-openai:v0.33.0
        args:
        - --model
        - Qwen/Qwen3-32B
        - --max-model-len
        - '32768'
        - --tensor-parallel-size
        - '2'
        - --gpu-memory-utilization
        - '0.90'
        - --disable-prefix-caching
        - --lora-modules
        - sql=/adapters/sql
        - support=/adapters/support
        - --enable-lora
        - --port
        - '8000'
        env:
        - name: HF_TOKEN
          valueFrom:
            secretKeyRef:
              name: hf
              key: token
        - name: VLLM_LOGGING_LEVEL
          value: INFO
        ports:
        - containerPort: 8000
        resources:
          limits:
            nvidia.com/gpu: 2
