本教程介绍如何使用 GKE Inference Gateway 在 Google Kubernetes Engine (GKE) 上部署大语言模型 (LLM)。本教程包含集群设置、模型部署、GKE Inference Gateway 配置和处理 LLM 请求的步骤。
本教程适用于机器学习 (ML) 工程师、平台管理员和运维人员，以及希望使用 GKE 推理网关在 GKE 上部署和管理 LLM 应用的数据和 AI 专家。
在阅读本页面内容之前，请先熟悉以下内容：
背景
本部分介绍本教程中使用的关键技术。 如需详细了解模型服务概念和术语，以及 GKE 生成式 AI 功能如何提升和支持模型部署性能，请参阅 GKE 上的模型推理简介。
vLLM
vLLM 是一种经过高度优化的开源 LLM 服务框架，可提高 GPU 上的服务吞吐量。主要功能包括：
- 具有 PagedAttention 且经过优化的 Transformer（转换器）实现
- 连续批处理，可提高整体服务吞吐量
- 多个 GPU 上的张量并行处理和分布式服务
如需了解详情，请参阅 vLLM 文档。
GKE 推理网关
GKE Inference Gateway 增强了 GKE 在部署大语言模型方面的功能。它通过以下功能优化推理工作负载：
- 基于负载指标的推理优化负载均衡。
- 支持 LoRA 适配器的密集型多工作负载部署。
- 模型感知型路由，可简化操作。
如需了解详情，请参阅 GKE 推理网关简介。
目标
准备工作
-
In the Google Cloud console, on the project selector page, select or create a Google Cloud project.
Roles required to select or create a project
- Select a project: Selecting a project doesn't require a specific IAM role—you can select any project that you've been granted a role on.
-
Create a project: To create a project, you need the Project Creator role
(
roles/resourcemanager.projectCreator), which contains theresourcemanager.projects.createpermission. Learn how to grant roles.
-
Verify that billing is enabled for your Google Cloud project.
Enable the required API.
Roles required to enable APIs
To enable APIs, you need the
serviceusage.services.enablepermission. If you created the project, then you likely already have this permission through the Owner role (roles/owner). Otherwise, you can get this permission through the Service Usage Admin role (roles/serviceusage.serviceUsageAdmin). Learn how to grant roles.-
确保您对项目拥有以下角色： roles/container.admin、roles/iam.serviceAccountAdmin
检查角色
-
在 Google Cloud 控制台中，前往 IAM 页面。
转到 IAM - 选择项目。
-
在正文 列中，找到标识您或您所属的所有组的所有行。如需了解您所属的组，请与您的 管理员联系。
- 对于指定或包含您的所有行，请检查角色 列，以查看角色列表是否包含所需角色。
授予角色
-
在 Google Cloud 控制台中，前往 IAM 页面。
转到 IAM - 选择项目。
- 点击 授予访问权限。
-
在新正文 字段中，输入您的用户标识符。 这通常是员工身份池中的用户的标识符。如需了解详情， 请参阅 在 IAM 政策中表示员工池用户，或与您的管理员联系。
- 点击选择角色，然后搜索该角色。
- 如需授予其他角色，请点击 添加其他角色 ，然后添加其他各个角色。
- 点击 Save （保存）。
-
- 如果您还没有 Hugging Face 账号，请创建一个。
- 确保您的项目具有足够的 H100 GPU 配额。如需了解详情，请参阅规划 GPU 配额和分配配额。
获取对模型的访问权限
如需将 Llama3.1 模型部署到 GKE，请签署许可同意协议并生成 Hugging Face 访问令牌。
签署许可同意协议
您必须签署同意协议才能使用 Llama3.1 模型。请按照以下说明操作：
- 访问同意页面，并验证您是否同意使用您的 Hugging Face 账号。
- 接受模型条款。
生成一个访问令牌
如需通过 Hugging Face 访问模型，您需要 Hugging Face 令牌。
如果您还没有令牌，请按照以下步骤生成新令牌：
- 点击您的个人资料 > 设置 > 访问令牌。
- 选择新建令牌 (New Token)。
- 指定您选择的名称和一个至少为
Read的角色。 - 选择生成令牌。
- 将生成的令牌复制到剪贴板。
准备环境
在本教程中，您将使用 Cloud Shell 来管理Google Cloud上托管的资源。Cloud Shell 中预安装了本教程所需的软件，包括 kubectl 和 gcloud CLI。
如需使用 Cloud Shell 设置您的环境，请执行以下步骤：
在 Google Cloud 控制台中，点击 Google Cloud 控制台中的
激活 Cloud Shell 以启动 Cloud Shell 会话。此操作会在 Google Cloud 控制台的底部窗格中启动会话。设置默认环境变量：
gcloud config set project PROJECT_ID gcloud config set billing/quota_project PROJECT_ID export PROJECT_ID=$(gcloud config get project) export REGION=REGION export CLUSTER_NAME=CLUSTER_NAME export HF_TOKEN=HF_TOKEN替换以下值：
PROJECT_ID：您的 Google Cloud项目 ID。REGION：支持要使用的加速器类型的区域，例如适用于 H100 GPU 的us-central1。CLUSTER_NAME：您的集群的名称。HF_TOKEN：您之前生成的 Hugging Face 令牌。
创建和配置 Google Cloud 资源
创建 GKE 集群和节点池
在 GKE Autopilot 或 Standard 集群中的 GPU 上部署 LLM。我们建议您使用 Autopilot 集群获得全托管式 Kubernetes 体验。如需选择最适合您的工作负载的 GKE 操作模式，请参阅选择 GKE 操作模式。
Autopilot
在 Cloud Shell 中，运行以下命令：
gcloud container clusters create-auto CLUSTER_NAME \
--project=PROJECT_ID \
--location=CONTROL_PLANE_LOCATION \
--release-channel=rapid
替换以下值：
PROJECT_ID：您的 Google Cloud项目 ID。CONTROL_PLANE_LOCATION：集群控制平面的 Compute Engine 区域。提供支持要使用的加速器类型的区域，例如适用于 H100 GPU 的us-central1。CLUSTER_NAME：您的集群的名称。
GKE 会根据所部署的工作负载的请求，创建具有所需 CPU 和 GPU 节点的 Autopilot 集群。
Standard
在 Cloud Shell 中，运行以下命令以创建 Standard 集群：
gcloud container clusters create CLUSTER_NAME \ --project=PROJECT_ID \ --location=CONTROL_PLANE_LOCATION \ --workload-pool=PROJECT_ID. \ --release-channel=rapid \ --num-nodes=1 \ --enable-managed-prometheus \ --monitoring=SYSTEM,DCGM \ --gateway-api=standard替换以下值：
PROJECT_ID：您的 Google Cloud项目 ID。CONTROL_PLANE_LOCATION：集群控制平面的 Compute Engine 区域。提供支持要使用的加速器类型的区域，例如适用于 H100 GPU 的us-central1。CLUSTER_NAME：您的集群的名称。
集群创建可能需要几分钟的时间。
如需创建具有适当磁盘大小的 节点池 以运行
Llama-3.1-8B-Instruct模型，请运行以下命令：gcloud container node-pools create gpupool \ --accelerator type=nvidia-h100-80gb,count=2,gpu-driver-version=latest \ --project=PROJECT_ID \ --location=CONTROL_PLANE_LOCATION \ --node-locations=CONTROL_PLANE_LOCATION-a \ --cluster=CLUSTER_NAME \ --machine-type=a3-highgpu-2g \ --num-nodes=1 \ --disk-type="pd-balanced"GKE 会创建一个节点池，其中包含一个 H100 GPU。
设置授权以爬取指标
如需设置授权以爬取指标，请创建 inference-gateway-sa-metrics-reader-secret Secret。
将以下清单保存为
metrics-auth.yaml：应用清单：
kubectl apply -f metrics-auth.yaml
为 Hugging Face 凭据创建 Kubernetes Secret
在 Cloud Shell 中，执行以下操作：
配置
kubectl，使其能够与您的集群通信：gcloud container clusters get-credentials CLUSTER_NAME \ --location=REGION替换以下值：
REGION：支持要使用的加速器类型的区域，例如适用于 L4 GPU 的us-central1。CLUSTER_NAME：您的集群的名称。
创建包含 Hugging Face 令牌的 Kubernetes Secret：
kubectl create secret generic hf-secret \ --from-literal=hf_api_token=${HF_TOKEN} \ --dry-run=client -o yaml | kubectl apply -f -将
HF_TOKEN替换为您之前生成的 Hugging Face 令牌。
安装 InferenceObjective 和 InferencePool CRD
在本部分中，您将安装 GKE Inference Gateway 所需的自定义资源定义 (CRD)。
CRD 扩展了 Kubernetes API。这样一来，您就可以定义新的资源类型。如需使用 GKE Inference Gateway，请在 GKE 集群中安装 InferencePool 和 InferenceObjective CRD，方法是运行以下命令：
kubectl apply -f https://github.com/kubernetes-sigs/gateway-api-inference-extension/releases/download/v1.0.0/manifests.yaml
部署模型服务器
此示例使用 vLLM 模型服务器部署 Llama3.1 模型。相应部署将被标记为 app:vllm-llama3.1-8b-instruct。此部署还使用了 Hugging Face 中的两个名为 food-review 和 cad-fabricator 的 LoRA 适配器。您可以使用自己的模型服务器和模型容器、服务端口和部署名称更新此部署。您可以选择在部署中配置 LoRA 适配器，或部署基础模型。
如需部署在
nvidia-h100-80gb加速器类型上，请将以下清单保存为vllm-llama3.1-8b-instruct.yaml。此清单定义了一个包含模型和模型服务器的 Kubernetes Deployment：apiVersion: apps/v1 kind: Deployment metadata: name: vllm-llama3.1-8b-instruct spec: replicas: 3 selector: matchLabels: app: vllm-llama3.1-8b-instruct template: metadata: labels: app: vllm-llama3.1-8b-instruct spec: containers: - name: vllm image: "vllm/vllm-openai:v0.13.0" imagePullPolicy: Always command: ["python3", "-m", "vllm.entrypoints.openai.api_server"] args: - "--model" - "meta-llama/Llama-3.1-8B-Instruct" - "--tensor-parallel-size" - "1" - "--port" - "8000" - "--enable-lora" - "--max-loras" - "2" - "--max-cpu-loras" - "12" - "--compilation-config" - '{"cudagraph_specialize_lora": "False"}' # As workaround for https://github.com/vllm-project/vllm/issues/29049 env: - name: PORT value: "8000" - name: HUGGING_FACE_HUB_TOKEN valueFrom: secretKeyRef: name: hf-secret key: hf_api_token - name: VLLM_ALLOW_RUNTIME_LORA_UPDATING value: "true" ports: - containerPort: 8000 name: http protocol: TCP lifecycle: preStop: # vLLM stops accepting connections when it receives SIGTERM, so we need to sleep # to give upstream gateways a chance to take us out of rotation. The time we wait # is dependent on the time it takes for all upstreams to completely remove us from # rotation. Older or simpler load balancers might take upwards of 30s, but we expect # our deployment to run behind a modern gateway like Envoy which is designed to # probe for readiness aggressively. sleep: # Upstream gateway probers for health should be set on a low period, such as 5s, # and the shorter we can tighten that bound the faster that we release # accelerators during controlled shutdowns. However, we should expect variance, # as load balancers may have internal delays, and we don't want to drop requests # normally, so we're often aiming to set this value to a p99 propagation latency # of readiness -> load balancer taking backend out of rotation, not the average. # # This value is generally stable and must often be experimentally determined on # for a given load balancer and health check period. We set the value here to # the highest value we observe on a supported load balancer, and we recommend # tuning this value down and verifying no requests are dropped. # # If this value is updated, be sure to update terminationGracePeriodSeconds. # seconds: 30 # # IMPORTANT: preStop.sleep is beta as of Kubernetes 1.30 - for older versions # replace with this exec action. #exec: # command: # - /usr/bin/sleep # - 30 livenessProbe: httpGet: path: /health port: http scheme: HTTP # vLLM's health check is simple, so we can more aggressively probe it. Liveness # check endpoints should always be suitable for aggressive probing. periodSeconds: 1 successThreshold: 1 # vLLM has a very simple health implementation, which means that any failure is # likely significant. However, any liveness triggered restart requires the very # large core model to be reloaded, and so we should bias towards ensuring the # server is definitely unhealthy vs immediately restarting. Use 5 attempts as # evidence of a serious problem. failureThreshold: 5 timeoutSeconds: 1 readinessProbe: httpGet: path: /health port: http scheme: HTTP # vLLM's health check is simple, so we can more aggressively probe it. Readiness # check endpoints should always be suitable for aggressive probing, but may be # slightly more expensive than readiness probes. periodSeconds: 1 successThreshold: 1 # vLLM has a very simple health implementation, which means that any failure is # likely significant, failureThreshold: 1 timeoutSeconds: 1 # We set a startup probe so that we don't begin directing traffic or checking # liveness to this instance until the model is loaded. startupProbe: # Failure threshold is when we believe startup will not happen at all, and is set # to the maximum possible time we believe loading a model will take. In our # default configuration we are downloading a model from HuggingFace, which may # take a long time, then the model must load into the accelerator. We choose # 10 minutes as a reasonable maximum startup time before giving up and attempting # to restart the pod. # # IMPORTANT: If the core model takes more than 10 minutes to load, pods will crash # loop forever. Be sure to set this appropriately. failureThreshold: 600 # Set delay to start low so that if the base model changes to something smaller # or an optimization is deployed, we don't wait unnecessarily. initialDelaySeconds: 2 # As a startup probe, this stops running and so we can more aggressively probe # even a moderately complex startup - this is a very important workload. periodSeconds: 1 httpGet: # vLLM does not start the OpenAI server (and hence make /health available) # until models are loaded. This may not be true for all model servers. path: /health port: http scheme: HTTP resources: limits: nvidia.com/gpu: 1 requests: nvidia.com/gpu: 1 volumeMounts: - mountPath: /data name: data - mountPath: /dev/shm name: shm - name: adapters mountPath: "/adapters" # This is the second container in the Pod, a sidecar to the vLLM container. # It watches the ConfigMap and downloads LoRA adapters. - name: lora-adapter-syncer image: us-central1-docker.pkg.dev/k8s-staging-images/gateway-api-inference-extension/lora-syncer:main imagePullPolicy: Always env: - name: DYNAMIC_LORA_ROLLOUT_CONFIG value: "/config/configmap.yaml" volumeMounts: # DO NOT USE subPath, dynamic configmap updates don't work on subPaths - name: config-volume mountPath: /config restartPolicy: Always # vLLM allows VLLM_PORT to be specified as an environment variable, but a user might # create a 'vllm' service in their namespace. That auto-injects VLLM_PORT in docker # compatible form as `tcp://<IP>:<PORT>` instead of the numeric value vLLM accepts # causing CrashLoopBackoff. Set service environment injection off by default. enableServiceLinks: false # Generally, the termination grace period needs to last longer than the slowest request # we expect to serve plus any extra time spent waiting for load balancers to take the # model server out of rotation. # # An easy starting point is the p99 or max request latency measured for your workload, # although LLM request latencies vary significantly if clients send longer inputs or # trigger longer outputs. Since steady state p99 will be higher than the latency # to drain a server, you may wish to slightly this value either experimentally or # via the calculation below. # # For most models you can derive an upper bound for the maximum drain latency as # follows: # # 1. Identify the maximum context length the model was trained on, or the maximum # allowed length of output tokens configured on vLLM (llama2-7b was trained to # 4k context length, while llama3-8b was trained to 128k). # 2. Output tokens are the more compute intensive to calculate and the accelerator # will have a maximum concurrency (batch size) - the time per output token at # maximum batch with no prompt tokens being processed is the slowest an output # token can be generated (for this model it would be about 10ms TPOT at a max # batch size around 50, or 100 tokens/sec) # 3. Calculate the worst case request duration if a request starts immediately # before the server stops accepting new connections - generally when it receives # SIGTERM (for this model that is about 4096 / 100 ~ 40s) # 4. If there are any requests generating prompt tokens that will delay when those # output tokens start, and prompt token generation is roughly 6x faster than # compute-bound output token generation, so add 40% to the time from above (40s + # 16s = 56s) # # Thus we think it will take us at worst about 56s to complete the longest possible # request the model is likely to receive at maximum concurrency (highest latency) # once requests stop being sent. # # NOTE: This number will be lower than steady state p99 latency since we stop receiving # new requests which require continuous prompt token computation. # NOTE: The max timeout for backend connections from gateway to model servers should # be configured based on steady state p99 latency, not drain p99 latency # # 5. Add the time the pod takes in its preStop hook to allow the load balancers to # stop sending us new requests (56s + 30s = 86s). # # Because the termination grace period controls when the Kubelet forcibly terminates a # stuck or hung process (a possibility due to a GPU crash), there is operational safety # in keeping the value roughly proportional to the time to finish serving. There is also # value in adding a bit of extra time to deal with unexpectedly long workloads. # # 6. Add a 50% safety buffer to this time (86s * 1.5 ≈ 130s). # # One additional source of drain latency is that some workloads may run close to # saturation and have queued requests on each server. Since traffic in excess of the # max sustainable QPS will result in timeouts as the queues grow, we assume that failure # to drain in time due to excess queues at the time of shutdown is an expected failure # mode of server overload. If your workload occasionally experiences high queue depths # due to periodic traffic, consider increasing the safety margin above to account for # time to drain queued requests. terminationGracePeriodSeconds: 130 nodeSelector: cloud.google.com/gke-accelerator: "nvidia-h100-80gb" volumes: - name: data emptyDir: {} - name: shm emptyDir: medium: Memory - name: adapters emptyDir: {} - name: config-volume configMap: name: vllm-llama3.1-8b-adapters --- apiVersion: v1 kind: ConfigMap metadata: name: vllm-llama3.1-8b-adapters data: configmap.yaml: | vLLMLoRAConfig: name: vllm-llama3.1-8b-instruct port: 8000 defaultBaseModel: meta-llama/Llama-3.1-8B-Instruct ensureExist: models: - id: food-review source: Kawon/llama3.1-food-finetune_v14_r8 - id: cad-fabricator source: redcathode/fabricator --- kind: HealthCheckPolicy apiVersion: networking.gke.io/v1 metadata: name: health-check-policy namespace: default spec: targetRef: group: "inference.networking.k8s.io" kind: InferencePool name: vllm-llama3.1-8b-instruct default: config: type: HTTP httpHealthCheck: requestPath: /health port: 8000将清单应用到您的集群：
kubectl apply -f vllm-llama3.1-8b-instruct.yaml
创建 InferencePool 资源
InferencePool Kubernetes 自定义资源定义了一组具有通用基础 LLM 和计算配置的 Pod。
InferencePool 自定义资源包含以下关键字段：
selector：指定哪些 Pod 属于此池。此选择器中的标签必须与应用于模型服务器 Pod 的标签完全一致。targetPort：定义 Pod 中模型服务器使用的端口。
InferencePool 资源使 GKE Inference Gateway 能够将流量路由到模型服务器 Pod。
如需使用 Helm 创建 InferencePool，请执行以下步骤：
helm install vllm-llama3.1-8b-instruct \
--set inferencePool.modelServers.matchLabels.app=vllm-llama3.1-8b-instruct \
--set provider.name=gke \
--set healthCheckPolicy.create=false \
--version v1.0.0 \
oci://registry.k8s.io/gateway-api-inference-extension/charts/inferencepool
更改以下字段以与您的 Deployment 相符：
inferencePool.modelServers.matchLabels.app：用于选择模型服务器 Pod 的标签的键。
此命令会创建一个 InferencePool 对象，该对象在逻辑上表示您的模型服务器部署，并引用 Selector 选择的 Pod 中的模型端点服务。
创建具有服务重要性的 InferenceObjective 资源
InferenceObjective 自定义资源定义了模型的服务参数，包括其优先级。您必须创建 InferenceObjective 资源来定义在 InferencePool 上部署哪些模型。这些资源可以引用 InferencePool 中模型服务器支持的基础模型或 LoRA 适配器。
metadata.name 字段指定模型的名称，priority 字段设置其服务重要性，poolRef 字段链接到部署模型所处的 InferencePool。
如需创建 InferenceObjective，请执行以下步骤：
将以下示例清单保存为
inferenceobjective.yaml：apiVersion: inference.networking.x-k8s.io/v1alpha2 kind: InferenceObjective metadata: name: MODEL_NAME spec: priority: VALUE poolRef: name: INFERENCE_POOL_NAME kind: "InferencePool"替换以下内容：
MODEL_NAME：基础模型或 LoRA 适配器的名称。例如food-review。VALUE：推理目标的优先级。这是一个整数，值越大表示请求越重要。例如10。INFERENCE_POOL_NAME：您在上一步中创建的InferencePool的名称。例如vllm-llama3.1-8b-instruct。
将示例清单应用于集群：
kubectl apply -f inferenceobjective.yaml
以下示例会创建两个 InferenceObjective 对象。第一个对象在 vllm-llama3.1-8b-instruct
InferencePool 上配置 food-review LoRA 模型，优先级为 10。第二个对象将 llama3-base-model 配置为以更高的优先级部署，为 20。
apiVersion: inference.networking.k8s.io/v1alpha1
kind: InferenceObjective
metadata:
name: food-review
spec:
priority: 10
poolRef:
name: vllm-llama3.1-8b-instruct
kind: "InferencePool"
---
apiVersion: inference.networking.k8s.io/v1alpha1
kind: InferenceObjective
metadata:
name: llama3-base-model
spec:
priority: 20
poolRef:
name: vllm-llama3.1-8b-instruct
kind: "InferencePool"
创建网关
网关资源充当外部流量进入 Kubernetes 集群的入口点。它定义用于接受传入连接的监听器。
GKE Inference Gateway 支持 gke-l7-rilb 和 gke-l7-regional-external-managed 网关类。如需了解详情，请参阅
GKE 文档中关于网关
类的内容。
如需创建网关，请执行以下步骤：
将以下示例清单保存为
gateway.yaml：apiVersion: gateway.networking.k8s.io/v1 kind: Gateway metadata: name: GATEWAY_NAME spec: gatewayClassName: gke-l7-regional-external-managed listeners: - protocol: HTTP # Or HTTPS for production port: 80 # Or 443 for HTTPS name: http将
GATEWAY_NAME替换为网关资源的唯一名称。例如inference-gateway。将清单应用到您的集群：
kubectl apply -f gateway.yaml
创建 HTTPRoute 资源
在本部分中，您将创建一个 HTTPRoute 资源，以定义网关如何将传入的 HTTP 请求路由到 InferencePool。
HTTPRoute 资源定义 GKE 网关如何将传入的 HTTP 请求路由到后端服务，即 InferencePool。它指定匹配规则（例如，标头或路径）以及应将流量转发到的后端。
如需创建 HTTPRoute，请执行以下步骤：
将以下示例清单保存为
httproute.yaml：apiVersion: gateway.networking.k8s.io/v1 kind: HTTPRoute metadata: name: HTTPROUTE_NAME spec: parentRefs: - name: GATEWAY_NAME rules: - matches: - path: type: PathPrefix value: PATH_PREFIX backendRefs: - name: INFERENCE_POOL_NAME group: inference.networking.k8s.io kind: InferencePool替换以下内容：
HTTPROUTE_NAME：HTTPRoute资源的唯一名称。例如my-route。GATEWAY_NAME：您创建的Gateway资源的名称。例如inference-gateway。PATH_PREFIX：用于匹配传入请求的路径前缀。例如，/可匹配所有路径。INFERENCE_POOL_NAME：要将流量路由到的InferencePool资源的名称。例如vllm-llama3.1-8b-instruct。
将清单应用到您的集群：
kubectl apply -f httproute.yaml
发送推理请求
配置 GKE 推理网关后，您便可以向已部署的模型发送推理请求。
如需发送推理请求，请执行以下步骤：
- 检索网关端点。
- 构建格式正确的 JSON 请求。
- 使用
curl向/v1/completions端点发送请求。
这样一来，您就可以根据输入提示和指定参数生成文本。
如需获取网关端点，请运行以下命令：
IP=$(kubectl get gateway/GATEWAY_NAME -o jsonpath='{.status.addresses[0].value}') PORT=80将
GATEWAY_NAME替换为您的网关资源名称。如需使用
curl向/v1/completions端点发送请求，请运行以下命令：curl -i -X POST http://${IP}:${PORT}/v1/completions \ -H "Content-Type: application/json" \ -d '{ "model": "MODEL_NAME", "prompt": "PROMPT_TEXT", "max_tokens": MAX_TOKENS, "temperature": "TEMPERATURE" }'替换以下内容：
MODEL_NAME：要使用的模型或 LoRA 适配器的名称。PROMPT_TEXT：模型的输入提示。MAX_TOKENS：回答中可生成的 token 数量上限。TEMPERATURE：控制输出的随机性。使用值0可获得确定性输出，使用更高的值则可获得更具创造性的输出。
请注意以下几点：
- 请求正文：请求正文可以包含其他参数，例如
stop和top_p。如需查看完整的选项列表，请参阅 OpenAI API 规范。 - 错误处理：在客户端代码中实现适当的错误处理，以处理响应中可能出现的错误。例如，检查
curl响应中的 HTTP 状态代码。非 200 状态代码通常表示错误。 - 身份验证和授权：对于生产部署，请使用身份验证和授权机制保护您的 API 端点。在请求中添加相应的标头（例如
Authorization）。
为推理网关配置可观测性
GKE Inference Gateway 可提供对推理工作负载的健康状况、性能和行为的可观测性。这有助于您发现和解决问题、优化资源利用率，并确保应用的可靠性。您可以通过 Metrics Explorer 在 Cloud Monitoring 中查看这些可观测性指标。
如需为 GKE 推理网关配置可观测性，请参阅配置 可观测性。
删除已部署的资源
为避免因您在本指南中创建的资源导致您的 Google Cloud 账号产生费用，请运行以下命令：
gcloud container clusters delete CLUSTER_NAME \
--location=CONTROL_PLANE_LOCATION
替换以下值：
CONTROL_PLANE_LOCATION：集群控制平面的 Compute Engine 区域。CLUSTER_NAME：您的集群的名称。