Update strategy for Deployments DeploymentUpdate: type: RollingUpdate maxUnavailable: 0 maxSurge: 100% ## Update strategy for StatefulSets StatefulSetUpdate: updateStrategy: RollingUpdate ## Pod priority settings ## ref: https://kubernetes.io/docs/concepts/configuration/pod-priority-preemption/ ## priorityClassName: "" ## Set default rootUser, rootPassword ## AccessKey and secretKey is generated when not set ## Distributed MinIO ref: https://docs.minio.io/docs/distributed-minio-quickstart-guide ##...

Update strategy for Deployments DeploymentUpdate: type: RollingUpdate maxUnavailable: 0 maxSurge: 100% ## Update strategy for StatefulSets StatefulSetUpdate: updateStrategy: RollingUpdate ## Pod priority settings ## ref: https://kubernetes.io/docs/concepts/configuration/pod-priority-preemption/ ## priorityClassName: "" ## Set default rootUser, rootPassword ## AccessKey and secretKey is generated when not set ## Distributed MinIO ref: https://docs.minio.io/docs/distributed-minio-quickstart-guide ##...

Update strategy for Deployments DeploymentUpdate: type: RollingUpdate maxUnavailable: 0 maxSurge: 100% ## Update strategy for StatefulSets StatefulSetUpdate: updateStrategy: RollingUpdate ## Pod priority settings ## ref: https://kubernetes.io/docs/concepts/configuration/pod-priority-preemption/ ## priorityClassName: "" ## Set default rootUser, rootPassword ## AccessKey and secretKey is generated when not set ## Distributed MinIO ref: https://docs.minio.io/docs/distributed-minio-quickstart-guide ##...

Update strategy for Deployments DeploymentUpdate: type: RollingUpdate maxUnavailable: 0 maxSurge: 100% ## Update strategy for StatefulSets StatefulSetUpdate: updateStrategy: RollingUpdate ## Pod priority settings ## ref: https://kubernetes.io/docs/concepts/configuration/pod-priority-preemption/ ## priorityClassName: "" ## Set default rootUser, rootPassword ## AccessKey and secretKey is generated when not set ## Distributed MinIO ref: https://docs.minio.io/docs/distributed-minio-quickstart-guide ##...

Update strategy for Deployments DeploymentUpdate: type: RollingUpdate maxUnavailable: 0 maxSurge: 100% ## Update strategy for StatefulSets StatefulSetUpdate: updateStrategy: RollingUpdate ## Pod priority settings ## ref: https://kubernetes.io/docs/concepts/configuration/pod-priority-preemption/ ## priorityClassName: "" ## Set default rootUser, rootPassword ## AccessKey and secretKey is generated when not set ## Distributed MinIO ref: https://docs.minio.io/docs/distributed-minio-quickstart-guide ##...

让我们回顾一下之前的部署概念：

* 安全性 - HTTPS
* 启动时运行
* 重新启动
* **复制（运行的进程数）**
* 内存
* 启动前的先前步骤

到目前为止，通过文档中的所有教程，您可能已经在**单个进程**上运行了像 Uvicorn 这样的**服务器程序**。

部署应用程序时，您可能希望进行一些**进程复制**，以利用**多核**并能够处理更多请求。

正如您在上一章有关[部署概念](concepts.md){.internal-link target=_blank}中看到的，您可以使用多种策略。

在这里我将向您展示如何将 <a href="https://gunicorn.org/" class="external-link" target="_blank">**Gunicorn**</a> 与 **Uvicorn worker 进程** 一起使用。

/// info

	atomic.AddInt32(&st.s3RequestsInQueue, i)
}

func (st *HTTPStats) incS3RequestsIncoming() {
	// Golang automatically resets to zero if this overflows
	atomic.AddUint64(&st.s3RequestsIncoming, 1)
}

// Converts http stats into struct to be sent back to the client.
func (st *HTTPStats) toServerHTTPStats(toLowerKeys bool) ServerHTTPStats {
	serverStats := ServerHTTPStats{}

* 💽 🛠️
* 💽 🔬
* 🏧 🧾

## 🎁 🆎 & 🔬

↖️ ⚪️➡️ 😐 ⭐ 🆎 💖 `str`, `int`, `float`, ♒️. 👆 💪 ⚙️ 🌅 🏗 ⭐ 🆎 👈 😖 ⚪️➡️ `str`.

👀 🌐 🎛 👆 ✔️, 🛒 🩺 <a href="https://docs.pydantic.dev/latest/concepts/types/" class="external-link" target="_blank">Pydantic 😍 🆎</a>. 👆 🔜 👀 🖼 ⏭ 📃.

🖼, `Image` 🏷 👥 ✔️ `url` 🏑, 👥 💪 📣 ⚫️ ↩️ `str`, Pydantic `HttpUrl`:

//// tab | 🐍 3️⃣.6️⃣ &amp; 🔛

   * {@code key} while the new value was loading then a removal notification will be sent for the
   * new value.
   *
   * <p><b>Warning:</b> this method silently converts checked exceptions to unchecked exceptions,
   * and should not be used with cache loaders which throw checked exceptions. In such cases use
   * {@link #get} instead.
   *

    status->status = tensorflow::errors::NotFound(
        op_name, " does not have number_attr() defined.");
    return nullptr;
  }

  // The returned string is owned by OpRegistry, so liveness is not a concern.
  return input_arg.number_attr().c_str();
}

int TF_OpIsStateful(const char* op_type, TF_Status* status) {
  const tensorflow::OpRegistrationData* op_reg_data;
  status->status =