> ## Documentation Index
> Fetch the complete documentation index at: https://enterprise-docs.dify.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# 生产环境部署

## 要求

> 可以根据需要扩展资源，或参考我们的[性能调优文档](/versions/3-9-x/zh-cn/deployment/advanced-configuration/performance)。

* Kubernetes 集群
  * 6 个工作节点：每个 8 CPU，32 GB RAM
* 容器镜像仓库（插件）
  * AWS ECR（with AK/SK or IRSA）/ Azure ACR / Aliyun ACR (with AK/SK) / Tencent TCR (with AK/SK) / Dockerhub / Google Artifact Registry
* 持久化存储
  * s3, azure-blob, aliyun-oss, google-storage, tencent-cos, volcengine-tos, huawei-obs
  * 512 GB 存储
* Postgres 数据库
  * 版本 14+
  * 8 CPU，16 GB RAM
  * 至少 512G SSD
* Redis 数据库
  * 版本 6+
  * 2 GB RAM
* [向量数据库虚拟机 - Qdrant 集群](/versions/3-0-x/zh-cn/deployment/advanced-configuration/qdrant/qdrant-cluster)
  * 3 个虚拟机
  * 8 CPU，64 GB RAM
  * 100 GB 存储

## Dify 服务资源的 `limits`

<Warning>
  `workerBeat` 应该始终作为单个副本运行。
</Warning>

| Services            | Replicas | CPU | RAM - GB  | Total CPU | Total RAM - GB |
| ------------------- | -------- | --- | --------- | --------- | -------------- |
| api                 | 6        | 1   | 2         | 9         | 12             |
| worker              | 9        | 1   | 2         | 6         | 18             |
| workerBeat          | 1        | 1   | 1         | 1         | 1              |
| web                 | 3        | 1   | 1         | 3         | 3              |
| sandbox             | 3        | 1   | 2         | 3         | 6              |
| enterprise          | 3        | 1   | 2         | 3         | 6              |
| enterpriseAudit     | 3        | 1   | 2         | 3         | 6              |
| enterpriseFrontend  | 3        | 1   | 1         | 3         | 3              |
| enterpriseCollector | 1        | 1   | 1         | 1         | 1              |
| ssrfProxy           | 3        | 1   | 1         | 3         | 3              |
| unstructured        | 3        | 1   | 2         | 3         | 6              |
| plugin\_daemon      | 3        | 1   | 3         | 3         | 9              |
| plugin\_controller  | 3        | 1   | 2         | 3         | 6              |
| plugin\_connector   | 3        | 1   | 2         | 3         | 6              |
| plugin\_manager     | 3        | 1   | 2         | 3         | 6              |
| gateway             | 3        | 1   | 2         | 3         | 6              |
|                     |          |     | **Total** | 52        | 98             |

## 部署

### 1. 获取 Helm 仓库信息

```bash theme={null}
helm repo add dify https://langgenius.github.io/dify-helm
helm repo update
```

有关更多信息，请参阅 [Dify Helm Chart](/versions/3-0-x/zh-cn/deployment/advanced-configuration/dify-helm-chart)。

### 2. 保存 Helm Chart 值

* 将 Helm chart 值保存到名为 `values.yaml` 的文件中。
* 使用你的配置更新 `values.yaml` 文件。

<Warning>
  - 请将 `#REPLACE_ME#` 占位符替换为你的实际值。
  - 请将 "dify123456" 替换为你自己的值。
  - `global.appSecretKey` 和 `enterprise.appSecretKey` 必须不同。
</Warning>

```yaml theme={null}
###################################
# Please replace "dify123456" with your own value
###################################
global:
  appSecretKey: 'dify123456'
  innerApiKey: 'dify123456'
  consoleApiDomain: "console.dify.local"
  consoleWebDomain: "console.dify.local"
  serviceApiDomain: "api.dify.local"
  appApiDomain: "app.dify.local"
  appWebDomain: "app.dify.local"
  filesDomain: "upload.dify.local"
  enterpriseDomain: "enterprise.dify.local"
  # Trigger service domain (used by API/workers for trigger functionality; if empty, triggers may not work)
  triggerDomain: "trigger.dify.local"

ingress:
  enabled: true
  className: "nginx"
  annotations: {
    # set file upload size limit
    nginx.ingress.kubernetes.io/proxy-body-size: "15m"
  }

api:
  replicas: 6
  serverWorkerAmount: 1
  resources:
    limits:
      cpu: 1000m
      memory: 2048Mi
    requests:
      cpu: 500m
      memory: 1024Mi
worker:
  replicas: 9
  celeryWorkerAmount: 1
  resources:
    limits:
      cpu: 1000m
      memory: 2048Mi
    requests:
      cpu: 500m
      memory: 1024Mi
workerBeat:
  resources:
    limits:
      cpu: 1000m
      memory: 1024Mi
    requests:
      cpu: 500m
      memory: 512Mi
web:
  replicas: 3
  resources:
    limits:
      cpu: 1000m
      memory: 1024Mi
    requests:
      cpu: 500m
      memory: 512Mi
sandbox:
  replicas: 3
  resources:
    limits:
      cpu: 1000m
      memory: 2048Mi
    requests:
      cpu: 500m
      memory: 1024Mi
enterprise:
  replicas: 3
  appSecretKey: "dify123456"
  adminAPIsSecretKeySalt: "dify123456"
  passwordEncryptionKey: "dify123456"
  resources:
    limits:
      cpu: 1000m
      memory: 2048Mi
    requests:
      cpu: 500m
      memory: 1024Mi
enterpriseAudit:
  replicas: 3
  resources:
    limits:
      cpu: 1000m
      memory: 2048Mi
    requests:
      cpu: 500m
      memory: 1024Mi
enterpriseFrontend:
  replicas: 3
  resources:
    limits:
      cpu: 1000m
      memory: 1024Mi
    requests:
      cpu: 500m
      memory: 512Mi
ssrfProxy:
  replicas: 3
  resources:
    limits:
      cpu: 1000m
      memory: 1024Mi
    requests:
      cpu: 500m
      memory: 512Mi
unstructured:
  replicas: 3
  resources:
    limits:
      cpu: 1000m
      memory: 2048Mi
    requests:
      cpu: 500m
      memory: 1024Mi
plugin_daemon:
  replicas: 3
  resources:
    limits:
      cpu: 1000m
      memory: 3072Mi
    requests:
      cpu: 500m
      memory: 1536Mi
plugin_controller:
  replicas: 3
  resources:
    limits:
      cpu: 1000m
      memory: 2048Mi
    requests:
      cpu: 500m
      memory: 1024Mi
plugin_connector:
  replicas: 3
  resources:
    limits:
      cpu: 1000m
      memory: 2048Mi
    requests:
      cpu: 500m
      memory: 1024Mi
plugin_manager:
  replicas: 3
  resources:
    limits:
      cpu: 1000m
      memory: 2048Mi
    requests:
      cpu: 500m
      memory: 1024Mi
gateway:
  replicas: 3
  resources:
    limits:
      cpu: 1000m
      memory: 2048Mi
    requests:
      cpu: 500m
      memory: 1024Mi

###################################
# Persistence Configration
###################################
persistence:
  type: "s3"
  s3:
    endpoint: "https://xxx.r2.cloudflarestorage.com"
    accessKey: "#REPLACE_ME#"
    secretKey: "#REPLACE_ME#"
    region: "us-east-1"
    bucketName: "your-bucket-name"
    addressType: ""
    useAwsManagedIam: false
    useAwsS3: true

###################################
# External database (Postgres/MySQL/TiDB) 配置
###################################
externalDatabase:
  enabled: false
  # postgres | mysql | tidb
  engine: "postgres"
  host: "localhost"

  # Default ports: postgres=5432, mysql=3306, tidb=4000
  # Note: when switching engine, remember to update port accordingly
  port: 5432

  # 全局数据库账号（默认使用）
  user: "postgres"
  password: "#REPLACE_ME#"

  timezone: "UTC"

  databases:
    dify: "dify"
    plugin_daemon: "dify_plugin_daemon"
    enterprise: "enterprise"
    audit: "audit"

  ###################################
  # Per-database credentials（可选）
  ###################################
  # 用于为每个数据库单独配置账号密码
  # 优先级高于全局 user/password
  # 如果为空，则回退使用全局账号
  # databaseCredentials：自 3.9.1 起可用（可选；覆盖上方的全局 user/password，留空则回退到全局值）。
  databaseCredentials:
    dify:
      user: ""
      password: ""
    enterprise:
      user: ""
      password: ""
    audit:
      user: ""
      password: ""
    plugin_daemon:
      user: ""
      password: ""

  dialectOptions:
    postgres:
      sslMode: "require"
      uriScheme: "postgresql"
      # extras: "options=-c search_path=your-schema -c your-other-option=xxx"
      extras: ""

    mysql:
      # MySQL/TiDB params, e.g. charset=utf8mb4&parseTime=true&loc=UTC
      params: "charset=utf8mb4&parseTime=true&loc=UTC"
      tls: false

###################################
# External Redis
###################################
externalRedis:
  enabled: true
  host: "#REPLACE_ME#"
  port: 6379
  username: ""
  password: "#REPLACE_ME#"
  useSSL: false

###################################
# External Qdrant
###################################
vectorDB:
  useExternal: true
  externalType: "qdrant"
  externalQdrant:
    endpoint: "http://your-qdrant-cluster-url.qdrant.tech/"
    apiKey: "#REPLACE_ME#"

imagePullSecrets: []
```

### 3. 配置 `容器镜像仓库`

* 3.1 配置 `插件` 容器镜像仓库
  * 更多信息，请参考 [插件的容器镜像仓库配置](/versions/3-0-x/zh-cn/deployment/advanced-configuration/container-registry-for-plugins)。
* 3.2 配置 `服务` 容器镜像仓库
  * 更多信息，请参考 [容器镜像仓库配置](/versions/3-0-x/zh-cn/deployment/advanced-configuration/container-registry)。

### 4. 配置 `Persistence Storage`

* 更多信息，请参考 [Persistence Storage](/versions/3-0-x/zh-cn/deployment/advanced-configuration/persistent-storage)。

### 5. 配置 `External Database`

* 更多信息，请参考 [External Database](/versions/3-8-x/zh-cn/deployment/advanced-configuration/database)。

### 6. 配置 `External Redis`

* 更多信息，请参考 [External Redis](/versions/3-0-x/zh-cn/deployment/advanced-configuration/redis)。

### 7. 配置 `External Qdrant`

* 更多信息，请参考 [External Qdrant Cluster](/versions/3-0-x/zh-cn/deployment/advanced-configuration/qdrant/qdrant-cluster)。

### 8. 配置 `Ingress Controller`

* 更多信息，请参考 [Ingress Controller 配置](/versions/3-0-x/zh-cn/deployment/advanced-configuration/nginx-ingress-controller)。

### 9. 安装 Dify Helm Chart

```bash theme={null}
helm upgrade -i dify -f values.yaml dify/dify
```

### 10. 验证安装结果

* 更多信息，请参考 [服务状态验证](/versions/3-0-x/zh-cn/deployment/verify-services-status)。

## 高级配置

### 1. 域名与 SSL 证书

* 更多信息，请参考 [域名与 SSL 证书配置](/versions/3-0-x/zh-cn/deployment/advanced-configuration/domain-and-ssl-certificate)。

### 2. 邮件服务提供商

* 更多信息，请参考 [邮件服务配置](/versions/3-0-x/zh-cn/deployment/advanced-configuration/email-provider)。

### 3. 性能调优

* 更多信息，请参考 [性能调优](/versions/3-9-x/zh-cn/deployment/advanced-configuration/performance)。
* 如需将上面的兜底 `worker` 拆分为按队列划分的专用 Worker（`additionalWorkers`），请参考 [按队列拆分 Worker](/versions/3-9-x/zh-cn/deployment/advanced-configuration/performance#2-按队列拆分-worker)。