如何用Serverless Framework为多环境配置DynamoDB ProvisionedThroughput

完全可以在同一个serverless.yml里实现不同环境的DynamoDB配置隔离，甚至能灵活开启/关闭自动扩缩容功能。核心思路是利用自定义变量插值结合CloudFormation条件判断，让配置根据部署的环境（stage）自动切换。下面给你两种实用的实现方案：

方案1：内置自定义变量+条件渲染 #

这是最直接的方式，不需要额外插件，完全依赖Serverless Framework的原生能力。

步骤1：在serverless.yml中定义多环境配置 #

首先在custom块里按环境（比如dev/prod）定义好DynamoDB的专属配置：

service: my-serverless-app

custom:
  dynamodbConfigs:
    dev:
      rcu: 1
      wcu: 1
      enableAutoScaling: false
    prod:
      rcu: 5
      wcu: 5
      enableAutoScaling: true
      autoScalingMax: 100
      targetUsagePercent: 70

# 指定默认环境，避免部署时忘记传stage
provider:
  name: aws
  runtime: nodejs18.x
  stage: ${opt:stage, 'dev'} # 默认用dev环境
  region: us-east-1

步骤2：根据环境渲染DynamoDB表配置 #

接下来在resources块里定义DynamoDB表，通过${self:custom.dynamodbConfigs.${self:provider.stage}.xxx}引用对应环境的配置，同时用CloudFormation的Fn::If来判断是否启用自动扩缩容：

resources:
  Resources:
    MyDynamoDBTable:
      Type: AWS::DynamoDB::Table
      Properties:
        TableName: MyTable-${self:provider.stage}
        AttributeDefinitions:
          - AttributeName: id
            AttributeType: S
        KeySchema:
          - AttributeName: id
            KeyType: HASH
        # 基础读写容量，自动匹配当前环境
        ProvisionedThroughput:
          ReadCapacityUnits: ${self:custom.dynamodbConfigs.${self:provider.stage}.rcu}
          WriteCapacityUnits: ${self:custom.dynamodbConfigs.${self:provider.stage}.wcu}
        BillingMode: PROVISIONED # 必须用PROVISIONED才能开启自动扩缩容
        StreamSpecification:
          StreamViewType: NEW_IMAGE

    # 针对生产环境添加读容量自动扩缩容策略
    ReadScalingPolicy:
      Type: AWS::ApplicationAutoScaling::ScalingPolicy
      Condition: IsProd # 只在prod环境创建
      Properties:
        PolicyName: DynamoDBReadCapacityScalingPolicy
        PolicyType: TargetTrackingScaling
        ScalingTargetId: !Ref ReadScalingTarget
        TargetTrackingScalingPolicyConfiguration:
          TargetValue: ${self:custom.dynamodbConfigs.prod.targetUsagePercent}
          PredefinedMetricSpecification:
            PredefinedMetricType: DynamoDBReadCapacityUtilization

    ReadScalingTarget:
      Type: AWS::ApplicationAutoScaling::ScalableTarget
      Condition: IsProd
      Properties:
        MaxCapacity: ${self:custom.dynamodbConfigs.prod.autoScalingMax}
        MinCapacity: ${self:custom.dynamodbConfigs.prod.rcu}
        ResourceId: table/${self:resources.Resources.MyDynamoDBTable}
        ScalableDimension: dynamodb:table:ReadCapacityUnits
        ServiceNamespace: dynamodb

    # 同理添加写容量自动扩缩容策略
    WriteScalingPolicy:
      Type: AWS::ApplicationAutoScaling::ScalingPolicy
      Condition: IsProd
      Properties:
        PolicyName: DynamoDBWriteCapacityScalingPolicy
        PolicyType: TargetTrackingScaling
        ScalingTargetId: !Ref WriteScalingTarget
        TargetTrackingScalingPolicyConfiguration:
          TargetValue: ${self:custom.dynamodbConfigs.prod.targetUsagePercent}
          PredefinedMetricSpecification:
            PredefinedMetricType: DynamoDBWriteCapacityUtilization

    WriteScalingTarget:
      Type: AWS::ApplicationAutoScaling::ScalableTarget
      Condition: IsProd
      Properties:
        MaxCapacity: ${self:custom.dynamodbConfigs.prod.autoScalingMax}
        MinCapacity: ${self:custom.dynamodbConfigs.prod.wcu}
        ResourceId: table/${self:resources.Resources.MyDynamoDBTable}
        ScalableDimension: dynamodb:table:WriteCapacityUnits
        ServiceNamespace: dynamodb

  # 定义环境判断条件
  Conditions:
    IsProd: !Equals [${self:provider.stage}, 'prod']

步骤3：部署指定环境 #

部署时只需指定stage参数即可自动加载对应配置：

# 部署测试环境
serverless deploy --stage dev

# 部署生产环境
serverless deploy --stage prod

方案2：使用独立环境配置文件 #

如果你的配置项很多，或者想把环境配置和主配置分离，可以用单独的YAML文件来存储各环境的配置，比如创建env/dev.yml和env/prod.yml：

env/dev.yml内容：

dynamodb:
  rcu: 1
  wcu: 1
  enableAutoScaling: false

env/prod.yml内容：

dynamodb:
  rcu: 5
  wcu: 5
  enableAutoScaling: true
  autoScalingMax: 100
  targetUsagePercent: 70

然后在serverless.yml中通过file()函数加载对应环境的配置：

custom:
  # 加载对应stage的配置文件
  envConfig: ${file(env/${self:provider.stage}.yml)}

provider:
  name: aws
  runtime: nodejs18.x
  stage: ${opt:stage, 'dev'}
  region: us-east-1

# 后续引用配置时用${self:custom.envConfig.dynamodb.rcu}即可，其他逻辑和方案1一致

这种方式的好处是主配置文件更简洁，环境配置单独管理，适合团队协作场景。

关键注意事项 #

• 自动扩缩容仅支持PROVISIONED计费模式，不能用PAY_PER_REQUEST
• 部署生产环境时，确保AWS账号有ApplicationAutoScaling相关的权限
• 如果需要给GSI配置自动扩缩容，只需复制上面的ScalingPolicy和ScalableTarget资源，修改ResourceId为table/${tableName}/index/${indexName}即可

内容的提问来源于stack exchange，提问作者rhythm