Terraform部署Azure多区域最佳实践：如何避免代码重复？

核心思路：for_each迭代区域+模块化封装 #

不用为每个区域重复编写模块调用代码，通过集中定义区域列表+for_each批量实例化模块实现多区域部署，同时保留区域配置的灵活性。

1. 定义全局区域变量 #

在根模块中统一管理要部署的目标区域，后续新增/移除区域只需修改这个变量：

variable "target_regions" {
  type        = list(string)
  description = "Azure regions to deploy infrastructure"
  default     = ["eastus", "westeurope", "southeastasia"]
}

2. 封装区域通用资源模块 #

把每个区域需要部署的资源（资源组、VNet、VM、存储账户等）封装到独立模块，模块核心参数为目标区域：

# ./modules/regional_infra/main.tf
variable "region" {
  type        = string
  description = "Target Azure region for this module instance"
}

variable "resource_prefix" {
  type        = string
  description = "Prefix for all resource names in this region"
}

# 区域资源组
resource "azurerm_resource_group" "regional_rg" {
  name     = "${var.resource_prefix}-rg-${var.region}"
  location = var.region
}

# 区域虚拟网络（用区域索引生成唯一网段）
resource "azurerm_virtual_network" "regional_vnet" {
  name                = "${var.resource_prefix}-vnet-${var.region}"
  address_space       = ["10.${index(var.target_regions, var.region)}.0.0/16"]
  location            = var.region
  resource_group_name = azurerm_resource_group.regional_rg.name
}

# 其他区域级资源（VM、存储账户等）可按需添加...

3. 用for_each批量调用模块 #

在根模块中通过for_each迭代区域列表，自动为每个区域创建模块实例，无需重复编写代码：

module "regional_deployments" {
  source = "./modules/regional_infra"

  # 将列表转为集合，实现逐个区域迭代
  for_each = toset(var.target_regions)

  region          = each.value
  resource_prefix = "my-production-app"
}

4. 处理区域特定配置 #

如果不同区域需要差异化参数（比如VM SKU、存储冗余策略、标签），用map类型变量定义区域专属配置：

# 根模块中定义区域特定配置
variable "region_specific_settings" {
  type = map(object({
    vm_sku               = string
    storage_account_tier = string
    tags                 = map(string)
  }))
  default = {
    eastus = {
      vm_sku               = "Standard_D4s_v3"
      storage_account_tier = "Premium_LRS"
      tags                 = { environment = "prod", role = "primary" }
    }
    westeurope = {
      vm_sku               = "Standard_D2s_v3"
      storage_account_tier = "Standard_GRS"
      tags                 = { environment = "prod", role = "secondary" }
    }
  }
}

# 模块调用时传入对应配置
module "regional_deployments" {
  source = "./modules/regional_infra"

  for_each = toset(var.target_regions)

  region          = each.value
  resource_prefix = "my-production-app"
  vm_sku          = var.region_specific_settings[each.value].vm_sku
  storage_tier    = var.region_specific_settings[each.value].storage_account_tier
  resource_tags   = var.region_specific_settings[each.value].tags
}

5. 全局资源单独处理 #

对于Azure全局资源（比如Traffic Manager、Front Door），不需要按区域迭代，直接在根模块中部署，并通过模块输出关联区域资源：

# 部署Traffic Manager关联多区域端点
resource "azurerm_traffic_manager_profile" "global_profile" {
  name                = "my-app-traffic-manager"
  resource_group_name = azurerm_resource_group.global_rg.name
  traffic_routing_method = "Priority"

  endpoint {
    name               = "eastus-endpoint"
    target_resource_id = module.regional_deployments["eastus"].public_ip_id
    priority           = 1
  }

  endpoint {
    name               = "westeurope-endpoint"
    target_resource_id = module.regional_deployments["westeurope"].public_ip_id
    priority           = 2
  }
}

6. 状态与输出 #

• 状态管理：推荐用Azure Blob存储作为远程后端，集中管理多区域部署状态；若需隔离区域状态，可按区域创建Terraform工作区，但集中状态更便于统一维护。
• 批量输出：用for表达式批量导出所有区域的资源信息：

output "regional_resource_groups" {
  type = map(string)
  value = {
    for region in var.target_regions : region => module.regional_deployments[region].rg_id
  }
}

内容的提问来源于stack exchange，提问作者Chris Taverner