Terraform创建GCP带静态IP实例报错：count引用上下文错误

错误原因 #

你遇到的Reference to "count" in non-counted context错误，核心是**count和for_each是Terraform中两种独立的多实例创建模式，不能在使用for_each的资源里引用count.index**。你的静态IP资源用count迭代创建，但虚拟机用for_each，两者的迭代上下文不兼容，无法直接交叉引用。

解决方案 #

下面提供两种可靠的实现方式，优先推荐第一种（基于名称关联，更稳定）：

方案1：通过名称关联静态IP与虚拟机（推荐） #

这种方式让每个虚拟机配置明确指定对应的静态IP名称，避免依赖顺序，更适合生产环境。

1. 调整虚拟机变量结构

在var.virtual_machines中新增ip_name字段，指定该虚拟机要绑定的静态IP名称：

variable "virtual_machines" {
  type = map(object({
    vm_size = string
    zone = string
    ip_name = string # 指定对应静态IP的名称
  }))
  example = {
    vm1 = {
      vm_size = "n1-standard-1"
      zone = "us-central1-a"
      ip_name = "ip1"
    },
    vm2 = {
      vm_size = "n1-standard-1"
      zone = "us-central1-b"
      ip_name = "ip2"
    }
  }
}

2. 修改静态IP资源为for_each迭代

把静态IP的创建方式从count改为for_each，用IP名称作为键，方便后续通过名称查找：

resource "google_compute_address" "static" {
  for_each = toset(var.ext_ip_name)
  project      = var.project_id
  name         = each.key
  region       = var.region
  address_type = "EXTERNAL"
}

3. 在虚拟机中绑定静态IP

在network_interface的access_config里，通过虚拟机配置的ip_name关联对应的静态IP：

resource "google_compute_instance" "vm" {
  for_each = var.virtual_machines
  name         = each.key
  machine_type = each.value.vm_size
  zone         = each.value.zone
  project      = var.project_id
  tags         = ["foo", "bar"]

  boot_disk {
    initialize_params {
      image  = "debian-cloud/debian-11"
      size   = 100
      type   = "pd-standard"
      labels = { my_label = "value" }
    }
  }

  network_interface {
    network = var.network
    subnetwork = "k8s-general-prod-gke-subnet"
    subnetwork_project = var.project_id

    access_config {
      nat_ip       = google_compute_address.static[each.value.ip_name].address
      network_tier = "PREMIUM"
    }
  }

  metadata = { foo = "bar" }

  service_account {
    email  = data.google_compute_default_service_account.default.email
    scopes = ["cloud-platform"]
  }
}

方案2：通过索引关联（依赖顺序，不推荐） #

如果不想修改虚拟机变量结构，且能保证虚拟机和静态IP的数量、顺序完全对应，可以把虚拟机的for_each改为count，统一用索引关联：

# 静态IP资源保持count不变
resource "google_compute_address" "static" {
  count = length(var.ext_ip_name)
  project      = var.project_id
  name         = var.ext_ip_name[count.index]
  region       = var.region
  address_type = "EXTERNAL"
}

# 虚拟机改用count迭代
resource "google_compute_instance" "vm" {
  count = length(var.virtual_machines)
  name         = keys(var.virtual_machines)[count.index]
  machine_type = values(var.virtual_machines)[count.index].vm_size
  zone         = values(var.virtual_machines)[count.index].zone
  project      = var.project_id
  tags         = ["foo", "bar"]

  boot_disk {
    initialize_params {
      image  = "debian-cloud/debian-11"
      size   = 100
      type   = "pd-standard"
      labels = { my_label = "value" }
    }
  }

  network_interface {
    network = var.network
    subnetwork = "k8s-general-prod-gke-subnet"
    subnetwork_project = var.project_id

    access_config {
      nat_ip       = google_compute_address.static[count.index].address
      network_tier = "PREMIUM"
    }
  }

  metadata = { foo = "bar" }

  service_account {
    email  = data.google_compute_default_service_account.default.email
    scopes = ["cloud-platform"]
  }
}

注意：这种方式依赖var.virtual_machines的键顺序，如果后续调整虚拟机顺序，会导致Terraform销毁重建实例，风险较高。

内容的提问来源于stack exchange，提问作者zakabluk