如何让Terraform仅重新部署代码变更的Cloud Function？

你当前的问题核心在于所有Cloud Function共用了同一套代码归档和存储桶对象资源，导致单个函数代码变更会触发全局MD5更新，进而所有函数都被重新部署。要解决这个问题，关键是让每个函数拥有独立的代码归档、存储桶对象资源，具体实现如下：

1. 为每个函数创建独立的代码归档 #

每个函数的代码目录独立，单独定义data.archive_file资源，这样只有对应目录下的代码变更时，该资源的output_md5才会变化。

示例代码：

# 函数A的代码归档
data "archive_file" "function_a_source" {
  type        = "zip"
  source_dir  = "./functions/function-a" # 函数A专属代码目录
  output_path = "./build/function-a.zip"
}

# 函数B的代码归档
data "archive_file" "function_b_source" {
  type        = "zip"
  source_dir  = "./functions/function-b" # 函数B专属代码目录
  output_path = "./build/function-b.zip"
}

2. 为每个函数创建独立的存储桶对象 #

每个函数对应独立的google_storage_bucket_object，用自身归档的MD5命名，确保只有该函数代码变更时，存储桶对象名称才会更新。

示例代码：

# 函数A的存储桶对象
resource "google_storage_bucket_object" "function_a_zip" {
  name   = "function-a#${data.archive_file.function_a_source.output_md5}"
  bucket = var.bucket.name
  source = data.archive_file.function_a_source.output_path
}

# 函数B的存储桶对象
resource "google_storage_bucket_object" "function_b_zip" {
  name   = "function-b#${data.archive_file.function_b_source.output_md5}"
  bucket = var.bucket.name
  source = data.archive_file.function_b_source.output_path
}

3. 每个Cloud Function关联自身的存储桶对象 #

让每个google_cloudfunctions_function资源只引用对应函数的存储桶对象名称，这样只有当该对象名称变化时，函数才会触发重新部署。

示例代码：

# 函数A的Cloud Function配置
resource "google_cloudfunctions_function" "function_a" {
  name    = "function-a"
  runtime = "java11"

  available_memory_mb   = 256
  source_archive_bucket = var.bucket.name
  source_archive_object = google_storage_bucket_object.function_a_zip.name
  timeout               = 120
  entry_point           = "com.example.FunctionA"

  event_trigger {
    event_type = "google.storage.object.finalize"
    resource   = "projects/${var.project_id}/buckets/${var.bucket.name}"
  }

  environment_variables = {
    PROJECT_ID = var.project_id
  }
}

# 函数B的Cloud Function配置
resource "google_cloudfunctions_function" "function_b" {
  name    = "function-b"
  runtime = "java11"

  available_memory_mb   = 256
  source_archive_bucket = var.bucket.name
  source_archive_object = google_storage_bucket_object.function_b_zip.name
  timeout               = 120
  entry_point           = "com.example.FunctionB"

  event_trigger {
    event_type = "google.pubsub.topic.publish"
    resource   = "projects/${var.project_id}/topics/my-topic"
  }

  environment_variables = {
    PROJECT_ID = var.project_id
  }
}

进阶：用Terraform模块封装（推荐） #

如果函数数量较多，重复写上述配置会很繁琐，用模块封装单个Cloud Function的逻辑，通过模块调用创建多个函数，既保证资源独立，又简化维护。

模块目录结构 #

modules/
  cloud-function/
    main.tf
    variables.tf
    outputs.tf

模块main.tf示例 #

# 模块内的代码归档
data "archive_file" "source" {
  type        = "zip"
  source_dir  = var.source_dir
  output_path = "${var.build_dir}/${var.function_name}.zip"
}

# 模块内的存储桶对象
resource "google_storage_bucket_object" "zip_file" {
  name   = "${var.function_name}#${data.archive_file.source.output_md5}"
  bucket = var.bucket_name
  source = data.archive_file.source.output_path
}

# 模块内的Cloud Function
resource "google_cloudfunctions_function" "function" {
  name    = var.function_name
  runtime = var.runtime

  available_memory_mb   = var.memory
  source_archive_bucket = var.bucket_name
  source_archive_object = google_storage_bucket_object.zip_file.name
  timeout               = var.timeout
  entry_point           = var.entry_point

  event_trigger {
    event_type = var.event_trigger_type
    resource   = var.event_trigger_resource
  }

  environment_variables = var.environment_variables

  timeouts {
    create = "60m"
  }
}

根模块调用示例 #

# 创建函数A
module "function_a" {
  source = "./modules/cloud-function"

  function_name        = "function-a"
  runtime              = "java11"
  memory               = 256
  timeout              = 120
  entry_point          = "com.example.FunctionA"
  source_dir           = "./functions/function-a"
  build_dir            = "./build"
  bucket_name          = var.bucket.name
  event_trigger_type   = "google.storage.object.finalize"
  event_trigger_resource = "projects/${var.project_id}/buckets/${var.bucket.name}"
  environment_variables = {
    PROJECT_ID = var.project_id
  }
}

# 创建函数B
module "function_b" {
  source = "./modules/cloud-function"

  function_name        = "function-b"
  runtime              = "java11"
  memory               = 256
  timeout              = 120
  entry_point          = "com.example.FunctionB"
  source_dir           = "./functions/function-b"
  build_dir            = "./build"
  bucket_name          = var.bucket.name
  event_trigger_type   = "google.pubsub.topic.publish"
  event_trigger_resource = "projects/${var.project_id}/topics/my-topic"
  environment_variables = {
    PROJECT_ID = var.project_id
  }
}

这样修改后，只有代码发生变更的函数对应的归档、存储桶对象会更新，进而触发该函数的重新部署，其他函数不受影响。

内容的提问来源于stack exchange，提问作者GiuxG97