Terraform嵌套循环创建AWS Glue Workflow资源报错排查

错误原因分析 #

你当前代码中使用toset(local.my_map)将对象列表转为集合后，each.key变成了对象的字符串化表示，而非原对象，导致element函数无法从字符串中提取元素，最终触发报错。要实现嵌套遍历并创建触发器，需要先扁平化所有需要的触发关系，再通过for_each遍历处理。

修正方案 #

1. 重构本地变量，生成扁平化的触发关系 #

修改locals块，将启动job到各flow首job的触发、flow内部job链的触发全部整理为一维的触发规则集合：

locals {
    workflow_name = "my_example"
    first_job = "Job_start"
    my_map = [
        {
            flow = ["JOB-A1", "JOB-A2", "JOB-A3"]
        },
        {
            flow = ["JOB-B1", "JOB-B2", "JOB-B3"]
        }
    ]

    # 生成从启动job到每个flow首个job的触发规则
    start_triggers = [
        for flow in local.my_map : {
            trigger_name = "trigger-start-to-${lower(replace(flow.flow[0], "_", "-"))}"
            predicate_job = replace(lower(local.first_job), "_", "-")
            action_job = replace(lower(flow.flow[0]), "_", "-")
        }
    ]

    # 生成每个flow内部相邻job的触发规则，并用flatten转为一维列表
    inner_triggers = flatten([
        for flow in local.my_map : [
            for i in range(length(flow.flow) - 1) : {
                trigger_name = "trigger-${lower(replace(flow.flow[i], "_", "-"))}-to-${lower(replace(flow.flow[i+1], "_", "-"))}"
                predicate_job = replace(lower(flow.flow[i]), "_", "-")
                action_job = replace(lower(flow.flow[i+1]), "_", "-")
            }
        ]
    ])

    # 合并所有触发规则
    all_triggers = concat(local.start_triggers, local.inner_triggers)

    # 转为map格式，用唯一的trigger_name作为key，适配for_each要求
    triggers_map = {
        for t in local.all_triggers : t.trigger_name => t
    }
}

2. 修改触发器资源定义 #

基于扁平化的triggers_map创建触发器，代码更简洁且避免索引错误：

resource "aws_glue_trigger" "this" {
  for_each      = local.triggers_map
  name          = each.key
  type          = "CONDITIONAL"
  workflow_name = aws_glue_workflow.example.name

  predicate {
    conditions {
      job_name = each.value.predicate_job
      state    = "SUCCEEDED"
    }
  }

  actions {
    job_name = each.value.action_job
  }
}

效果说明 #

执行terraform plan后，会生成以下触发器：

• trigger-start-to-job-a1：启动job触发JOB-A1
• trigger-start-to-job-b1：启动job触发JOB-B1
• trigger-job-a1-to-job-a2：JOB-A1触发JOB-A2
• trigger-job-a2-to-job-a3：JOB-A2触发JOB-A3
• trigger-job-b1-to-job-b2：JOB-B1触发JOB-B2
• trigger-job-b2-to-job-b3：JOB-B2触发JOB-B3

内容的提问来源于stack exchange，提问作者Alberto.Garau