Terraform创建CloudWatch告警失败：实例ID类型不匹配求助

我来帮你分析下这个问题：你遇到的错误核心是CloudWatch告警的InstanceId维度需要单个字符串值，但你的模块输出的是列表类型，直接赋值就会触发类型不兼容的问题。

错误信息 #

Error: Error running plan: 1 error(s) occurred: * aws_cloudwatch_metric_alarm.queue_1: dimensions (InstanceId): '' expected type 'string', got unconvertible type '[]interface {}'

问题原因 #

你的模块superb_queue_api-a的output.tf里，ec2-instance输出的是aws_instance.ec2_instances.*.id——这是列表类型（哪怕你设置no_of_inst = 1，Terraform的splat语法*.id依然会返回包含单个元素的列表）。而CloudWatch告警的InstanceId维度要求传入单个字符串格式的实例ID，直接使用列表自然会报错。

你的配置参考 #

主Terraform配置 #

module "superb_queue_api-a" {
  source = "../../Modules/Services"
  no_of_inst = 1
  ami_id = "${data.aws_ami.ubuntu.id}"
  instance_type = "t2.micro"
  subnets = "${data.terraform_remote_state.state_pull.prod-pvt-sub-a}"
  key = "sup_queue_inst"
  sec-group = ["${aws_security_group.ec2_sg.id}"]
  tp = "${var.is_production ? true : false}"
  instance-name = "sup-queue-api-a"
}

resource "aws_cloudwatch_metric_alarm" "queue_1" {
  alarm_name = "sup-queue-api-a"
  comparison_operator = "GreaterThanOrEqualToThreshold"
  evaluation_periods = "1"
  metric_name = "CPUUtilization"
  namespace = "AWS/EC2"
  period = "120"
  statistic = "Average"
  threshold = "90"

  dimensions {
    InstanceId = "${module.superb_queue_api-a.ec2-instance}"
  }

  alarm_description = "This metric monitors ec2 cpu utilization"
  alarm_actions = ["${var.terraform_elb_name}"]
}

模块的EC2实例配置 #

resource "aws_instance" "ec2_instances" {
  count = "${var.no_of_inst}"
  ami = "${var.ami_id}"
  instance_type = "${var.instance_type}"
  subnet_id = "${var.subnets}"
  key_name = "${var.key}"
  vpc_security_group_ids = ["${var.sec-group}"]
  disable_api_termination = "${var.tp}"

  root_block_device {
    volume_size = 100
    volume_type = "gp2"
  }

  tags {
    Name = "${var.instance-name}"
  }
}

模块的output.tf #

output "ec2-instance" {
  value = "${aws_instance.ec2_instances.*.id}"
}

output "private_ip" {
  value = "${aws_instance.ec2_instances.*.private_ip}"
}

解决方案 #

根据你的实际需求，有两种处理方式：

方式1：针对单个实例的场景（当前no_of_inst=1） #

直接取列表的第一个元素，修改告警里的InstanceId赋值：

dimensions {
  InstanceId = "${module.superb_queue_api-a.ec2-instance[0]}"
}

如果你的Terraform版本是0.12+，可以简化为更简洁的写法：

dimensions {
  InstanceId = module.superb_queue_api-a.ec2-instance[0]
}

方式2：支持多个实例的场景（未来可能扩展no_of_inst>1） #

如果以后需要创建多个EC2实例并为每个实例单独配置告警，可以用count关联实例数量，自动生成对应数量的告警：

resource "aws_cloudwatch_metric_alarm" "queue" {
  # 根据模块输出的实例数量创建对应数量的告警
  count = length(module.superb_queue_api-a.ec2-instance)

  alarm_name = "sup-queue-api-${count.index}"
  comparison_operator = "GreaterThanOrEqualToThreshold"
  evaluation_periods = "1"
  metric_name = "CPUUtilization"
  namespace = "AWS/EC2"
  period = "120"
  statistic = "Average"
  threshold = "90"

  dimensions {
    InstanceId = module.superb_queue_api-a.ec2-instance[count.index]
  }

  alarm_description = "This metric monitors CPU utilization for EC2 instance ${module.superb_queue_api-a.ec2-instance[count.index]}"
  alarm_actions = [var.terraform_elb_name]
}

这样不管你设置多少个实例，每个实例都会自动匹配对应的CloudWatch告警。

内容的提问来源于stack exchange，提问作者Nishant Singh