terraformを使ってAWSのauto recoveryを設定する

バージョン

• Terraform v0.7.0
• AWSのドキュメントは 2016/08/11時点

コード

aws.tf

variable "aws_access_key" {}
variable "aws_secret_key" {}
variable "aws_default_region" {}
variable "subnet_id" {}
variable "security_groups" {}
variable "key_name" {}

variable "const" {
  default = {
    count = 2
  }
}
provider "aws" {
  access_key    = "${var.aws_access_key}"
  secret_key    = "${var.aws_secret_key}"
  region        = "${var.aws_default_region}"
}


resource "aws_instance" "web" {
  count         = "${lookup(var.const, "count")}"
  ami           = "ami-6869aa05"
  instance_type = "m3.medium"
  subnet_id     = "${var.subnet_id}"
  security_groups =  ["${split(",", var.security_groups)}"]
  monitoring = true
  key_name      = "${var.key_name}"
  tags {
    Name = "${format("web%02d", count.index + 1)}"
  }
}

# Cloud watch
# 1分ごとに2回失敗するとリカバリが走る
# http://docs.aws.amazon.com/ja_jp/AmazonCloudWatch/latest/monitoring/UsingAlarmActions.html
# namespace: https://docs.aws.amazon.com/ja_jp/AmazonCloudWatch/latest/DeveloperGuide/aws-namespaces.html
# dimenstion: http://docs.aws.amazon.com/ja_jp/AmazonCloudWatch/latest/monitoring/ec2-metricscollected.html#ec2-metric-dimensions
resource "aws_cloudwatch_metric_alarm" "recovery" {
  alarm_name            = "recovery ${format("web%02d", count.index + 1)}"
  comparison_operator   = "GreaterThanOrEqualToThreshold"
  evaluation_periods    = 1
  metric_name           = "StatusCheckFailed_System"
  namespace             = "AWS/EC2"
  period                = 60
  statistic             = "Minimum"
  threshold             = 2
  alarm_actions         = ["arn:aws:automate:${var.aws_default_region}:ec2:recover"]
  dimensions = {
    InstanceId = "${element(aws_instance.web.*.id, count.index)}"
  }
  count                 = "${lookup(var.const, "count")}"
  # テスト用
  insufficient_data_actions = ["arn:aws:automate:${var.aws_default_region}:ec2:recover"]
  # webに依存させる
  depends_on            = ["aws_instance.web"]
}