cxmcc/tf_aws_ec2_auto_recovery

A Terraform template for EC2 Auto Recovery

tf_aws_ec2_auto_recovery

A Terraform template for EC2 Auto Recovery

This terraform module implements

• Auto Recovery for Amazon EC2
• Route53 Healthchecks and Failover

This is useful when

• You need to have a group of EC2 instances that need static public IPs (EIP)
• The software on EC2 instances is relatively reliable and you only need to worry about EC2 system errors
• EC2 instances type is one of C3, C4, M3, R3, and T2

Alternatives

• When static IP (EIP) is not needed, ELB with auto scaling group should be considered.
• An auto scaling group with (1,1,1) configured and attach EIP at provision time. (need IAM role for instance)

Components

• aws_eip.eip: an EIP to be assigned to the EC2 instance
• aws_instance.instance: the EC2 instance
• aws_cloudwatch_metric_alarm.auto_recovery_alarm: the Cloudwatch alarm to trigger EC2 instance Auto Recovery
• aws_route53_health_check.route53_health_check: an HTTP health check for the health of the EC2 instance (Assuming the instance serves HTTP traffic and has a health check)
• aws_route53_record.route53_record: a Route53 record point to the EIP attached to the EC2 instnace with healthcheck.

Example

module "static-ip-ec2-instances" {
  source                 = "github.com/cxmcc/tf_aws_ec2_auto_recovery"
  count                  = 2
  domain                 = "static-ip.example.com"
  route53_hosted_zone_id = "ZZZZFFFFEEEENNNN"
  ami                    = "ami-1234567"
  health_check_path      = "/route53_health_check"
  instance_type          = "m3.medium"
  name                   = "static-ip-instance"
  key_name               = "default"
  subnet_ids             = ["subnet-123456", "subnet-234567"]
  vpc_security_group_ids = ["sg-1234567"]

  user_data = <<EOF
#cloud-config
hostname: static-ip-instance
manage_etc_hosts: true
EOF
}

tf_aws_ec2_auto_recovery's Issues

Should the CloudWatch alarm's Statistic be set to "Maximum"?

Hi there,

In the CloudWatch alarm, the Statistic is set to "Minimum" and is being evaluated over 3 periods. Let's say the System Status Check came back as 1, 0, then 1 (fail, pass, fail), then with Statistic set to "Minimum", the auto-recover action won't be triggered (Threshold >= 1). Now this may be what you're going for, but personally I'd rather the action be triggered in this scenario, so I'd set the Statistic to Maximum.

Please let me know if my logic and/or understanding of CW is incorrect.
Thanks!

-- Khalid