terraform-aws-bastion-host-ssm

This Terraform module installs a bastion host accessible via SSM only. The underlying EC2 instance has no ports opened. All data is encrypted and a resource_prefix can be specified to integrate into your naming schema.

The implemented connection method allows port forwarding for one port only. Multiple port forwardings can be realized by the user by creating multiple connections to the bastion host.

Check the examples directory for the module usage.

Cost Estimation (for version 2.4.0)

Name                                                                    Monthly Qty  Unit                   Monthly Cost
module.bastion_host.aws_autoscaling_group.on_spot[0]
 └─ module.bastion_host.aws_launch_template.manual_start
    ├─ Instance usage (Linux/UNIX, on-demand, t3.nano)                          730  hours                         $4.38
      └─ root_block_device
         └─ Storage (general purpose SSD, gp3)                                   16  GB                            $1.52
    └─ Instance usage (Linux/UNIX, spot, t3.nano)                               730  hours                         $1.31
      └─ root_block_device
         └─ Storage (general purpose SSD, gp3)                                   16  GB                            $1.52

 module.bastion_host.aws_cloudwatch_log_group.panic_button_off
 ├─ Data ingested                                               Monthly cost depends on usage: $0.63 per GB
 ├─ Archival Storage                                            Monthly cost depends on usage: $0.0324 per GB
 └─ Insights queries data scanned                               Monthly cost depends on usage: $0.0063 per GB

 module.bastion_host.aws_cloudwatch_log_group.panic_button_on
 ├─ Data ingested                                               Monthly cost depends on usage: $0.63 per GB
 ├─ Archival Storage                                            Monthly cost depends on usage: $0.0324 per GB
 └─ Insights queries data scanned                               Monthly cost depends on usage: $0.0063 per GB

 module.bastion_host.aws_lambda_function.panic_button_off
 ├─ Requests                                                    Monthly cost depends on usage: $0.20 per 1M requests
 └─ Duration                                                    Monthly cost depends on usage: $0.0000166667 per GB-seconds

 module.bastion_host.aws_lambda_function.panic_button_on
 ├─ Requests                                                    Monthly cost depends on usage: $0.20 per 1M requests
 └─ Duration                                                    Monthly cost depends on usage: $0.0000166667 per GB-seconds

 OVERALL TOTAL                                                                                                     $8.73

Features

• use autoscaling groups to replace dead instances
• have a schedule to shutdown the instance at night
• Keepass support for AWS credentials
• use spot instances to save some money
• provide IAM role for easy access
• provide a script to connect to the bastion from your local machine
• panic switch to enable the bastions or disable them immediately

Panic Switch

Two lambda functions are provided. One to enable the bastion host, e.g. if you have to work at night and the bastion hosts are deactivated. The second lambda function disables the bastion host immediately no matter what.

As both functions are destructive (they modify the autoscaling group), you have to re-apply this module as soon as possible to restore the auto scaling setting (especially the schedules).

Keepass Support For IAM User Credentials

In case you are not using SSO or similar techniques you have to store the credentials for the user able to connect to the bastion host somewhere. We provide a little helper script to handle this scenario in a secure way.

Create a Keepass database and add the KPScript plugin. The scripts/export_aws_credentials_from_keypass.sh will read and export the credentials from the Keepass database.

Schedules

Schedules allow to start and shutdown the instance at certain times. If your work hours are from 9 till 5 in Berlin, add

module "bastion" {
  # ...
  schedule {
    start = "0 9 * * MON-FRI"
    stop = "0 17 * * MON-FRI"

    time_zone = "Europe/Berlin"
  }
}

The bastion host will automatically start at 9 and shuts down at 17 from monday to friday (Berlin time). Depending on the instance_type you will save more or less money. Do not forget to adjust the timezone.

In case you have to start a bastin host outside the working hours use the launch template provided by the module and launch the new instance from the AWS CLI or Console. Don't forget to shut it down if you are done.

Encryption

In case you are using spot instances don't forget to allow AWSServiceRoleForAutoScaling to access your keys.

data "aws_iam_policy_document" "key_policy" {
    # ...

    statement {
    sid    = "AdminKMSManagement"
    effect = "Allow"

    principals {
      identifiers = [
        "arn:aws:iam::${data.aws_caller_identity.current.account_id}:root"
      ]
      type = "AWS"
    }

    actions = [
      "kms:*"
    ]
    resources = ["*"]
    }

    statement {
    sid    = "Allow spot instances use of the customer managed key"
    effect = "Allow"

    principals {
      identifiers = ["arn:aws:iam::${var.aws_account_id}:role/aws-service-role/autoscaling.amazonaws.com/AWSServiceRoleForAutoScaling"]
      type        = "AWS"
    }

    actions = ["kms:Encrypt",
      "kms:Decrypt",
      "kms:ReEncrypt*",
      "kms:GenerateDataKey*",
      "kms:DescribeKey"
    ]
    resources = ["*"]
  }

  statement {
    sid    = "Allow attachment of persistent resources"
    effect = "Allow"

    principals {
      identifiers = ["arn:aws:iam::${var.aws_account_id}:role/aws-service-role/autoscaling.amazonaws.com/AWSServiceRoleForAutoScaling"]
      type        = "AWS"
    }

    actions   = ["kms:CreateGrant"]
    resources = ["*"]
    condition {
      test     = "Bool"
      variable = "kms:GrantIsForAWSResource"
      values   = ["true"]
    }
  }
}

Connect To The Bastion Host

The Session Manager Plugin is needed to connect via SSM to the bastion host.

AWS-Gate

AWS-Gate.

AWS CLI

instance_id="my-bastion-instance-id"
az="az-of-the-bastion"

export AWS_ACCESS_KEY_ID="xxxxx"
export AWS_SECRET_ACCESS_KEY="yyyyy"
export AWS_SESSION_TOKEN=""

aws sts assume-role --role-arn the-bastion-role-arn --role-session-profile bastion
echo "export the credentials from above!"

echo -e 'y\n' | ssh-keygen -t rsa -f bastion_key -N '' >/dev/null 2>&1
ssh_public_key=$(cat bastion_key.pub)

aws ec2-instance-connect send-ssh-public-key --instance-id "${instance_id}" --availability-zone "${az}" \
      --instance-os-user ec2-user --ssh-public-key "${ssh_public_key}"

ssh "ec2-user@${instance_id}" -i bastion_key -N -L "12345:my.cloud.http:80" -o "UserKnownHostsFile=/dev/null" \
    -o "StrictHostKeyChecking=no" -o ProxyCommand="aws ssm start-session --target %h --document AWS-StartSSHSession \
      --parameters portNumber=%p"

curl http://localhost:12345/

AWS CLI With Menu

1. Export the AWS credentials for the user able to connect to the bastion host.
2. Execute scripts/connect_bastion.sh. Make sure to add the port forwarding and change the role ARN and bastion instance name.
3. Access the forwarded service through the local port.

Direct access to the bastion host is not granted but the specified port is forwarded. This way you can access the database, Redis cluster, ... directly from your localhost.

A Bastion Host

• allows access to the infrastructure which is not exposed to the internet
• designed to withstand attacks
• also known as jump host

Wikipedia

Module Documentation

Requirements

Providers

Modules

Resources

Inputs

object({
    type              = string # EC2 instance type
    desired_capacity  = number # number of EC2 instances to run
    root_volume_size  = number # in GB
    enable_monitoring = bool
    enable_spot       = bool
    profile_name      = string
  })

{
  "desired_capacity": 1,
  "enable_monitoring": false,
  "enable_spot": false,
  "profile_name": "",
  "root_volume_size": 8,
  "type": "t3.nano"
}

object({
    on_demand_base_capacity                  = number # absolute minimum amount of on_demand instances
    on_demand_percentage_above_base_capacity = number # percentage split between on-demand and Spot instances
    spot_allocation_strategy                 = string
  })

{
  "on_demand_base_capacity": 0,
  "on_demand_percentage_above_base_capacity": 0,
  "spot_allocation_strategy": "lowest-price"
}

object({
    prefix    = string
    separator = string
  })

{
  "prefix": "bastion",
  "separator": "-"
}

object({
    start     = string
    stop      = string
    time_zone = string
  })

Outputs