» AMI Builder (chroot)

Type: amazon-chroot

The amazon-chroot Packer builder is able to create Amazon AMIs backed by an EBS volume as the root device. For more information on the difference between instance storage and EBS-backed instances, see the "storage for the root device" section in the EC2 documentation.

The difference between this builder and the amazon-ebs builder is that this builder is able to build an EBS-backed AMI without launching a new EC2 instance. This can dramatically speed up AMI builds for organizations who need the extra fast build.

The builder does not manage AMIs. Once it creates an AMI and stores it in your account, it is up to you to use, delete, etc., the AMI.

» How Does it Work?

This builder works by creating a new EBS volume from an existing source AMI and attaching it into an already-running EC2 instance. Once attached, a chroot is used to provision the system within that volume. After provisioning, the volume is detached, snapshotted, and an AMI is made.

Using this process, minutes can be shaved off the AMI creation process because a new EC2 instance doesn't need to be launched.

There are some restrictions, however. The host EC2 instance where the volume is attached to must be a similar system (generally the same OS version, kernel versions, etc.) as the AMI being built. Additionally, this process is much more expensive because the EC2 instance must be kept running persistently in order to build AMIs, whereas the other AMI builders start instances on-demand to build AMIs as needed.

» Configuration Reference

There are many configuration options available for the builder. They are segmented below into two categories: required and optional parameters. Within each category, the available configuration keys are alphabetized.

» Required:

  • access_key (string) - The access key used to communicate with AWS. Learn how to set this

  • ami_name (string) - The name of the resulting AMI that will appear when managing AMIs in the AWS console or via APIs. This must be unique. To help make this unique, use a function like timestamp (see template engine for more info).

  • secret_key (string) - The secret key used to communicate with AWS. Learn how to set this

  • source_ami (string) - The source AMI whose root volume will be copied and provisioned on the currently running instance. This must be an EBS-backed AMI with a root volume snapshot that you have access to. Note: this is not used when from_scratch is set to true.

» Optional:

  • ami_architecture (string) - what architecture to use when registering the final AMI; valid options are "x86_64" or "arm64". Defaults to "x86_64".

  • ami_description (string) - The description to set for the resulting AMI(s). By default this description is empty. This is a template engine, see Build template data for more information.

  • ami_groups (array of strings) - A list of groups that have access to launch the resulting AMI(s). By default no groups have permission to launch the AMI. all will make the AMI publicly accessible.

  • ami_product_codes (array of strings) - A list of product codes to associate with the AMI. By default no product codes are associated with the AMI.

  • ami_regions (array of strings) - A list of regions to copy the AMI to. Tags and attributes are copied along with the AMI. AMI copying takes time depending on the size of the AMI, but will generally take many minutes.

  • ami_users (array of strings) - A list of account IDs that have access to launch the resulting AMI(s). By default no additional users other than the user creating the AMI has permissions to launch it.

  • ami_virtualization_type (string) - The type of virtualization for the AMI you are building. This option is required to register HVM images. Can be paravirtual (default) or hvm.

  • chroot_mounts (array of array of strings) - This is a list of devices to mount into the chroot environment. This configuration parameter requires some additional documentation which is in the Chroot Mounts section. Please read that section for more information on how to use this.

  • command_wrapper (string) - How to run shell commands. This defaults to {{.Command}}. This may be useful to set if you want to set environmental variables or perhaps run it with sudo or so on. This is a configuration template where the .Command variable is replaced with the command to be run. Defaults to {{.Command}}.

  • copy_files (array of strings) - Paths to files on the running EC2 instance that will be copied into the chroot environment prior to provisioning. Defaults to /etc/resolv.conf so that DNS lookups work. Pass an empty list to skip copying /etc/resolv.conf. You may need to do this if you're building an image that uses systemd.

  • custom_endpoint_ec2 (string) - This option is useful if you use a cloud provider whose API is compatible with aws EC2. Specify another endpoint like this https://ec2.custom.endpoint.com.

  • decode_authorization_messages (boolean) - Enable automatic decoding of any encoded authorization (error) messages using the sts:DecodeAuthorizationMessage API. Note: requires that the effective user/role have permissions to sts:DecodeAuthorizationMessage on resource *. Default false.

  • device_path (string) - The path to the device where the root volume of the source AMI will be attached. This defaults to "" (empty string), which forces Packer to find an open device automatically.

  • ena_support (boolean) - Enable enhanced networking (ENA but not SriovNetSupport) on HVM-compatible AMIs. If set, add ec2:ModifyInstanceAttribute to your AWS IAM policy. If false, this will disable enhanced networking in the final AMI as opposed to passing the setting through unchanged from the source. Note: you must make sure enhanced networking is enabled on your instance. See Amazon's documentation on enabling enhanced networking.

  • encrypt_boot (boolean) - Whether or not to encrypt the resulting AMI when copying a provisioned instance to an AMI. By default, Packer will keep the encryption setting to what it was in the source image. Setting false will result in an unencrypted image, and true will result in an encrypted one.

  • force_deregister (boolean) - Force Packer to first deregister an existing AMI if one with the same name already exists. Default false.

  • force_delete_snapshot (boolean) - Force Packer to delete snapshots associated with AMIs, which have been deregistered by force_deregister. Default false.

  • insecure_skip_tls_verify (boolean) - This allows skipping TLS verification of the AWS EC2 endpoint. The default is false.

  • kms_key_id (string) - ID, alias or ARN of the KMS key to use for boot volume encryption. This only applies to the main region, other regions where the AMI will be copied will be encrypted by the default EBS KMS key. For valid formats see KmsKeyId in the AWS API docs - CopyImage. This field is validated by Packer, when using an alias, you will have to prefix kms_key_id with alias/.
  • from_scratch (boolean) - Build a new volume instead of starting from an existing AMI root volume snapshot. Default false. If true, source_ami is no longer used and the following options become required: ami_virtualization_type, pre_mount_commands and root_volume_size. The below options are also required in this mode only:

  • ami_block_device_mappings (array of block device mappings) - Add one or more block device mappings to the AMI. These will be attached when booting a new instance from your AMI. If this field is populated, and you are building from an existing source image, the block device mappings in the source image will be overwritten. This means you must have a block device mapping entry for your root volume, root_volume_size, and root_device_name. `Your options here may vary depending on the type of VM you use. The block device mappings allow for the following configuration:

    • delete_on_termination (boolean) - Indicates whether the EBS volume is deleted on instance termination. Default false. NOTE: If this value is not explicitly set to true and volumes are not cleaned up by an alternative method, additional volumes will accumulate after every build.

    • device_name (string) - The device name exposed to the instance (for example, /dev/sdh or xvdh). Required for every device in the block device mapping.

    • encrypted (boolean) - Indicates whether or not to encrypt the volume. By default, Packer will keep the encryption setting to what it was in the source image. Setting false will result in an unencrypted device, and true will result in an encrypted one.

    • iops (number) - The number of I/O operations per second (IOPS) that the volume supports. See the documentation on IOPs for more information

    • kms_key_id (string) - The ARN for the KMS encryption key. When specifying kms_key_id, encrypted needs to be set to true. For valid formats see KmsKeyId in the AWS API docs - CopyImage.
    • no_device (boolean) - Suppresses the specified device included in the block device mapping of the AMI.

    • snapshot_id (string) - The ID of the snapshot.

    • virtual_name (string) - The virtual device name. See the documentation on Block Device Mapping for more information.

    • volume_size (number) - The size of the volume, in GiB. Required if not specifying a snapshot_id.

    • volume_type (string) - The volume type. gp2 for General Purpose (SSD) volumes, io1 for Provisioned IOPS (SSD) volumes, st1 for Throughput Optimized HDD, sc1 for Cold HDD, and standard for Magnetic volumes.

  • region_kms_key_ids (map of strings) - a map of regions to copy the ami to, along with the custom kms key id (alias or arn) to use for encryption for that region. Keys must match the regions provided in ami_regions. If you just want to encrypt using a default ID, you can stick with kms_key_id and ami_regions. If you want a region to be encrypted with that region's default key ID, you can use an empty string "" instead of a key id in this map. (e.g. "us-east-1": "") However, you cannot use default key IDs if you are using this in conjunction with snapshot_users -- in that situation you must use custom keys. For valid formats see KmsKeyId in the AWS API docs - CopyImage.

    This option supercedes the kms_key_id option -- if you set both, and they are different, Packer will respect the value in region_kms_key_ids for your build region and silently disregard the value provided in kms_key_id.

  • root_device_name (string) - The root device name. For example, xvda.

  • mfa_code (string) - The MFA TOTP code. This should probably be a user variable since it changes all the time.

  • mount_path (string) - The path where the volume will be mounted. This is where the chroot environment will be. This defaults to /mnt/packer-amazon-chroot-volumes/{{.Device}}. This is a configuration template where the .Device variable is replaced with the name of the device where the volume is attached.

  • mount_partition (string) - The partition number containing the / partition. By default this is the first partition of the volume, (for example, xvda1) but you can designate the entire block device by setting "mount_partition": "0" in your config, which will mount xvda instead.

  • mount_options (array of strings) - Options to supply the mount command when mounting devices. Each option will be prefixed with -o and supplied to the mount command ran by Packer. Because this command is ran in a shell, user discretion is advised. See this manual page for the mount command for valid file system specific options.

  • nvme_device_path (string) - When we call the mount command (by default mount -o device dir), the string provided in nvme_mount_path will replace device in that command. When this option is not set, device in that command will be something like /dev/sdf1, mirroring the attached device name. This assumption works for most instances but will fail with c5 and m5 instances. In order to use the chroot builder with c5 and m5 instances, you must manually set nvme_device_path and device_path.

  • pre_mount_commands (array of strings) - A series of commands to execute after attaching the root volume and before mounting the chroot. This is not required unless using from_scratch. If so, this should include any partitioning and filesystem creation commands. The path to the device is provided by {{.Device}}.

  • profile (string) - The profile to use in the shared credentials file for AWS. See Amazon's documentation on specifying profiles for more details.

  • post_mount_commands (array of strings) - As pre_mount_commands, but the commands are executed after mounting the root device and before the extra mount and copy steps. The device and mount path are provided by {{.Device}} and {{.MountPath}}.

  • root_volume_size (number) - The size of the root volume in GB for the chroot environment and the resulting AMI. Default size is the snapshot size of the source_ami unless from_scratch is true, in which case this field must be defined.

  • root_volume_type (string) - The type of EBS volume for the chroot environment and resulting AMI. The default value is the type of the source_ami, unless from_scratch is true, in which case the default value is gp2. You can only specify io1 if building based on top of a source_ami which is also io1.

  • root_volume_tags (object of key/value strings) - Tags to apply to the volumes that are launched. This is a template engine, see Build template data for more information.

  • skip_region_validation (boolean) - Set to true if you want to skip validation of the ami_regions configuration option. Default false.

  • snapshot_tags (object of key/value strings) - Tags to apply to snapshot. They will override AMI tags if already applied to snapshot. This is a template engine, see Build template data for more information.

  • snapshot_groups (array of strings) - A list of groups that have access to create volumes from the snapshot(s). By default no groups have permission to create volumes from the snapshot(s). all will make the snapshot publicly accessible.

  • snapshot_users (array of strings) - A list of account IDs that have access to create volumes from the snapshot(s). By default no additional users other than the user creating the AMI has permissions to create volumes from the backing snapshot(s).

  • source_ami_filter (object) - Filters used to populate the source_ami field. Example:

    "source_ami_filter": {
      "filters": {
        "virtualization-type": "hvm",
        "name": "ubuntu/images/*ubuntu-xenial-16.04-amd64-server-*",
        "root-device-type": "ebs"
      "owners": ["099720109477"],
      "most_recent": true

    This selects the most recent Ubuntu 16.04 HVM EBS AMI from Canonical. NOTE: This will fail unless exactly one AMI is returned. In the above example, most_recent will cause this to succeed by selecting the newest image.

    • filters (map of strings) - filters used to select a source_ami. NOTE: This will fail unless exactly one AMI is returned. Any filter described in the docs for DescribeImages is valid.
    • owners (array of strings) - Filters the images by their owner. You may specify one or more AWS account IDs, "self" (which will use the account whose credentials you are using to run Packer), or an AWS owner alias: for example, "amazon", "aws-marketplace", or "microsoft". This option is required for security reasons.
    • most_recent (boolean) - Selects the newest created image when true. This is most useful for selecting a daily distro build.

    You may set this in place of source_ami or in conjunction with it. If you set this in conjunction with source_ami, the source_ami will be added to the filter. The provided source_ami must meet all of the filtering criteria provided in source_ami_filter; this pins the AMI returned by the filter, but will cause Packer to fail if the source_ami does not exist.

  • sriov_support (boolean) - Enable enhanced networking (SriovNetSupport but not ENA) on HVM-compatible AMIs. If true, add ec2:ModifyInstanceAttribute to your AWS IAM policy. Note: you must make sure enhanced networking is enabled on your instance. See Amazon's documentation on enabling enhanced networking. Default false.

  • tags (object of key/value strings) - Tags applied to the AMI. This is a template engine, see Build template data for more information.

  • vault_aws_engine (object) - Get credentials from Hashicorp Vault's aws secrets engine. You must already have created a role to use. For more information about generating credentials via the Vault engine, see the Vault docs. If you set this flag, you must also set the below options:

    • name (string) - Required. Specifies the name of the role to generate credentials against. This is part of the request URL.
    • engine_name (string) - The name of the aws secrets engine. In the Vault docs, this is normally referred to as "aws", and Packer will default to "aws" if engine_name is not set.
    • role_arn (string)- The ARN of the role to assume if credential_type on the Vault role is assumed_role. Must match one of the allowed role ARNs in the Vault role. Optional if the Vault role only allows a single AWS role ARN; required otherwise.
    • ttl (string) - Specifies the TTL for the use of the STS token. This is specified as a string with a duration suffix. Valid only when credential_type is assumed_role or federation_token. When not specified, the default_sts_ttl set for the role will be used. If that is also not set, then the default value of 3600s will be used. AWS places limits on the maximum TTL allowed. See the AWS documentation on the DurationSeconds parameter for AssumeRole (for assumed_role credential types) and GetFederationToken (for federation_token credential types) for more details.

      Example: json { "vault_aws_engine": { "name": "myrole", "role_arn": "myarn", "ttl": "3600s" } }

» Basic Example

Here is a basic example. It is completely valid except for the access keys:

  "type": "amazon-chroot",
  "access_key": "YOUR KEY HERE",
  "secret_key": "YOUR SECRET KEY HERE",
  "source_ami": "ami-e81d5881",
  "ami_name": "packer-amazon-chroot {{timestamp}}"

» Chroot Mounts

The chroot_mounts configuration can be used to mount specific devices within the chroot. By default, the following additional mounts are added into the chroot by Packer:

  • /proc (proc)
  • /sys (sysfs)
  • /dev (bind to real /dev)
  • /dev/pts (devpts)
  • /proc/sys/fs/binfmt_misc (binfmt_misc)

These default mounts are usually good enough for anyone and are sane defaults. However, if you want to change or add the mount points, you may using the chroot_mounts configuration. Here is an example configuration which only mounts /proc and /dev:

  "chroot_mounts": [
    ["proc", "proc", "/proc"],
    ["bind", "/dev", "/dev"]

chroot_mounts is a list of a 3-tuples of strings. The three components of the 3-tuple, in order, are:

  • The filesystem type. If this is "bind", then Packer will properly bind the filesystem to another mount point.

  • The source device.

  • The mount directory.

» Parallelism

A quick note on parallelism: it is perfectly safe to run multiple separate Packer processes with the amazon-chroot builder on the same EC2 instance. In fact, this is recommended as a way to push the most performance out of your AMI builds.

Packer properly obtains a process lock for the parallelism-sensitive parts of its internals such as finding an available device.

» Gotchas

» Unmounting the Filesystem

One of the difficulties with using the chroot builder is that your provisioning scripts must not leave any processes running or packer will be unable to unmount the filesystem.

For debian based distributions you can setup a policy-rc.d file which will prevent packages installed by your provisioners from starting services:

  "type": "shell",
  "inline": [
    "echo '#!/bin/sh' > /usr/sbin/policy-rc.d",
    "echo 'exit 101' >> /usr/sbin/policy-rc.d",
    "chmod a+x /usr/sbin/policy-rc.d"

// ...

  "type": "shell",
  "inline": [
    "rm -f /usr/sbin/policy-rc.d"

» Ansible provisioner

Running ansible against amazon-chroot requires changing the Ansible connection to chroot and running Ansible as root/sudo.

» Using Instances with NVMe block devices.

In C5, C5d, M5, and i3.metal instances, EBS volumes are exposed as NVMe block devices reference. In order to correctly mount these devices, you have to do some extra legwork, involving the nvme_device_path option above. Read that for more information.

A working example for mounting an NVMe device is below:

  "variables": {
    "region" : "us-east-2"
  "builders": [
      "type": "amazon-chroot",
      "region": "{{user `region`}}",
      "source_ami_filter": {
        "filters": {
        "virtualization-type": "hvm",
        "name": "amzn-ami-hvm-*",
        "root-device-type": "ebs"
        "owners": ["137112412989"],
        "most_recent": true
      "ena_support": true,
      "ami_name": "amazon-chroot-test-{{timestamp}}",
      "nvme_device_path": "/dev/nvme1n1p",
      "device_path": "/dev/sdf"

  "provisioners": [
      "type": "shell",
      "inline": ["echo Test > /tmp/test.txt"]

Note that in the nvme_device_path you must end with the p; if you try to define the partition in this path (e.g. nvme_device_path: /dev/nvme1n1p1) and haven't also set the "mount_partition": 0, a 1 will be appended to the nvme_device_path and Packer will fail.

» Building From Scratch

This example demonstrates the essentials of building an image from scratch. A 15G gp2 (SSD) device is created (overriding the default of standard/magnetic). The device setup commands partition the device with one partition for use as an HVM image and format it ext4. This builder block should be followed by provisioning commands to install the os and bootloader.

  "type": "amazon-chroot",
  "ami_name": "packer-from-scratch {{timestamp}}",
  "from_scratch": true,
  "ami_virtualization_type": "hvm",
  "pre_mount_commands": [
    "parted {{.Device}} mklabel msdos mkpart primary 1M 100% set 1 boot on print",
    "mkfs.ext4 {{.Device}}1"
  "root_volume_size": 15,
  "root_device_name": "xvda",
  "ami_block_device_mappings": [
      "device_name": "xvda",
      "delete_on_termination": true,
      "volume_type": "gp2"

» Build template data

In configuration directives marked as a template engine above, the following variables are available:

  • BuildRegion - The region (for example eu-central-1) where Packer is building the AMI.
  • SourceAMI - The source AMI ID (for example ami-a2412fcd) used to build the AMI.
  • SourceAMIName - The source AMI Name (for example ubuntu/images/ebs-ssd/ubuntu-xenial-16.04-amd64-server-20180306) used to build the AMI.
  • SourceAMITags - The source AMI Tags, as a map[string]string object.