»Packer Commands (CLI)

Packer is controlled using a command-line interface. All interaction with Packer is done via the packer tool. Like many other command-line tools, the packer tool takes a subcommand to execute, and that subcommand may have additional options as well. Subcommands are executed with packer SUBCOMMAND, where "SUBCOMMAND" is the actual command you wish to execute.

If you run packer by itself, help will be displayed showing all available subcommands and a brief synopsis of what they do. In addition to this, you can run any packer command with the -h flag to output more detailed help for a specific subcommand.

In addition to the documentation available on the command-line, each command is documented on this website. You can find the documentation for a specific subcommand using the navigation to the left.

»Machine-Readable Output

By default, the output of Packer is very human-readable. It uses nice formatting, spacing, and colors in order to make Packer a pleasure to use. However, Packer was built with automation in mind. To that end, Packer supports a fully machine-readable output setting, allowing you to use Packer in automated environments.

Because the machine-readable output format was made with Unix tools in mind, it is awk/sed/grep/etc. friendly and provides a familiar interface without requiring you to learn a new format.

»Enabling Machine-Readable Output

The machine-readable output format can be enabled by passing the -machine-readable flag to any Packer command. This immediately enables all output to become machine-readable on stdout. Logging, if enabled, continues to appear on stderr. An example of the output is shown below:

$ packer -machine-readable version
1498365963,,ui,say,Packer v1.0.2

The format will be covered in more detail later. But as you can see, the output immediately becomes machine-friendly. Try some other commands with the -machine-readable flag to see!

~>; The -machine-readable flag is designed for automated environments and is mutually-exclusive with the -debug flag, which is designed for interactive environments.

»Format for Machine-Readable Output

The machine readable format is a line-oriented, comma-delimited text format. This makes it more convenient to parse using standard Unix tools such as awk or grep in addition to full programming languages like Ruby or Python.

The format is:


Each component is explained below:

  • timestamp is a Unix timestamp in UTC of when the message was printed.

  • target When you call packer build this can be either empty or individual build names, e.g. amazon-ebs. It is normally empty when builds are in progress, and the build name when artifacts of particular builds are being referred to.

  • type is the type of machine-readable message being outputted. The two most common types are ui and artifact

  • data is zero or more comma-separated values associated with the prior type. The exact amount and meaning of this data is type-dependent, so you must read the documentation associated with the type to understand fully.

Within the format, if data contains a comma, it is replaced with %!(PACKER_COMMA). This was preferred over an escape character such as \' because it is more friendly to tools like awk.

Newlines within the format are replaced with their respective standard escape sequence. Newlines become a literal \n within the output. Carriage returns become a literal \r.

»Machine-Readable Message Types

Here's an incomplete list of types you may see in the machine-readable output:

You'll see these data types when you run packer build:

  • ui: this means that the information being provided is a human-readable string that would be sent to stdout even if we aren't in machine-readable mode. There are three "data" subtypes associated with this type:

    • say: in a non-machine-readable format, this would be bolded. Normally it is used for anouncements about beginning new steps in the build process

    • message: the most commonly used message type, used for basic updates during the build process.

    • error: reserved for errors

  • artifact-count: This data type tells you how many artifacts a particular build produced.

  • artifact: This data type tells you information about what Packer created during its build. An example of output follows the pattern timestamp, buildname, artifact, artifact_number, key, value where key and value contain information about the artifact.

    For example:

      1539967803,,ui,say,\n==> Builds finished. The artifacts of successful builds are:
      1539967803,amazon-ebs,artifact,0,string,AMIs were created:\neu-west-1: ami-04d23aca8bdd36e30\n
      1539967803,,ui,say,--> amazon-ebs: AMIs were created:\neu-west-1: ami-04d23aca8bdd36e30\n
      2018/10/19 09:50:03 waiting for all plugin processes to complete...

You'll see these data types when you run packer version:

  • version: what version of Packer is running

  • version-prerelease: Data will contain dev if version is prerelease, and otherwise will be blank.

  • version-commit: The git hash for the commit that the branch of Packer is currently on; most useful for Packer developers.


The packer command features opt-in subcommand autocompletion that you can enable for your shell with packer -autocomplete-install. After doing so, you can invoke a new shell and use the feature.

For example, assume a tab is typed at the end of each prompt line:

$ packer p
plugin  build
$ packer build -
-color             -debug             -except            -force             -machine-readable  -on-error          -only              -parallel          -timestamp          -var               -var-file