packer-cn/website/source/docs/provisioners/file.html.md

---
description: |
    The file Packer provisioner uploads files to machines built by Packer. The
    recommended usage of the file provisioner is to use it to upload files, and
    then use shell provisioner to move them to the proper place, set permissions,
    etc.
layout: docs
page_title: 'File - Provisioners'
sidebar_current: 'docs-provisioners-file'
---

# File Provisioner

Type: `file`

The file Packer provisioner uploads files to machines built by Packer. The
recommended usage of the file provisioner is to use it to upload files, and then
use [shell provisioner](/docs/provisioners/shell.html) to move them to the
proper place, set permissions, etc.

The file provisioner can upload both single files and complete directories.

## Basic Example

``` json
{
  "type": "file",
  "source": "app.tar.gz",
  "destination": "/tmp/app.tar.gz"
}
```

## Configuration Reference

The available configuration options are listed below. All elements are required.

-   `source` (string) - The path to a local file or directory to upload to
    the machine. The path can be absolute or relative. If it is relative, it is
    relative to the working directory when Packer is executed. If this is a
    directory, the existence of a trailing slash is important. Read below on
    uploading directories.

-   `destination` (string) - The path where the file will be uploaded to in
    the machine. This value must be a writable location and any parent
    directories must already exist.

-   `direction` (string) - The direction of the file transfer. This defaults to
    "upload". If it is set to "download" then the file "source" in the machine
    will be downloaded locally to "destination"

-   `generated` (boolean) - For advanced users only. If true, check the file
    existence only before uploading, rather than upon pre-build validation.
    This allows to upload files created on-the-fly. This defaults to false. We
    don't recommend using this feature, since it can cause Packer to become
    dependent on system state. We would prefer you generate your files before
    the Packer run, but realize that there are situations where this may be
    unavoidable.

## Directory Uploads

The file provisioner is also able to upload a complete directory to the remote
machine. When uploading a directory, there are a few important things you should
know.

First, the destination directory must already exist. If you need to create it,
use a shell provisioner just prior to the file provisioner in order to create
the directory. If the destination directory does not exist, the file
provisioner may succeed, but it will have undefined results.

Next, the existence of a trailing slash on the source path will determine
whether the directory name will be embedded within the destination, or whether
the destination will be created. An example explains this best:

If the source is `/foo` (no trailing slash), and the destination is `/tmp`, then
the contents of `/foo` on the local machine will be uploaded to `/tmp/foo` on
the remote machine. The `foo` directory on the remote machine will be created by
Packer.

If the source, however, is `/foo/` (a trailing slash is present), and the
destination is `/tmp`, then the contents of `/foo` will be uploaded into `/tmp`
directly.

This behavior was adopted from the standard behavior of rsync. Note that under
the covers, rsync may or may not be used.

## Uploading files that don't exist before Packer starts

In general, local files used as the source **must** exist before Packer is run.
This is great for catching typos and ensuring that once a build is started,
that it will succeed. However, this also means that you can't generate a file
during your build and then upload it using the file provisioner later.
A convenient workaround is to upload a directory instead of a file. The
directory still must exist, but its contents don't. You can write your
generated file to the directory during the Packer run, and have it be uploaded
later.

## Symbolic link uploads

The behavior when uploading symbolic links depends on the communicator. The
Docker communicator will preserve symlinks, but all other communicators will
treat local symlinks as regular files. If you wish to preserve symlinks when
uploading, it's recommended that you use `tar`. Below is an example of what
that might look like:

``` text
$ ls -l files
total 16
drwxr-xr-x  3 mwhooker  staff  102 Jan 27 17:10 a
lrwxr-xr-x  1 mwhooker  staff    1 Jan 27 17:10 b -> a
-rw-r--r--  1 mwhooker  staff    0 Jan 27 17:10 file1
lrwxr-xr-x  1 mwhooker  staff    5 Jan 27 17:10 file1link -> file1
$ ls -l toupload
total 0
-rw-r--r--  1 mwhooker  staff    0 Jan 27 17:10 files.tar
```

``` json
{
  "provisioners": [
    {
      "type": "shell-local",
      "command": "tar cf toupload/files.tar files"
    },
    {
      "destination": "/tmp/",
      "source": "./toupload",
      "type": "file"
    },
    {
      "inline": [
        "cd /tmp && tar xf toupload/files.tar",
        "rm toupload/files.tar"
      ],
      "type": "shell"
    }
  ]
}
```

## Slowness when transferring large files over WinRM.

Because of the way our WinRM transfers works, it can take a very long time to
upload and download even moderately sized files. If you're experiencing
slowness using the file provisioner on Windows, it's suggested that you set up
an SSH server and use the [ssh
communicator](/docs/templates/communicator.html#ssh-communicator). If you only
want to transfer files to your guest, and if your builder supports it, you may
also use the `http_directory` directive. This will cause that directory to be
available to the guest over http, and set the environment variable
`PACKER_HTTP_ADDR` to the address.