Template User Variables
Note: This page is about older-style JSON Packer templates. JSON templates are still supported by the Packer core, but new features added to the Packer core may not be implemented for JSON templates. We recommend you transition to HCL templates as soon as is convenient for you, in order to have the best possible experience with Packer. To help you upgrade your templates, we have written an hcl2_upgrade command command.
User variables allow your templates to be further configured with variables from the command-line, environment variables, Vault, or files. This lets you parameterize your templates so that you can keep secret tokens, environment-specific data, and other types of information out of your templates. This maximizes the portability of the template.
Using user variables expects you to know how configuration templates work. If you don't know how configuration templates work yet, please read that page first.
Usage
In order to set a user variable, you must define it either within the
variables
section within your template, or using the command-line -var
or
-var-file
flags.
Even if you want a user variable to default to an empty string, it is best to explicitly define it. This explicitness helps reduce the time it takes for newcomers to understand what can be modified using variables in your template.
The variables
section is a key/value mapping of the user variable name to a
default value. A default value can be the empty string. An example is shown
below:
In the above example, the template defines two user variables: aws_access_key
and aws_secret_key
. They default to empty values. Later, the variables are
used within the builder we defined in order to configure the actual keys for
the Amazon builder.
If the default value is null
, then the user variable will be required. This
means that the user must specify a value for this variable or template
validation will fail.
User variables are used by calling the {{user}}
function in the form of
{{user 'variable'}}
. This function can be used in any value but type
within the template: in builders, provisioners, anywhere outside the variables
section.
User variables are available globally within the rest of the template.
Environment Variables
Environment variables can be used within your template using user variables.
The env
function is available only within the default value of a user
variable, allowing you to default a user variable to an environment variable.
An example is shown below:
This will default "my_secret" to be the value of the "MY_SECRET" environment variable (or an empty string if it does not exist).
Why can't I use environment variables elsewhere? User variables are
the single source of configurable input to a template. We felt that having
environment variables used anywhere in a template would confuse the user
about the possible inputs to a template. By allowing environment variables only
within default values for user variables, user variables remain as the single
source of input to a template that a user can easily discover using
packer inspect
.
Why can't I use ~
for home variable? ~
is an special variable
that is evaluated by shell during a variable expansion. As Packer doesn't run
inside a shell, it won't expand ~
.
Consul keys
Consul keys can be used within your template using the consul_key
function.
This function is available only within the default value of a user variable,
for reasons similar to environment variables above.
This will default soft_versions
to the value of the key
my_image/softs_versions/next
in consul.
The configuration for consul (address, tokens, ...) must be specified as environment variables, as specified in the Documentation.
Vault Variables
Secrets can be read from Vault and used within
your template as user variables. the vault
function is available only
within the default value of a user variable, allowing you to default a user
variable to a vault secret.
An example of using a v2 kv engine:
If you store a value in vault using vault kv put secret/hello foo=world
, you
can access it using the following template engine:
which will assign "my_secret": "world"
An example of using a v1 kv engine:
If you store a value in vault using:
You can access it using the following template engine:
This example accesses the Vault path secret/data/foo
and returns the value
stored at the key bar
, storing it as "my_secret".
In order for this to work, you must set the environment variables VAULT_TOKEN
and VAULT_ADDR
to valid values.
The api tool we use allows for more custom configuration of the Vault client via environment variables.
The full list of available environment variables is:
and detailed documentation for usage of each of those variables can be found here.
AWS Secrets Manager Variables
Secrets can be read from AWS Secrets Manager
and used within your template as user variables. The aws_secretsmanager
function is
available only within the default value of a user variable, allowing you to default
a user variable to an AWS Secrets Manager secret.
Plaintext Secrets
In the example above it is assumed that the secret globalpassword
is not
stored as a key pair but as a single non-JSON string value. Which the
aws_secretsmanager
function will return as a raw string.
Single Key Secrets
In the example above it is assumed that only one key is stored in
sample/app/password
if there are multiple keys stored in it then you need
to indicate the specific key you want to fetch as shown below.
Multiple Key Secrets
In order to use this function you have to configure valid AWS credentials using one of the following methods:
Using array values
Some templates call for array values. You can use template variables for these,
too. For example, the amazon-ebs
builder has a configuration parameter called
ami_regions
, which takes an array of regions that it will copy the AMI to.
You can parameterize this by using a variable that is a list of regions, joined
by a ,
. For example:
Setting Variables
Now that we covered how to define and use user variables within a template, the next important point is how to actually set these variables. Packer exposes two methods for setting user variables: from the command line or from a file.
From the Command Line
To set user variables from the command line, the -var
flag is used as a
parameter to packer build
(and some other commands). Continuing our example
above, we could build our template using the command below. The command is
split across multiple lines for readability, but can of course be a single
line.
As you can see, the -var
flag can be specified multiple times in order to set
multiple variables. Also, variables set later on the command-line override any
earlier set variable of the same name.
warning If you are calling Packer from cmd.exe, you should double-quote your variables rather than single-quoting them. For example:
packer build -var "aws_secret_key=foo" template.json
From a File
Variables can also be set from an external JSON file. The -var-file
flag
reads a file containing a key/value mapping of variables to values and sets
those variables. An example JSON file may look like this:
It is a single JSON object where the keys are variables and the values are the
variable values. Assuming this file is in variables.json
, we can build our
template using the following command:
The -var-file
flag can be specified multiple times and variables from
multiple files will be read and applied. As you'd expect, variables read from
files specified later override a variable set earlier.
Combining the -var
and -var-file
flags together also works how you'd
expect. Variables set later in the command override variables set earlier. So,
for example, in the following command with the above variables.json
file:
Results in the following variables:
Variable | Value |
---|---|
aws_access_key | foo |
aws_secret_key | baz |
Sensitive Variables
If you use the environment to set a variable that is sensitive, you probably don't want that variable printed to the Packer logs. You can make sure that sensitive variables won't get printed to the logs by adding them to the "sensitive-variables" list within the Packer template:
The above snippet of code will function exactly the same as if you did not set
"sensitive-variables", except that the Packer UI and logs will replace all
instances of "bar" and of whatever the value of "my_secret" is with
<sensitive>
. This allows you to be confident that you are not printing
secrets in plaintext to our logs by accident.
Recipes
Making a provisioner step conditional on the value of a variable
There is no specific syntax in Packer templates for making a provisioner step
conditional, depending on the value of a variable. However, you may be able to
do this by referencing the variable within a command that you execute. For
example, here is how to make a shell-local
provisioner only run if the
do_nexpose_scan
variable is non-empty.
Using HOME Variable
In order to use $HOME
variable, you can create a home
variable in Packer:
And this will be available to be used in the rest of the template, i.e.: