Fetches terraform outputs and lets you:
- print them to stdout in different formats
- call another command with them exposed in the environment
- export them to your current shell
You need to install this module and then run it's command from somewhere that has a property initialized terraform environment.
npm install -g tf-output
Assuming you have a module called database.tf with an output called DATABASE_URL
in a folder called terraform/database and you have already set up terraform...
tf-output database
Would print out something like:
DATABASE_URL="http://database.totallysecure.com"
And:
tf-output database -- node src/app.js
Would call node src/app.js
with the DATABASE_URL
available in the process environment.
You can also
export $(tf-output database)
Which will populate outputs in to your current shell.
-m
or --module
specifies which module to obtain output from. -m
alone uses the dir
name as the value. So
tf-output api -m
calls terraform output -m api
in terraform/api
-f
or --format
allows you to specify an alternate output format, currently json
is supported. This flag is ignored if a command is specified.
-a
or --auto-init
allows you to auto-initialize terraform (terraform init
) prior to copying outputs.
-g
or --auto-init-get
will pull down modules mentioned in the root module during auto-initialization.
-i
or --init-opts
allows you to pass additional options to terraform init
(e.g. -backend-config
).
-fl
or --flatten
will flatten output values that are objects by concatenating key names.
-fd
or --flatten-delimiter
allows you to specify the delimiter to use while concatenating key names (_
by default).
-c
or --check-plan
checks if terraform plan has any unapplied changes (and aborts if it does).
-o
or --plan-opts
allows you to pass additional options to terraform plan
(e.g. -var-file
).
-p
or -path
specifies a path template. tf-output looks for modules according to a path template, which by default is terraform/{dir}
- so tf-output api database
would look for a module in two directories:
- terraform/database
- terraform/api
You can use more complex path templates. Imagine you deploy your app using the following command:
deploy --stage=dev --region=us-east-1
You might organize your terraform definitions in a number of ways and you can customize where tf-output looks to suit your needs.
tf-output database -p {stage}/{region}/terraform/{dir}
This command would fail, because while it knows the module you are trying to load, it doesn't know what stage
and region
should be.
You can specify them:
tf-output database api -p {stage}/{region}/terraform/{dir} --stage=dev --region=us-east-1
This would cause tf-output to load modules from:
dev/us-east-1/terraform/api
dev/us-east-1/terraform/database
If you run a command through tf-output, it will look ahead for substitution values. This works the same:
tf-output database api -p {stage}/{region}/terraform/{dir} -- deploy --stage=dev --region=us-east-1
Except instead of printing the outputs out to stdout, it would exec deploy --stage=dev --region=us-east-1
with environment variables set, and the path template would still be substituted with the right stage and region. This saves you havin to repeat region, stage, etc.
In many cases the arguments you will specify to tf-output
are always going to be the same, for example the -p
path template argument. You can put these in a .tfoutput file which will be read on every execution.
Example .tfoutput
file:
{
"path": "terraform/{dir}/{stage}/{region}",
"module": true,
"auto-init": true,
"auto-init-get": true
}