It seems like I can use either user_data
with a template file or a \"remote-exec\" provisioner
with inline commands to bootstrap. So which one is consi
You should use user_data
. The user data field is idiomatic because it's native to AWS, whereas the remote-exec provisioner is specific to Terraform, which is just one of many ways to call the AWS API.
Also, the user-data is viewable in the AWS console, and often an important part of using Auto Scaling Groups in AWS, where you want each EC2 Instance to execute the same config code when it launches. It's not possible to do that with Terraform's remote-exec provisioner.
Though I do agree with Josh, if there are no run time changes to the instance you can use packer to build an ami and then use that in the launch config. That way you don't have to wait for user-data to run.
Packer is part of the Hashicorp family of tools
https://www.packer.io/docs/builders/amazon-ebs.html