Non-interactive environment variable setting (e.g. CI pipeline)

If I am using arc to deploy an app via a CI pipeline and I want to add environment variables to my Lambda, should I be doing this like so:

$ arc deploy
$ arc env -e <environment> --add FOO bar

If so, when this command runs in CI, it hangs because the arc env command is requesting input:

Env Would you like to create a local preferences file? [Y|n]

Is there a better way to deploy with these environment variables via CI, or is there a non-interactive mode? Should arc pick up on the presence of a CI environment variable from my pipeline and not prompt for such interactions?

[ryanblock]

@thescientist13 that’s right! If you need an example of how we do it: https://github.com/architect/arc.codes/blob/main/.github/workflows/build.yml#L89-L103

[ryanblock]

We can add a flag to suppress that behavior, for sure. However, how you’re doing it above is not correct or necessary. Here’s how env vars flow through the system (broadly):

• You add an env var with arc env command
• Architect saves that env var to your app’s SSM namespace
• You deploy your app with arc deploy
• Architect queries your app’s SSM namespace for all the env vars you added to a given environment (eg staging) prior to deployment, then adds them to your CloudFormation template
• Architect then deploys your app environment with its environment-specific env vars

Architect does not update live env vars in your running application when you run arc env. Those env vars don’t go live until the next deployment.

Hopefully that addresses your question as well, @thescientist13, but just to be clear: no secrets are uploaded to mysterious servers / services, Architect safely and securely stores your env vars in your app’s own secured SSM store within AWS, and those env vars are only accessed by your creds at deploy time.