Approval Steps Inside Orb

Build Environment
1

Is it possible to have an approval step inside of an orb action like the following?

commands:
  invoke:
   steps:
     - when:
          condition:
            equal: [ true, << parameters.wait_for_deployment >> ]
          steps:
            - run:
                name: Waiting For Deployment To Notify Finished
                command: |
                  echo "This is an approval step that the deployment will auto notify when done."
                  echo "WORKFLOW ID: $CIRCLECI_WORKFLOW_ID"
                  echo "JOB ID: $CIRCLE_WORKFLOW_JOB_ID"
2

I know you can have an approval step inside a “workflow” but as you can see this is a “command”, so there would need to be an approval step inside of this individual step.

3

I could be wrong, but I think not, for the reason you mentioned - manual approval steps are part of workflows vs. jobs or commands.

4

That’s too bad as that’s extremely limiting as you can’t have workflows inside built orbs, which was the goal here. Oh well. I’ll work around it.

5

What you seem to be describing is some form of event system/queue to allow a process to continue rather than a manual approval step that is part of the provided environment.

I can only guess, but it is likely that CircleCI did not want to provide manual approval features from within an active resource just consider a X-large instance that is costing maybe hundreds of credits each minute it is on hold waiting for someone to click the approve button.

As for a basic event system I have tried to put something together for my own needs as I have a nighly CI task that deploys a complete production environment starting from the creation of the underlying VM. Currently in production, this just uses a sleep command to wait 40 minutes, but I have also been trying the following

  • Deploy or sign up for some form of Key-Value service that can be accessed by all parts of the system. I’ve been using Doppler as I use it for other tasks.

  • Have your workflow start the external process, such as the deployment

  • Within the workflow you can then look for a known Key-Value to change in some way, while also managing a time-out solution.

  • The external process sets the Key-Value pair to the required value or status code when finished/failed.

  • The workflow sees the Key-Value change and acts on the result.

All of this depends on some type of third-party solution, but Doppler has a free tier to work with, and depending on the tools you already use Amazon’s dynamodb offering could also be an option.

6

I’m actually putting something together right now. Basically the pipeline will trigger the deploy and go into a waiting (approval) step. The service that is doing the deploying will run a job that notifies back to the pipeline when it succeed by simply approving that approval step. If the deployment fails it cancels the pipeline.