Dependencies

Once you have a few pipelines running to deploy infrastructure and services, they'll almost certainly need to share data. Deploying a service may require details of the cluster where it is to be deployed. A frontend serverless function may need to know the URL of a backend service. In an event driven system, you may need to make sure the producer exists before the consumer.

To handle these challenges, Ocuroot allows you to define dependencies between release pipelines. Each Call or Deployment has a set of Inputs and Outputs that can be connected with the Inputs and Outputs of others.

Declaring Inputs

Both the call and deploy functions accept an inputs parameter, which is a dictionary of input names to either static values, ref, or input objects.

For example, the below creates an input called image_tag that is populated from the tag output of the most recent build call:

inputs={
    "image_tag": ref("./@/call/build#output/tag"),
},

The ref is declared as relative to the current config file. The fragment #output/tag will point directly to the tag value in the outputs from this build. These outputs would have been returned from the build function:

def build(ctx):
    # ...
    tag = "myapp:{}".format(build_number)
    return done(
        output={
            "tag": tag,
        }
    )

You can also declare input values with both a ref and a default value. This allows for a work item to refer to itself.

inputs={
    "prev_build_number": input(ref="./call/build#output/build_number", default=0),
},

Note the lack of an @ portion in the ref above. This is because it is relative to the release where it is evaluated.

How Dependencies are Resolved

After completing any work item, Ocuroot will check for any inputs that depend on its outputs. If the value of the output is different than the most recent value for any existing deployments, they will be re-deployed with the new value.