Define an output that acts as a Knative eventing source

[afrittoli]

A CI/CD pipeline may include tasks that should be executed asynchronously from the main pipeline, for instance post processing of test results or logs, notifications and more; so things that should not affect the duration / reliability of the pipeline and can be executed "eventually" once the pipeline is complete.

This could be achieved by exploiting knative eventing. We could define a pipelineresource to be exclusively used as taskrun / pipelinerun output, that can store generic medatada about the *run and that can act an eventing source for a configured sink.

Examples of metadata are the location of the produced artifacts (a container image, logs, machine readable test results), the name of the task / pipeline, details about the triggering event (the git branch , github PR, release number, etc).

[bobcatfish]

A CI/CD pipeline may include tasks that should be executed asynchronously from the main pipeline

That's a cool idea @afrittoli !

Would the success/failure and completion of the PipelineRun depend on the completion of this processing as well?

[afrittoli]

That's a good question. I can see both behaviours as valid, i.e. submitting the event to the channel successfully might or might not affect the success / fail status of the pipeline.

On one side one might not want the health of the channel to impact critical pipelines; on the other side not reporting a failure in submitting to the channel could live the issue undetected.

[afrittoli]

Proposed design: https://docs.google.com/document/d/1Ry1B37QZlN6wwjgYj4TUdNwq_Wtex0A3nj11yizfqZI/edit?ts=5cd1879c

[afrittoli]

The CloudEventPipelineResource is implemented, so this can be closed now.