Product discovery for switching serverless builds from knative to GitLab CI
Background
Our current serverless implementation uses Knative build to create build template using the Knative builder
. This presents a couple of issues:
-
We cannot use our security features outside of GitLab CI
-
We cannot surface build errors without additional work which is subject to change based on knative's alpha status
What questions are we trying to answer
- What is the level of effort to surface errors and other valuable Knative troubleshooting data in the GitLab GUI?
- What is the level of effort to extend GitLab CI to handle serverless builds? (assuming we provide a limited set of runtimes and maintain them)
What is the backstory of the project? How does it impact the approach
We want to abstract as much as possible from the user. that is, the user should ideally only worry about the function code they need to write instead of standing up templates, infra, etc. Bringing builds to CI should make it easier for users to build and troubleshoot.
What do we already know about the areas we're exploring?
Knative build works well but it requires lots of extra steps to do simple things like surface errors.
What does success look like at the end of discovery?
Edited by Daniel Gruesso