Docker-in-Docker vs Kaniko as recommended method for building Docker images with CI?
In determining the best way to handle building Docker images using GitLab CI with a customer, we went over the pros and cons of using Docker-in-Docker (DinD) vs. using Kaniko and it seemed in our review that Kaniko addresses most of the concerns of DinD with few, if any, real drawbacks. My customer asked why Kaniko was not the recommended method based on how our documentation described each approach, so I'm bringing the question here for clarification.
- Why do we still recommended Docker-in-Docker over Kaniko in our docs?
- Should we revise our documentation to direct people to Kaniko as our preferred method?
Edited by Christiaan Conover