potiuk edited a comment on pull request #14911: URL: https://github.com/apache/airflow/pull/14911#issuecomment-803641595
Hey @ecerulm @grantav @mik-laj (also @kaxil and @ashb) - while improving the automation of image building, I also realized that after some of the recent changes we got very close to a "target" prod image setting I Imagined when I started it. Finally the original "development" image turned into a much closer "production" one - much more user friendly to build and use. The nice thing is that we still kept the "development" friendliness (and especially CI friendliness) of the image that I started it from and we have now a very nice (I think) way of describing it. Following @mik-laj separating out of the docker-stack in our documentation I restructured it internally and removed some of the old arguments/fixed examples, I also made a very clear progression of complexity of building the image - from very easy extending the image going through more complex cases, ending with rather complex "building image in restricted environment" (which was BTW a real case we had with two of our customers independently so it is not an isolated or rare case). I restructured and split the "build-args" reference following the same pattern - first you have the basic args that you will almost always use when customizing the image, and then you have other args that are grouped by "customization cases" with links to the examples section where you can see how the args can be used. I made it very clear - I think when extension is a good choice, and when you should go with the customization route. I prepared an example Customize vs Extend where I got 1.1 GB "extended" vs 874 MB "Customized" with the same customization, this is a nice example of showing how you can get 20% optimization if you are a "Power-user". Moreover - I extracted all the examples (using `exampleinclude`) and I run them in CI (!) so that we will be sure that those examples work (I have to fix failing test but this is the goal to keep them tested automatically). @ecerulm -> you were looking at the image recently as the user, I would really appreciate your feedback and potential improvements in this PR - but anyone else's comments are most welcome. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected]
