Capture rationale and impact for using the debug images

[codefromthecrypt]

In a recent pull request to zipkin, @devinsba noticed we are using the debug distroless image. We should be aware if this is adding heft to our based image which has grown substantially since moving off JDK 1.8

In other repos, we are starting to add RATIONALE.md, and this could be a good thing to start with.

[codefromthecrypt]

We are about to cut a new zipkin version, but for starters here's where we are > 250MiB. The last version prior to distroless (2.12.3) was < 150MiB.

openzipkin/zipkin                  2.12.3              69d91daa0839        6 months ago        147MB
openzipkin/zipkin                  2.16.2              b4a4e6e796ee        10 days ago         253MB

We definitely grew otherwise, some of it I aim to cleanup in the next release or so. Ex 2.16.2 is dramatically larger than 2.12.3: however it isn't a hundred megs larger.

zipkin-server-2.12.3-exec.jar                     2019-03-01 16:06  39726640
zipkin-server-2.16.2-exec.jar                     2019-08-24 15:08  56548977

I tried to build w/o the -debug and it seems busybox isn't there.. No idea the space saving if we were able to avoid it.

Step 11/17 : RUN ["/busybox/sh", "-c", "adduser -g '' -h /zipkin -D zipkin"]
 ---> Running in e50ea9179b1f
OCI runtime create failed: container_linux.go:345: starting container process caused "exec: \"/busybox/sh\": stat /busybox/sh: no such file or directory": unknown

[anuraaga]

Can get the numbers into a RATIONALE next week. The debug image just adds BusyBox which was just a couple megabytes or so and is definitely good to keep. Looks like the Java image is getting larger by adding dependencies for swing type libraries, which I remember our old image was explicitly deleting since we don't have a gui. We'll probably need to go back to manual trimming in a base image to get those bytes back.

https://github.com/GoogleContainerTools/distroless/commits/master/java