#171 updating docker templates to fabric8

[Cho-William]

• updated archetype and foundation-mda templates to leverage fabric8 docker-maven-plugin in lieu of orphedomos
• sagemaker antora documentation was updated to reflect updated configuration of ECR registry URL
• machine learning pipeline and notifications to set the ECR registry URL has been updated
• fixed null pointer exception bug with check for onnx conversion generation (MachineLearningPipelineGenerator)
• release notes draft updated noting potential breaking changes regarding loss of extra orphedomos configs
• update the orphedomos to fabric8 migration to include the default skip configuration

IMPORTANT NOTE

Please note that this PR disables fabric8's docker builds as part of a default build for a downstream project:

• This configuration is consistent with the previous orphedomos configuration, although not leveraging this skip behavior would have been preferred
• Unfortunately, if a user were to generate a project and exclusively use mvn clean install rather than mvn clean generate-sources (which is how the archetype test script also does it) to create their project structure, the user may eventually encounter a broken build. Consider the following example:
  • The user begins generating a downstream project with only mvn clean install calls
  • The user's machine learning inference docker module is generated
  • The user builds again, and if the docker fabric8 build is not skipped, then that docker module will begin building
    • However, the generation for the machine learning inference pipeline step has not yet been generated at this point
  • Finally, the inference docker image expects a copy of ie the sdist requirements.txt from the inference pipeline step to be available in the docker build context/directory, but that ml pipeline inference step module has not yet been built yet
    • The reason why the inference pipeline step module has not yet been generated is because it is a sub-sub directory of the -pipelines module, and requires another generate-sources execution. Unfortunately, this source generation occurs after the aforementioned docker build.
    • Thus, previous archetype tests and downstream projects did not encounter this issue, as orphedomos docker builds were also disabled by default

Therefore, there are two options to remedy this issue for the archetype test script and downstream projects:

1. Simply keep the skip configuration as part of the default fabric8 docker build configuration
2. Update foundation-mda's generation logic to automatically include sub-sub step modules as declared child modules in their respective pipeline parent poms. Namely this applies to machine learning pipelines, and any other pipelines that house child modules.

Unfortunately, option 2 was not able to be incorporated in the scope of this ticket, however it is in the interest of the aiSSEMBLE team to eventually implement option 2 and revert the default skip behavior

[chang-annie]

looks like these bullets should be 1., 2., 3., instead of all 1's

[Cho-William]

That is intentional. That way if we insert another point, we don't have to manually renumber. Markdown is smart enough to increment the numbers when the README is rendered

[chang-annie]

oo good to know!

[d-ryan-ashcraft]

S: are we sure we need foundation-mda? It's not really a maven plugin, but rather a dependency that is fed to fermenter. I would have though that the normal reactor would pick it up for the cache without explicitly being set.

If it is that the build cache isn't picking up dependencies listed within plugins, I might just note that so it's easier to know what it is here in the future.

[Cho-William]

I believe it is indeed the latter case, because after adding these two dependencies, it resolved the inconsistent behavior with the archetype tests. I will add a comment to contextualize

[Cho-William]

Will also make sure the additions are not part of the plugins comment block

[ewilkins-csi]

A: Maybe "Upgrade Steps for Projects with Customized Orphedomos Configurations"? Usually we try to convey what conditions would cause the instructions to be relevant to a project in these titles.