Re: Cloud Foundry and Kubernetes Integration Efforts
That said, there are a few critiques I have that I hope can offer some constructive criticism along with a proposal for a smaller initiative in this same direction.
First off, the proposals feel focused on the use of a particular solution (Kubernetes primarily) rather than specific Operator or App Developer problems. Understanding and gathering evidence of problems Operators are currently facing with bosh, and that app developers are experiencing with Diego would be valuable on its own even if the research is built on the assumption that k8s can solve these problems. Ideally one could point to a particular metric (like deploy times) as a success metric so that there is a clear advantage that justifies the switch.
A second piece of feedback is that the proposals envision an attractive end state but do not provide small iterative milestones that could be used to ship incremental software experiments along the way. Taking on such ambitious goals is especially risky without problem statements and metrics.
Finally, these are great ideas! One potentially smaller step would be to fully support a containerized Loggregator that could be run as a unified egress model both on Cloud Foundry Container Runtime (CFCR) as well as the Cloud Foundry Application Runtime (CFAR). The Loggregator team has been thinking about modularity since the very first rumblings of a bosh deployed Kubernetes release and our former anchor and now dedicated committer to CFCR has put together an incubation release of this.