Re: BOSH OpenStack CPI: planned end of maintenance by end of 2019
Hi everyone, hi Daniel,
Thanks for the interest in the BOSH OpenStack CPI, I'm happy to give you some perspective on what we had been doing in the past and what the current status is.
* We've automated pretty much everything in terms of regular maintenance. This includes updates of the dependencies like rubygems , and consumption of the ruby-release in bosh packages . We have automated the update of ruby-release itself, such that ruby, rubygems, etc are automatically updated.
* All PRs are tested against the supported set of OpenStack APIs by using the cf-openstack-validator to run a base set of API interactions with OpenStack . We're using OpenLab  to provide virtualized OpenStack environments for us. Adding a new one is only a few lines of yml 
* New versions of bosh, a stemcell, or the cpi are automatically tested in a pipeline using BATs and e2e tests. While this pipeline currently runs against an SAP provided OpenStack, it is easy to set this up against a different OpenStack
* There are very few issues open (currently 7 issues and 3 of those are feature requests)
* There isn't really left-over work for now: Our backlog contains 37 open stories tagged with 'cpi' . Quite a few are crossed out (we decided not to do them, but keep them for future reference to document decisions), or even about different CPIs like AWS and GCP.
* The biggest pain point in the past have been fog-openstack and excon. Updates of these two libraries are always a gamble, they don't seem to be well-tested and introduced quite some annoyances over the last few years. One example is how the individual OpenStack projects adopted microversions in their APIs and how fog-openstack selects which version to use. More recently, this didn't have much impact – it mostly matters when you're trying to use functionality that isn't present below a certain microversion or has been removed starting from a version.
* We have been trying to support the community with their OpenStack issues in #openstack and #bosh on CF Slack. Interrupts don't come on a daily basis, but there is a need for looking at community questions.
* We tried to improve the ecosystem with projects like cf-openstack-validator , better OpenStack support in bbl , and .tf templates for OpenStack . Also maintaining the OpenStack part of the docs on bosh.io
Overall, I guess the amount of work you would need to put into the OpenStack CPI depends on what you're trying to do with it. My guess is that a pair of engineers is sufficient to address some of the more recent feature requests and keep the system alive and healthy.
Let me know if you need more information to get a feeling for what it means to take over this work. I'm happy to talk about individual aspects some more, if it helps.
From: <cf-bosh@...> on behalf of Daniel Jones <daniel.jones@...>
I should imagine there must be enough very large consumers of the OpenStack CPI to make it commercially viable for a small team to be funded that could maintain it, were all said consumers to pull together and contribute what I should imagine would be a small sliver of their total CF expenditure.
If any OpenStack CPI users want to talk to EngineerBetter about funded development, we'd be happy to have that conversation as we're doing some similar EOL work for other CF projects.
Similarly, if our chums at Anynines, Altoros, S&W, GrapeUp, EVoila, GStack, Armakuni, Novatec or others want to work together on this, we'd be open to that. I can also imagine that some chunk of Pivotal's revenue depends on this CPI, so perhaps some of our teal friends might be interested too? It could be like one of those Marvel films (that I don't watch) where we all team up to vanquish the evil villain of abandoned software.
Marco - how many pairs were working on these streams of work, and do you have a feel for the balance of vital maintenance, tech debt, new features, and quality-of-life improvements in the backlog?
On Wed, 21 Aug 2019 at 09:38, Marco Voelz <marco.voelz@...> wrote: