Merge pull request #5300 from saschagrunert/platforms

Add guide for shipping alternative platforms

Author: k8s-ci-robot

contributors/devel/sig-architecture/platforms.md

@@ -0,0 +1,128 @@
+# Shipping alternate platforms in Kubernetes release artifacts
+
+The default Kubernetes platform is Linux/amd64. This platform is fully tested
+and build and release systems initially supported only that. A while ago we
+started an [effort to support multiple architectures][0]. As part of this
+effort, we added support in our build/release pipelines for the architectures
+arm, arm64, ppc64le and s390x on different operating systems like Linux, Windows
+and macOS.
+
+[0]: https://github.com/kubernetes/kubernetes/issues/38067
+
+The main focus was to have binaries and container images to be available for
+these architectures/operating systems and for contributors that are interested
+to be able to take these artifacts and set up CI jobs to adequately test these
+platforms. Specifically to call out the ability to run conformance tests on
+these platforms.
+
+Target of this document is to provide a starting point for adding new platforms
+to Kubernetes from a SIG Architecture perspective. This does not include release
+mechanics or supportability in terms of functionality.
+
+# Step 1: crawling (Build)
+
+- docker based build infrastructure should support this architecture
+
+The above 2 implicitly require the following:
+
+- golang should support the architecture out-of-the-box.
+- All our dependencies, whether vendored or run separately, should support this
+  platform out-of-the-box.
+
+In other words, anyone in the community should be able to use our build infra to
+generate all artifacts required to stand up Kubernetes.
+
+More information about how to build Kubernetes can be found in [the build
+documentation][1].
+
+[1]: https://github.com/kubernetes/kubernetes/tree/3f7c09e/build#building-kubernetes
+
+# Step 2: walking (Test)
+
+It is not enough for builds to work as it gets bit-rotted quickly when we vendor
+in new changes, update versions of things we use etc. So we need a good set of
+tests that exercise a wide battery of jobs in this new architecture.
+
+A good starting point from a testing perspective are:
+
+- unit tests
+- e2e tests
+- node e2e tests
+
+This will ensure that community members can rely on these architectures on a
+consistent basis. This will give folks who are making changes a signal when they
+break things in a specific architecture.
+
+This implies a set of folks who stand up and maintain both post-submit and
+periodic tests, watch them closely and raise the flag when things break. They
+will also have to help debug and fix any platform specific issues as well.
+
+Creating custom testgrid dashboards can help to monitor platform specific tests.
+
+# Step 3: running (Release)
+
+So with the first 2 steps we have a reasonable expectation that there is a bunch
+of people taking care of an architecture and it mostly works ok ("works on my
+machine!"), if things break it gets fixed quickly.
+
+Getting to the next level is a big jump from here. We are talking about real
+users who are betting their business literally on the work we are doing here. So
+we need guarantees around "can we really ship this!?" question.
+
+Specifically we are talking about a set of CI jobs in our release-informing and
+release-blocking tabs of our testgrid. Kubernetes release team has a "CI signal"
+team that relies on the status(es) of these jobs to either ship or hold a
+release. Essentially, if things are mostly red with occasional green, it would
+be prudent to not even bother making this architecture as part of the release.
+CI jobs get added to release-informing first and when these get to a point where
+they work really well, then they get promoted to release-blocking.
+
+The problem here is once we start shipping something, users will start to rely
+on it, whether we like it or not. So it becomes a trust issue on this team that
+is talking care of a platform/architecture. Do we really trust this team not
+just for this release but on an ongoing basis. Do they show up consistently when
+things break, do they proactively work with testing/release on ongoing efforts
+and try to apply them to their architectures. It's very easy to setup a CI job
+as a one time thing, tick a box and advocate to get something added. It's a
+totally different ball game to be there consistently over time and show that you
+mean it. There has to be a consistent body of people working on this over time
+(life happens!).
+
+What are we looking for here, a strong green CI signal for release managers
+to cut a release and for folks to be able to report problems and them getting
+addressed. This includes [conformance testing][2] as use of the Kubernetes
+trademark is controlled through a conformance ensurance process. So we are
+looking for folks here to work with [the conformance sub project][3] in addition
+to testing and release.
+
+[2]: https://github.com/cncf/k8s-conformance
+[3]: http://bit.ly/sig-architecture-conformance
+
+# Step 4: profit!
+
+If you got this far, you really have made it! You have a clear engagement with
+the community, you are working seamlessly with all the relevant SIGs, you have
+your stuff in Kubernetes release and get end users to adopt your architecture.
+And having achieved conformance, you gain conditional use of the Kubernetes
+trademark relative to your offerings.
+
+# Rules of the game (Notes?)
+
+- We should keep it easy for folks to get into Step 1.
+- Step 1, by default things should not build and should be switched off.
+- Step 1, should not place undue burden on review or infrastructure (case in
+  point - WINDOWS!).
+- Once Step 2 is done, we could consider switching things on by default (but
+  still not in release artifacts).
+- Once Step 3 is done, binaries / images in arch can ship with release.
+- Step 2 is at least the default e2e-gce equivalent, PLUS the node e2e tests.
+  More the better.
+- Step 2 will involve 3rd party reporting to test-grid at the least.
+- Step 2 may end up needing boskos etc to run against clouds (with these arches)
+  where we have credits:
+- Step 3 is at least the conformance test suite. More the better. Using
+  community tools like prow/kubeadm is encouraged but not mandated.
+- Step 4 is where we take this up to CNCF trademark program.
+  for at least a year in Step 3 before we go to Step 4.
+- If at any stage things bit rot, we go back to a previous step, giving an
+  opportunity for the community to step up.