Hi folks,
as you might now, we run CouchDB builds on builds.apache.org for quite a
while now.
Looking at https://builds.apache.org/job/CouchDB/ I see an ocean of red
builds with a very occasional blue (successful) build in between. This
makes me sad.
Actually, each build bubble represents six sub builds (3 operating
systems, 2 Erlang versions). Anyway, obviously these builds are very
unreliable and provide zero benefit to the project.
A breakdown of *why* the builds fail is here:
https://github.com/basti1302/couchdb-ci/blob/master/utils/analyze-jenkins-logs/ci-errors.markdown
(Caveat: This report is generated by a little script that makes some
wild guesses and simplifications, so its results might be wrong)
My question: What can we do to make this better?
* Do we need to fix something in the various CouchDB source repos to
make the build more reliable?
* Is the build itself the culprit? (e. g. Docker issues etc.)
* Will everything resolve itself if we switch back to one big repository?
The discussion "On dependency management and CI issues associated with
it" brought up the point of dependency problems due to the multi-repo
approach, but I really don't think that this is even the main issue on
builds.a.o.
And, most importantly: Who is willing to help with this or better yet,
take the lead here?
Best regards
Bastian
