canonical-ci-engineering team mailing list archive

[Vincent Ladeuil <vila+ci@xxxxxxxxxxxxx>]

>>>>> Francis Ginther <francis.ginther@xxxxxxxxxxxxx> writes:

    > Once we have a handle on what we own and what it's used for, how can
    > we then report to our consumers the health of the system? Is this
    > something that belongs on the dashboard or in the wiki? It would be a
    > lot easier to field some support questions if developers knew to
    > automatically check a status and notices page first. Although I also
    > fear the headache that it would require to keep it up to date without
    > some automation behind it.

    > Thoughts?

>From a user point of view, it should always be green :)

The only thing that may matter to a user is: is there something
currently broken that is holding back *my* job.

So reporting the status should be: "it's green" or "here are the failing
parts". This is not really different from presenting test results for a
given job: "It's passing" or "Here are the failed tests."

I'm not sure we can define (yet) all the errors we will encounter along
the way, some of them won't probably make much sense to users.

So I'd rather start small with whatever tricks we are using ourselves
today: 

- are the private jenkins up,
- are there phones stucked after a reflash,
- are there nodes offline from jenkins pov,
- <your tricks here>.

For a first shot, I included the following:
- what are the jobs that are known failing,
- what are the jobs that are currently running.

But re-reading that, I think this is already different and may not be
strictly considered part of the 'status', we may want to present them
*even* if the ci engine is green.

So let's stay focused, in this thread, about the status and keep that
for another thread ;)

Starting small should also allow us to write automated tests for each of
the tricks, both in terms of how we find the info and how we present the
info to us and to our users.

        Vincent