Running validations without Mistral, and more
We’ve already talked about Validations, how to run them and how the new CLI will be awesome.
Now is time for some updates!
But first, some history :). The validations were mostly launched using Mistral, an OpenStack service. Since there are talks about removing this service, we had to think a bit about “how can we ensure we’ll still be able to run them”.
The current state of the Framework is: we’re almost freed of Mistral!
Indeed, there’s “only” one thing that needs Mistral right now: listing the validations. The quotes are due to the fact “listing” is also used when we want to run a validation group.
This means the following command will work, by default, without mistral:
source ~/stackrc
openstack tripleo validator run --validation-name undercloud-selinux-mode
And, as you can see in the asciicast, without any runing Mistral!
We’re actively working on a way to get rid of Mistral for the “listing” part, so stay tuned for more Mistral-less features :).
But. That’s not the only thing we can brag about!
Having validations is fine, having a nice way to run them is good. But how can you ensure the validations you’re running are, actually, checking the things?
We have to provide a level of trust, with proof that the validations are working as intended, so that Devs, Operators and Support can rely on them with confidence.
This is now possible, thanks to Molecule and the heavy work of my colleagues in order to integrate those tests, both in tripleo-validations repository, but also in the CI.
Running Molecule is easy, although it might be tricky to create the test suit.
The next steps are obvious: finish the Mistral-less changes, and work on unit-tests for the existing validations.
Another battle exists though: performances. We have to find a way to get a faster run. The current validations are “simple”, involving just a couple of tasks, but are, for some reasons, really slow, at least when launched via the CLI.
There are multiple reasons for that, one being the fact gathering. If we avoid gathering facts for nothing, we will gain time, especially when we have validations running against a 100+ compute infrastructure.
Another lead: the way we actually run ansible through the new “validator” CLI.
So the work is far from being over, but we’re seeing massive improvments.
Stay tuned! That’s it for today!