[GAPRINDASHVILI] Prevent queueing things for a zone that doesn't exist in the region

[bdunne]

Prevent queueing things for a zone that doesn't exist in the region
https://bugzilla.redhat.com/show_bug.cgi?id=1635764

(cherry picked from commit 01c28d5)

Based on #17987 (comment)

[jrafanie]

@bdunne I'm assuming this one will cause test failures on the other repos gaprindashvili branch after merge, much the way the original one did on master, right? I'm good with this, but it might require a few PRs on other repos to fix.

[jrafanie]

LGTM, see my comment above though.

[simaishi]

Is this everything?

ManageIQ/manageiq-ui-classic#4669
ManageIQ/manageiq-providers-ovirt#289
ManageIQ/manageiq-automation_engine#230
#18009

(plus 2 PRs that go with this PR, ManageIQ/manageiq-api#472 and #18036)

[bdunne]

@simaishi Yes, I think that's all of them

[cben]

This backport also broke providers-kubernetes & providers-openshift.
(https://travis-ci.org/ManageIQ/manageiq-providers-openshift/builds/442300654
https://travis-ci.org/ManageIQ/manageiq-providers-kubernetes/builds/442299612)
TLDR: all this is already fixed on master, so just need to backport some test fixes.

Took me a long time to diagnose, because chains of rails hooks lead to really disconnected symptoms (like what does null value in column "created_at" violates not-null constraint have to do with this validation !?)
The chain definitely involved here is EMS -> Authentication -> after_save hook -> raise_evm_event;
in the container providers we also have the scary chain EMS -> Endpoint -> after_create/after_destroy hooks -> create/destroy a "child" EMS, suspect but not sure that's involved too...

Took me even longer to debug/fix. The root problem was the following stubbing all over kubernetes & openshift tests:

allow(MiqServer).to receive(:my_zone).and_return("default")

this was affecting the zone with which events were added to MiqQueue to be "default" but no such zone...

So all these stubs can be replaced with the more robust setup (which we already use in some tests):

EvmSpecHelper.local_miq_server(:zone => Zone.seed)

(done in ManageIQ/manageiq-providers-kubernetes@db9bdff but that's not the fix we'll use)

For some reason only then, knowing my_zone is the problem, I found the master PRs that fixed this:
ManageIQ/manageiq-providers-kubernetes#289
ManageIQ/manageiq-providers-kubernetes#297
ManageIQ/manageiq-providers-openshift#114
sending backport PRs now...