Use only fake domain in our tests

[michalvavrik]

Summary

• closes Don't connect to external domains in tests #2111

The ReactiveRestClientProxyIT test really talked to http://example.com, now no external communication to example.com is run, however we assert that correct host was proxied by passing $host via Nginx back to the test.

Please select the relevant options.

• Bug fix (non-breaking change which fixes an issue)
• Dependency update
• Refactoring
• Backport
• New scenario (non-breaking change which adds functionality)
• This change requires a documentation update
• This change requires execution against OCP (use run tests phrase in comment)

Checklist:

• Methods and classes used in PR scenarios are meaningful
• Commits are well encapsulated and follow the best practices

[michalvavrik]

run tests

[michalvavrik]

run tests

[michalvavrik]

There doesn't seem to be a sense to rerun OCP now, I'll try it tomorrow as the cluster is in a bad shape.

[michalvavrik]

OCP test failures are not related, but I need to re-think that proxy stuff I guess. Not sure.

[michalvavrik]

I thought about it and what we need to test is that proxy options are set correctly, not the redirect to external domain.

[michalvavrik]

run tests

[michalvavrik]

@mocenas OCP native run didn't fail, I canceled it manually because it was started when OCP cluster was down and now it is waiting in a queue. I won't merge this till it is green, so please review when you get a time. Thanks

[mocenas]

I thought about it and what we need to test is that proxy options are set correctly, not the redirect to external domain.

I thought about it too. It "would be nice" if we could do an E2E test to verify that communication though proxy works. But as you states, we want to test connection and options-passing to the proxy, not proxy's ability to forward it outside.

[mocenas]

LGTM

[michalvavrik]

I have re-triggered OCP native manually in Jenkins several times, it seems like it fails randomly on OOM and failing tests are changing. I can see periodic OCP native CI is also failing over OOM (among other) so I dare say it is not related (also not sure how could changes here cause OOM). For that reason I think it should be safe to merge.