Add an integration test that verifies unchanged data on re-indexing

[aaron-collier]

@andrewjbtw has a test collection in stage that should be used for an integration test that captures all metadata and the public xml, then performs a reindex and verifies that none of the data changed.

This will help verify that changes to dor_indexing_app will not introduce unexpected indexing bugs into production.

[jcoyne]

How would we run that integration test on qa?

[aaron-collier]

Mirror the collection on QA? I believe it was manually setup in the first place, that could be done again.

[jcoyne]

What if the data gets screwed up or removed from those environments? If we just check for "it doesn't change", then it might not be as good as ensuring we have a test in DSA for each of the scenarios he's identified.

[aaron-collier]

I think those are questions/details that can be discussed when we can work on this. I'm just capturing the need for a test that we've seen changes not get caught in our existing testing. This may not be doable, but I'd like to at least capture the need for now.

[jcoyne]

I guess I'm wondering if the goal is for dor_indexing_app to behave a certain way, why not write the tests in dor_indexing_app? We should avoid adding new tests to infrastructure-integration-test unless there's a multi-service integration that can't be done in an test suite that we run on every commit.

[andrewjbtw]

To clarify: this collection has all the rights metadata but not all the options for other types of metadata.

[ndushay]

Note that there is access_indexing_spec already in integration tests that creates a single new item, and loops through various access settings and checks the indexed value for each access setting. This may be sufficient. Note that this test is off by default.

[ndushay]

It is useful to have an integration test that assures that changes in, say Argo, not only are stored correctly in SDR, but are also indexed correctly. The test I wrote changes the object and also looks for index changes.

[jcoyne]

If you want that, then the Argo test suite might be the way to go. It has all of those components running and it will be run automatically on every change, unlike this test suite which must be run manually.

[justinlittman]

But an Argo test won't test changes in dor-services-app or dor-indexing-app, right? (At least until the changes are committed to main and a new container is built and a change is made to Argo that causes the Argo test suite to be run.)