Add github action for testing notebooks

[hevansDev]

No description provided.

[petermarshallio]

Reckon @techdocsmith @hevansDev we should specific about what tests we are, and are not, attempting to catch with this test suite. Are we trying to be sure, for example, that the notebooks run "as code"? Or are we just looking for quality things?

For example...

• Should this test do anything that a user wouldn't do if they picked up the notebooks; i.e. I noticed that there's lots of http://localhost:8082/status calls when the tests start the containers, waiting for Druid to actually be live. Therefore, the notebooks themselves should have something in them that will check that all the services are up, and then throw an error after a timeout. Ie replace status_client = druid.status and status_client.version. (Add to this that one of the tests I did was to the wrong port number being put into the notebook, and I'm wondering if this should actually be a notebook based thing.)
• Some notebooks state that they only require druid-jupyter - if the test is run with all-services and it works, but it doesn't with what we say it does, is that a fail? This might also speed up the actual time it takes to run the tests.

Are we OK with such a long timeout for the tests? When I ran this with the wrong port for Druid, for example, it took ages to fail the test. I was wondering if, realistically, any notebook should take like 5 minutes - as a quality test more than a functional one.

[petermarshallio]

I'm wondering @hevansDev @techdocsmith if we only really know if this change becomes an issue when we start to commit notebooks. I do think we should find the other angle, though, like what common issues is this change looking to catch?

[petermarshallio]

@hevansDev what errors does this test throw across all notebooks rn? I wonder cos I found something in a notebook earlier that didn't seem to work any more, and wondered if it was me (!!!) or if it really was an issue.

[petermarshallio]

Great catches!! I suggest pulling out the amendments to the notebooks into a PR that gets merged before this one as these fix a separate issue.

[petermarshallio]

I don't think this change is necessary : when I tested this notebook, the table was correctly dropped.
If still required, note that the variable needs to be used rather than the explicit table name. The variable table_name is set in a previous cell.

[petermarshallio]

To conform with the template, all imports should be in a section at the top ("import additional modules").
100 seconds also seems a long time to pause.

[petermarshallio]

As above, can it be lower than 100 seconds?

[petermarshallio]

I suspect the authors intention here was to use latest_session_length throughout, rather than session_length, given that the ingestion is a GROUP BY, and session_length is high cardinality.

[petermarshallio]

Argh I merged the wrong PR and thus deleted the branch. I have reverted the changes and restored the branch.