There is some basic work on a docker-compose setup for AQS + cassandra + test data in its current state from another task here https://gerrit.wikimedia.org/r/c/analytics/aqs/+/679295

In T288160#7263723, @hnowlan wrote:

There is some basic work on a docker-compose setup for AQS + cassandra + test data in its current state from another task here https://gerrit.wikimedia.org/r/c/analytics/aqs/+/679295

Thanks!

I took a look at this with a mind toward extending it for our use, but I think I concluded otherwise. What's in r679295 starts an instance of Cassandra, and a container to run the legacy AQS service. This old version of RESTBase initializes the schema, and I assume the test data you're referring to is what the tests create.

What we need here is Cassandra (+ schema + some test data) and Druid (+ some test data), and (TTBMK) we will not need the legacy service running for anything. That leaves an intersection of mostly "Cassandra".

Additionally, it probably does not make sense for AQS 2.0 to reuse the existing repo since it isn't likely they'll share any usable history (especially if we implement using Golang, as has been discussed). The legacy repository would benefit from having r679295 in the interim though.

I stubbed out https://gitlab.wikimedia.org/eevans/aqs as a jumping off point (based on @hnowlan's gerrit).

Thoughts?

As an experiment: I used a query (ala cqlsh) on the production Cassandra cluster, plus some shell hackery, to create a COPY-compatible csv file and some import test data (see d6399bd). This gives us real data -data we could use (for example) to compare API results with those from production- and it's fairly straightforward to obtain.