pants ci: add integration tests jobs

[cognifloyd]

This adds 2 GHA jobs to the pants+pytest CI workflow:

• integration-tests
• integration-st2cluster-tests

Sadly, GHA is very repetitive, so a lot of the steps for these jobs are just copies of steps in other jobs. So, this should be fairly easy to review despite the line count. Go through each commit to see why I made a few changes to some steps.

Currently, we run Makefile-based integration tests in CI and Orquesta CI workflows. For the pants+pytest CI, I opted to do 2 jobs instead of separate workflows. Also, I found that the only integration tests that require a running st2cluster (started with tools/launchdev.sh) are the orquesta tests. Since st2cluster is not required by most integration tests, I was able to simplify the setup and avoid starting it. This also means there will be far fewer logs to go through when looking up failures about one of these integration tests. So, I used that requirement (needs a running st2cluster) as the dividing line between the 2 sets of integration tests. By using "st2cluster" instead of "orquesta" in the job name, I created an (hopefully) obvious home for future non-orquesta tests that need to use st2cluster.

These PRs were split out of this PR to keep each PR focused and easier to review:

• Pants: Add BUILD metadata for integration tests #6275
• Refactor some CI scripts #6276
• Add support for overriding st2.conf settings via env vars #6277
• pants-plugins/uses_services: add support for uses=st2cluster in BUILD metadata #6279
• pants-plugins/uses_services: respect ST2_DATABASE__* vars #6278
• Add new option [messaging].prefix to configure prefix of RabbitMQ exchange/queue names #6282
• Make tests safer to run in parallel by changing the Redis key namespace #6283

[cognifloyd]

Planned sequence of PRs:

flowchart LR
    A("✅ #6275: BUILD metadata") --> D("✅ #6279: pants-plugins/uses_services st2cluster")
    D ---> H("⏳ #6273: Add integration tests job")
    B("✅ #6276: Refactor CI scripts") ---> G("✅ #6283: Add redis namespace + ST2_* ST2TESTS_* vars in launchdev.sh")
    G --> H
    C("✅ #6277: env var support") --> E("✅ #6278: pants-plugins/uses_services refactor + mongo")
    E --> F("✅ #6282: Add messaging.prefix option") & G
    F --> H

Loading

• Pants: Add BUILD metadata for integration tests #6275
• Refactor some CI scripts #6276
• Add support for overriding st2.conf settings via env vars #6277
• pants-plugins/uses_services: add support for uses=st2cluster in BUILD metadata #6279
• pants-plugins/uses_services: respect ST2_DATABASE__* vars #6278
• Add new option [messaging].prefix to configure prefix of RabbitMQ exchange/queue names #6282
• Make tests safer to run in parallel by changing the Redis key namespace #6283
• pants ci: add integration tests jobs #6273

[winem]

Well done!

This is btw the first time that I used the Explain feature of the GItHub Copilot which was quite helpful to double-check my own interpretation of the code.

[cognifloyd]

These vars are used to skip some integration tests unless running in CI.

[cognifloyd]

With all of the recent changes to use redis and to make running tests in parallel safer, and with pants+pytest running the tests, I attempted to run tests that we've been skipping even in CI for a while. Happily, they seem much more reliable now, so this line re-enables the tests in pants+pytest CI.