For some reason, the MultiStoresImportImages test was causing the test runner to balloon to ~1.5G, which generates spurious OOMs in the gate, resulting in worker death and failure. This test was generating 10M of data with data_utils.random_bytes(), instead of the default of 1K, for no apparent reason. We are still not sure why, but this results in the memory ballooning, potentially because of BytesIO, or request.put() mishandling. Regardless, this is unnecessary, and causing failures in the gate, so this patch switches it back to the default, which works fine. Change-Id: I9f1ea0114531a735bd38ad54da7ce15fb2bf7f7c
|2 years ago|
|compute||2 years ago|
|identity||2 years ago|
|image||2 years ago|
|network||2 years ago|
|object_storage||2 years ago|
|volume||3 years ago|
|README.rst||6 years ago|
|__init__.py||9 years ago|
Tempest Field Guide to API tests
What are these tests?
One of Tempest's prime function is to ensure that your OpenStack cloud works with the OpenStack API as documented. The current largest portion of Tempest code is devoted to test cases that do exactly this.
It's also important to test not only the expected positive path on APIs, but also to provide them with invalid data to ensure they fail in expected and documented ways. The latter type of tests is called
negative tests in Tempest source code. Over the course of the OpenStack project Tempest has discovered many fundamental bugs by doing just this.
In order for some APIs to return meaningful results, there must be enough data in the system. This means these tests might start by spinning up a server, image, etc, then operating on it.
Why are these tests in Tempest?
This is one of the core missions for the Tempest project, and where it started. Many people use this bit of function in Tempest to ensure their clouds haven't broken the OpenStack API.
It could be argued that some of the negative testing could be done back in the projects themselves, and we might evolve there over time, but currently in the OpenStack gate this is a fundamentally important place to keep things.
Scope of these tests
API tests should always use the Tempest implementation of the OpenStack API, as we want to ensure that bugs aren't hidden by the official clients.
They should test specific API calls, and can build up complex state if it's needed for the API call to be meaningful.
They should send not only good data, but bad data at the API and look for error codes.
They should all be able to be run on their own, not depending on the state created by a previous test.