[Task]: Use turmoil to simulate client and server environments in tests

[laysakura]

What needs to happen?

See the description in #43

Issue Priority

Priority: 3 (nice-to-have improvement)

Issue Components

• Component: Python SDK
• Component: Java SDK
• Component: Go SDK
• Component: Typescript SDK
• Component: IO connector
• Component: Beam examples
• Component: Beam playground
• Component: Beam katas
• Component: Website
• Component: Spark Runner
• Component: Flink Runner
• Component: Samza Runner
• Component: Twister2 Runner
• Component: Hazelcast Jet Runner
• Component: Google Cloud Dataflow Runner

[rahulssheth]

@laysakura any chance I could help out here?

[laysakura]

@rahulssheth Sure! If you find anything to discuss with other contributors, feel free to start chatting in the discord (I found you in discord).

[rahulssheth]

Thanks @laysakura! Do we have any analogous frameworks used in tests of other SDKs? I have an idea of a few cases that could benefit from this but a reference would be helpful.

[laysakura]

@sjvanrossum Any idea?

[sjvanrossum]

Nothing truly analogous, but the Java, Go, and Python SDKs all ship with modules to scaffold test pipelines (e.g. https://github.com/apache/beam/tree/master/sdks/java/core/src/main/java/org/apache/beam/sdk/testing) and the direct runner for each of those SDKs is used for testing purposes. Integration tests are largely expressed through those constructs.

Turmoil sits a level below that and enables us to define deterministic integration tests for distributed components. The level of granularity there is mostly left up to us, so this can be used both to test individual services (e.g. BeamFnControl client, BeamFnData client) and to build the equivalent of TestPipeline and similar higher level APIs like a TestRunner.