ray-project / ray_beam_runner

Ray-based Apache Beam runner
Apache License 2.0
42 stars 12 forks source link

Add support for State in ParDo / DoFn operations #12

Closed pabloem closed 2 years ago

pabloem commented 2 years ago

The Ray Runner does not support accessing state by ParDo / DoFns. This test verifies this behavior:

https://github.com/ray-project/ray_beam_runner/blob/master/ray_beam_runner/portability/ray_runner_test.py#L328-L352

We have an implementation of a state manager (https://github.com/ray-project/ray_beam_runner/blob/master/ray_beam_runner/portability/execution.py#L205-L243), as an Actor that implements the state servicer interface (https://github.com/apache/beam/blob/044313637c9eea2e3c2b0baa60bc853a948c12ee/sdks/python/apache_beam/runners/portability/fn_api_runner/worker_handlers.py#L944). We need to verify that the state is plumbed through and working.