uranusjr commented on code in PR #65956: URL: https://github.com/apache/airflow/pull/65956#discussion_r3159365035
########## java-sdk/adr/0003-workload-execution.md: ########## @@ -0,0 +1,473 @@ +<!-- + Licensed to the Apache Software Foundation (ASF) under one + or more contributor license agreements. See the NOTICE file + distributed with this work for additional information + regarding copyright ownership. The ASF licenses this file + to you under the Apache License, Version 2.0 (the + "License"); you may not use this file except in compliance + with the License. You may obtain a copy of the License at + + http://www.apache.org/licenses/LICENSE-2.0 + + Unless required by applicable law or agreed to in writing, + software distributed under the License is distributed on an + "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + KIND, either express or implied. See the License for the + specific language governing permissions and limitations + under the License. + --> + +# ADR-0003: Workload Execution — Language-Specific Task Execution + +## Status + +Accepted + +## Context + +Airflow's standard task runner executes Python callables. To support tasks written in other languages, the pipeline needs an extension point where a language-specific coordinator can intercept the execution, delegate to an external runtime process, and bridge the Task SDK protocol so the external process can access Airflow services (connections, variables, XCom) during execution. + +This ADR details the task execution side of the coordinator architecture described in [ADR-0001](0001-java-sdk-airflow-integration.md). It starts with the generic model — the abstract contracts and expected behavior that any language must implement — then walks through Java as a concrete example. + +## Decision + +### Extension Point: `BaseCoordinator` + +The same `BaseCoordinator` base class that handles DAG parsing also handles task execution. It is registered in `provider.yaml` under `coordinators`. For task execution, a subclass must implement: + +| Method | Signature | Responsibility | +|---|---|---| +| `task_execution_cmd` | `(what, dag_rel_path, bundle_info, comm_addr, logs_addr) -> list[str]` | Return the full command to launch the language runtime for task execution. `comm_addr` and `logs_addr` are `host:port` strings the process must connect to. | + +The base class provides `run_task_execution()` as a concrete method that handles all TCP/process plumbing automatically (same pattern as `run_dag_parsing()` for the DAG parsing side). + +**Parameters passed to `run_task_execution()`:** + +| Parameter | Type | Description | +|---|---|---| +| `what` | `TaskInstance` | The task instance to execute (id, dag_id, task_id, run_id, try_number, etc.) | +| `dag_rel_path` | `str \| PathLike` | Relative path to the DAG file / bundle within the bundle root | +| `bundle_info` | `BundleInfo` | Bundle name and version | +| `startup_details` | `StartupDetails` | Full startup context (task instance, DAG rel path, bundle info, run context, start date) — already consumed from fd 0 | + +### Registration + +The same `coordinators` entry in `provider.yaml` covers both DAG parsing and task execution — no separate registration needed: + +```yaml +coordinators: + - airflow.providers.sdk.<lang>.coordinator.<LangCoordinator> +``` + +### Discovery: `_resolve_runtime_entrypoint()` + +When `task_runner.main()` starts, before any Python task execution: + +``` +task_runner.main() Review Comment: Since we use the existing executor mechanism in Airflow, the idea of a worker varies depending on the executor in use. LocalExecutor for example does not really have a long-running worker concept; CeleryExecutor would require you to have at least the same number of worker processes as you want parallelism. I think another way to put it is this design does not really go into how workers are managed. It simply assumes you have JRE set up to run Java, the same way as Python and needed provider packages are assumed when you run Python tasks. (This brings up the interesting case of Kubernetes, where we can have different images for different languages, but currently the design requires you to have an image that can handle both. This is a possible future improvement work.) -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
