jason810496 commented on code in PR #65958: URL: https://github.com/apache/airflow/pull/65958#discussion_r3294140153
########## task-sdk/src/airflow/sdk/coordinators/java/coordinator.py: ########## @@ -0,0 +1,261 @@ +# +# 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. +"""Java runtime coordinator that launches a JVM subprocess for Dag file processing and task execution.""" + +from __future__ import annotations + +import email +import os +import pathlib +import selectors +import socket +import subprocess +import time +import zipfile +from typing import TYPE_CHECKING, cast + +import attrs +import psutil +import structlog + +from airflow.sdk.execution_time.coordinator import BaseCoordinator +from airflow.sdk.execution_time.supervisor import ActivitySubprocess + +if TYPE_CHECKING: + from collections.abc import Sequence + + from structlog.typing import FilteringBoundLogger + from typing_extensions import Self + + from airflow.sdk.api.client import Client + from airflow.sdk.api.datamodels._generated import BundleInfo + from airflow.sdk.execution_time.workloads.task import TaskInstanceDTO + +log: FilteringBoundLogger = structlog.get_logger(logger_name="coordinators.java") + + +def _start_server() -> socket.socket: + server = socket.socket() + server.bind(("127.0.0.1", 0)) + server.setblocking(True) + server.listen(1) # Just need to listen to the child process. + return server + + +def _calculate_classpath(jars_root: Sequence[pathlib.Path]) -> str: + jars = (p.as_posix() for root in jars_root for p in root.iterdir() if p.suffix == ".jar") + return os.pathsep.join(jars) + + [email protected] +class _MainJar: + path: pathlib.Path + main_class: str + schema_version: str | None + + @classmethod + def find(cls, jars_root: Sequence[pathlib.Path]) -> Self: + for root in jars_root: + for p in root.iterdir(): + if p.suffix != ".jar": + continue + with zipfile.ZipFile(p) as zf: + with zf.open("META-INF/MANIFEST.MF") as f: + manifest = email.message_from_binary_file(f) + if main_class := manifest["Main-Class"]: + return cls(p, main_class, manifest.get("Airflow-SDK-Supervisor-Schema-Version")) + resolved_paths = os.pathsep.join(str(p.resolve()) for p in jars_root) + raise FileNotFoundError(f"cannot fine main class in {resolved_paths}") Review Comment: > Curious, if each runtime is expected to invent their own mechanism or we can plan something common like airflow-runtime-version.json which can provide the version for each bundle? It will be categorized by the programming language. e.g. compiled language (Go, Rust, C++, etc), interpreted language (TS in this case). I lean toward "self-contain" bundle for discovery. Take the https://github.com/apache/airflow/pull/67153 as example, the binary itself contains `airflow-metadata` (dagId, taskId, etc. I will include supervisor schema version as well) and the raw source code for UI visualization. We could introduce similar concept for interpreted language by having special comment header or footer that include the `airflow-metadata` spec. So that we could distinguish which bundle belongs to this task in task execution runtime without starting each foreign lang subprocess, just by file I/O scaning. -- 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]
