JudgeServer

Document

API示例

• Request POST

{
    "language": "py",
    "src": "import re\nprint(\"666\")",
    "max_cpu_time": 3000,
    "max_memory": 33554432,
    "test_case_id": "test",
    "output": false
}

• Response

result为Judger的判题结果。

{
    "err": null,
    "data": [
        {
            "cpu_time": 30,
            "real_time": 30,
            "memory": 9691136,
            "signal": 0,
            "exit_code": 0,
            "error": 0,
            "result": -1,
            "test_case": "1",
            "output_md5": "fae0b27c451c728867a567e8c1bb4e53",
            "output": null
        }
    ]
}

判题机参数

MAX_READ_BYTES = 64 * 1024 * 1024  # 最大读取输出大小 64M
MAX_RESP_BYTES = 16 * 1024  # 单个测试点最大返回输出大小 16K

并发配置

JUDGE_SERVER_WORKERS      Gunicorn worker 数，默认等于 MAX_WORKER_NUM（通常为 CPU 核数）
JUDGE_TEST_CASE_WORKERS   单次提交内部并行测试点进程数，默认 1
JUDGE_SERVER_BACKLOG      请求等待队列长度，默认 2048
JUDGE_SERVER_MAX_REQUESTS worker 处理指定请求数后回收，默认 2000

高并发场景建议保留 JUDGE_TEST_CASE_WORKERS=1，通过增加 JUDGE_SERVER_WORKERS 并行处理不同用户的提交；低负载、强调单次提交延迟时，可显式设置为 2 或 4。