{"id":106721,"date":"2026-09-18T07:32:33","date_gmt":"2026-09-17T23:32:33","guid":{"rendered":"https:\/\/www.wsisp.com\/helps\/106721.html"},"modified":"2026-09-18T07:32:33","modified_gmt":"2026-09-17T23:32:33","slug":"%e5%bd%93-gil-%e4%b8%8d%e5%86%8d%e6%98%af%e5%a4%a9%e8%8a%b1%e6%9d%bf%ef%bc%9afree-threaded-python-x-asyncio-%e7%9a%84%e5%a4%9a%e4%ba%8b%e4%bb%b6%e5%be%aa%e7%8e%af%e6%9e%b6%e6%9e%84%e5%ae%9e","status":"publish","type":"post","link":"https:\/\/www.wsisp.com\/helps\/106721.html","title":{"rendered":"\u5f53 GIL \u4e0d\u518d\u662f\u5929\u82b1\u677f\uff1aFree-Threaded Python \u00d7 asyncio \u7684\u591a\u4e8b\u4ef6\u5faa\u73af\u67b6\u6784\u5b9e\u8df5"},"content":{"rendered":"

\u5f53 GIL \u4e0d\u518d\u662f\u5929\u82b1\u677f：Free-Threaded Python \u00d7 asyncio \u7684\u591a\u4e8b\u4ef6\u5faa\u73af\u67b6\u6784\u5b9e\u8df5<\/h2>\n

\u8fc7\u53bb\u8c08 Python \u9ad8\u5e76\u53d1，\u6211\u4eec\u5f88\u5bb9\u6613\u5f62\u6210\u4e00\u79cd\u56fa\u5b9a\u601d\u7ef4：<\/p>\n

I\/O \u5bc6\u96c6\u578b\u7528 asyncio，CPU \u5bc6\u96c6\u578b\u7528\u591a\u8fdb\u7a0b\u3002<\/p>\n

\u8fd9\u5957\u7ecf\u9a8c\u957f\u671f\u4ee5\u6765\u6ca1\u6709\u9519，\u4f46 Free-Threaded CPython \u6b63\u5728\u6539\u53d8\u5b83\u80cc\u540e\u7684\u524d\u63d0\u3002<\/p>\n

PEP 703 \u8ba9 CPython \u53ef\u4ee5\u5728\u5173\u95ed GIL \u7684\u6a21\u5f0f\u4e0b\u8fd0\u884c\u3002Python 3.13 \u9996\u6b21\u63d0\u4f9b\u5b9e\u9a8c\u6027\u7684 free-threaded build；\u5230\u4e86 Python 3.14，PEP 779 \u5c06\u5176\u63a8\u8fdb\u5230\u5b98\u65b9\u652f\u6301\u7684 Phase II：\u5b83\u4ecd\u7136\u662f\u53ef\u9009\u6784\u5efa，\u800c\u4e0d\u662f\u9ed8\u8ba4\u6a21\u5f0f，\u4f46\u5df2\u7ecf\u4e0d\u518d\u53ea\u662f\u4e00\u4e2a\u5b9e\u9a8c\u9879\u76ee\u3002<\/p>\n

\u8fd9\u610f\u5473\u7740\u4e00\u4e2a\u8fc7\u53bb\u770b\u8d77\u6765\u6709\u4e9b\u201c\u5947\u602a\u201d\u7684\u67b6\u6784\u5f00\u59cb\u503c\u5f97\u8ba4\u771f\u8ba8\u8bba：<\/p>\n

\u4e00\u4e2a Python \u8fdb\u7a0b
\n\u2502
\n\u251c\u2500\u2500 Thread 1 \u2500\u2500 Event Loop 1
\n\u251c\u2500\u2500 Thread 2 \u2500\u2500 Event Loop 2
\n\u2514\u2500\u2500 Thread 3 \u2500\u2500 Event Loop 3<\/p>\n

\u95ee\u9898\u6765\u4e86：<\/p>\n

\u4e3a\u4ec0\u4e48\u4e0d\u7ee7\u7eed\u4f7f\u7528\u4e00\u4e2a event loop？\u4ec0\u4e48\u65f6\u5019\u591a\u4e2a\u7ebf\u7a0b\u5404\u8dd1\u4e00\u4e2a asyncio event loop \u53cd\u800c\u66f4\u5408\u7406？<\/p>\n

\u7b54\u6848\u7684\u6838\u5fc3\u5e76\u4e0d\u662f\u201casyncio \u53d8\u5feb\u4e86\u201d，\u800c\u662f：<\/p>\n

Free-threaded Python \u8ba9\u591a\u4e2a event loop \u6240\u5728\u7ebf\u7a0b\u6709\u673a\u4f1a\u771f\u6b63\u540c\u65f6\u6267\u884c Python \u4ee3\u7801\u3002<\/p>\n

\u8fd9\u6253\u5f00\u4e86\u4e00\u7247\u65b0\u7684\u8bbe\u8ba1\u7a7a\u95f4\u3002<\/p>\n

\n

\u4e00\u3001\u5148\u7406\u89e3\u672c\u8d28：Free-Threaded \u5e76\u4e0d\u4f1a\u8ba9\u4e00\u4e2a Event Loop \u5e76\u884c<\/h3>\n

\u8fd9\u662f\u6574\u4e2a\u95ee\u9898\u6700\u5bb9\u6613\u4ea7\u751f\u8bef\u89e3\u7684\u5730\u65b9\u3002<\/p>\n

\u5373\u4f7f\u5173\u95ed\u4e86 GIL，\u4e00\u4e2a asyncio event loop \u7684\u57fa\u672c\u8c03\u5ea6\u6a21\u578b\u4ecd\u7136\u6ca1\u6709\u6539\u53d8：<\/p>\n

Task A \u2500\u2500\u6267\u884c\u2500\u2500 await
\n \u2193
\nTask B \u2500\u2500\u6267\u884c\u2500\u2500 await
\n \u2193
\nTask C \u2500\u2500\u6267\u884c\u2500\u2500 await<\/p>\n

\u4e00\u4e2a event loop \u5728\u4e00\u4e2a\u7ebf\u7a0b\u4e2d\u6267\u884c callback \u548c Task\u3002\u67d0\u4e2a Task \u6b63\u5728\u6267\u884c Python \u4ee3\u7801\u65f6，\u540c\u4e00\u4e2a loop \u4e2d\u5176\u4ed6 Task \u4e0d\u4f1a\u540c\u65f6\u6267\u884c；\u53ea\u6709\u5f53\u524d\u4efb\u52a1\u6267\u884c\u5230 await \u5e76\u8ba9\u51fa\u63a7\u5236\u6743\u540e，loop \u624d\u80fd\u5207\u6362\u5230\u5176\u4ed6\u4efb\u52a1\u3002<\/p>\n

\u56e0\u6b64：<\/p>\n

1 Thread
\n \u2193
\n1 Event Loop
\n \u2193
\n10000 Coroutines<\/p>\n

\u4f9d\u7136\u4e3b\u8981\u5c5e\u4e8e concurrency（\u5e76\u53d1），\u800c\u4e0d\u662f Python \u5b57\u8282\u7801\u5c42\u9762\u7684\u771f\u6b63\u591a\u6838 parallelism\u3002<\/p>\n

Free-threaded \u771f\u6b63\u6709\u610f\u601d\u7684\u5730\u65b9\u5728\u8fd9\u91cc：<\/p>\n

CPU Core 0 CPU Core 1 CPU Core 2
\n \u2502 \u2502 \u2502
\nThread 1 Thread 2 Thread 3
\n \u2502 \u2502 \u2502
\nLoop 1 Loop 2 Loop 3
\n \u2502 \u2502 \u2502
\nTasks Tasks Tasks<\/p>\n

\u5f53 GIL \u88ab\u5173\u95ed\u540e，\u8fd9\u4e09\u4e2a OS \u7ebf\u7a0b\u53ef\u4ee5\u540c\u65f6\u6267\u884c Python \u4ee3\u7801\u3002<\/p>\n

\u4e8e\u662f\u51fa\u73b0\u4e86\u4e00\u79cd\u5f88\u6709\u610f\u601d\u7684\u67b6\u6784：<\/p>\n

Reactor-per-thread \/ Loop-per-thread\u3002<\/p>\n

\u8fd9\u79cd\u8bbe\u8ba1\u8fc7\u53bb\u5f53\u7136\u4e5f\u80fd\u5199，\u4f46 GIL \u4f1a\u4e25\u91cd\u9650\u5236\u591a\u4e2a\u7ebf\u7a0b\u540c\u65f6\u6267\u884c Python callback \u7684\u6536\u76ca；\u5728 free-threaded \u73af\u5883\u4e0b，\u5b83\u7b2c\u4e00\u6b21\u771f\u6b63\u6210\u4e3a\u503c\u5f97 benchmark \u7684\u591a\u6838 asyncio \u67b6\u6784\u3002<\/p>\n

\n

\u4e8c\u3001\u4ec0\u4e48\u65f6\u5019\u6211\u4f1a\u8003\u8651\u4e09\u4e2a\u7ebf\u7a0b\u3001\u4e09\u4e2a Event Loop？<\/h2>\n

\u6211\u4e0d\u4f1a\u628a\u5b83\u4f5c\u4e3a\u666e\u901a asyncio \u670d\u52a1\u7684\u9ed8\u8ba4\u67b6\u6784\u3002<\/p>\n

\u7edd\u5927\u591a\u6570 Web API\u3001\u6570\u636e\u5e93 CRUD\u3001HTTP \u805a\u5408\u670d\u52a1，\u9996\u5148\u5e94\u8be5\u4ece：<\/p>\n

1 Process
\n\u2514\u2500\u2500 1 Event Loop
\n \u2514\u2500\u2500 many coroutines<\/p>\n

\u5f00\u59cb\u3002<\/p>\n

\u53ea\u6709\u5f53 profiling \u660e\u786e\u544a\u8bc9\u6211 \u5355 loop \u672c\u8eab\u5f00\u59cb\u6210\u4e3a\u6267\u884c\u74f6\u9888，\u6211\u624d\u4f1a\u8003\u8651\u591a loop\u3002<\/p>\n

\u4e00\u4e2a\u5178\u578b\u573a\u666f\u662f：<\/p>\n

Socket I\/O
\n \u2193
\nProtocol decode
\n \u2193
\nJSON \/ message parsing
\n \u2193
\nValidation
\n \u2193
\nRouting
\n \u2193
\nBusiness rules
\n \u2193
\nAggregation
\n \u2193
\nNetwork I\/O<\/p>\n

\u8fd9\u91cc\u867d\u7136\u6574\u4f53\u770b\u8d77\u6765\u50cf\u201cI\/O \u670d\u52a1\u201d，\u4f46\u6bcf\u4e2a\u8bf7\u6c42\u4e4b\u95f4\u5b9e\u9645\u4e0a\u5939\u6742\u7740\u5927\u91cf Python \u8ba1\u7b97\u3002<\/p>\n

\u5047\u8bbe\u5355\u4e2a\u8bf7\u6c42\u53ea\u6709 300 \u5fae\u79d2 Python CPU \u5de5\u4f5c\u3002<\/p>\n

\u6bcf\u79d2 5000 \u4e2a\u8bf7\u6c42\u5c31\u662f：<\/p>\n

5000 \u00d7 300\u03bcs
\n\u2248 1.5 CPU seconds \/ second<\/p>\n

\u6b64\u65f6\u4e00\u4e2a event loop \u5df2\u7ecf\u4e0d\u53ef\u80fd\u9760\u201c\u66f4\u591a\u534f\u7a0b\u201d\u89e3\u51b3\u95ee\u9898\u3002<\/p>\n

\u534f\u7a0b\u89e3\u51b3\u7684\u662f\u7b49\u5f85\u3002<\/p>\n

\u5b83\u4e0d\u80fd\u5236\u9020\u7b2c\u4e8c\u9897 CPU\u3002<\/p>\n

\u5728 free-threaded Python \u4e0b，\u6211\u4eec\u53ef\u4ee5\u6309\u7167\u8fde\u63a5\u3001\u7528\u6237\u3001tenant\u3001partition \u6216 hash \u5bf9\u4efb\u52a1\u8fdb\u884c\u5206\u7247：<\/p>\n

hash(client_id) % 3<\/p>\n

\u250c\u2500\u2500 Loop 0 \/ Thread 0
\nRequest \u2500\u2500\u2500\u2500\u2500\u253c\u2500\u2500 Loop 1 \/ Thread 1
\n \u2514\u2500\u2500 Loop 2 \/ Thread 2<\/p>\n

\u6bcf\u4e2a loop \u72ec\u7acb\u5904\u7406\u81ea\u5df1\u7684\u8fde\u63a5\u548c\u72b6\u6001\u3002<\/p>\n

\u8fd9\u7c7b workload \u7279\u522b\u9002\u5408：<\/p>\n

\u9ad8\u5e76\u53d1\u7f51\u7edc\u670d\u52a1
\n\u5b9e\u65f6\u884c\u60c5\/Telemetry
\nWebSocket Gateway
\n\u6d88\u606f\u534f\u8bae\u89e3\u6790
\nIoT Gateway
\n\u6d41\u5f0f\u6570\u636e\u5904\u7406
\nRPC Proxy
\n\u5b9e\u65f6\u89c4\u5219\u8ba1\u7b97<\/p>\n

\u5b83\u7684\u4e00\u4e2a\u5de8\u5927\u4f18\u52bf\u662f：<\/p>\n

\u4e0d\u9700\u8981\u50cf multiprocessing \u4e00\u6837\u901a\u8fc7 IPC \u624d\u80fd\u8bbf\u95ee\u540c\u4e00\u5730\u5740\u7a7a\u95f4\u3002<\/p>\n

\u4f46\u8fd9\u4e5f\u662f\u6700\u5927\u7684\u5371\u9669\u3002<\/p>\n

\u56e0\u4e3a\u73b0\u5728\u5927\u5bb6\u771f\u7684\u5728\u540c\u65f6\u78b0\u540c\u4e00\u5757\u5185\u5b58\u3002<\/p>\n

\n

\u4e09\u3001Shared State：\u5171\u4eab\u5185\u5b58\u65e2\u662f\u4f18\u52bf，\u4e5f\u662f\u9677\u9631<\/h2>\n

Free-threaded \u6700\u8bf1\u4eba\u7684\u5730\u65b9\u4e4b\u4e00\u662f：<\/p>\n

shared_cache =<\/span> {<\/span>}<\/span><\/p>\n

\u4e09\u4e2a\u7ebf\u7a0b\u90fd\u80fd\u76f4\u63a5\u8bbf\u95ee\u5b83\u3002<\/p>\n

\u4f46\u201cPython \u4e0d\u5d29\u201d\u4e0e\u201c\u7a0b\u5e8f\u903b\u8f91\u6b63\u786e\u201d\u5b8c\u5168\u662f\u4e24\u56de\u4e8b\u3002<\/p>\n

Free-threaded CPython \u5bf9 dict\u3001list\u3001set \u7b49\u5185\u7f6e\u5bb9\u5668\u589e\u52a0\u4e86\u5185\u90e8\u540c\u6b65，\u4ee5\u907f\u514d\u5f88\u591a\u5e76\u53d1\u4fee\u6539\u9020\u6210\u7684\u6570\u636e\u7ed3\u6784\u635f\u574f；\u5b98\u65b9\u6587\u6863\u540c\u65f6\u660e\u786e\u5efa\u8bae，\u4e0d\u8981\u628a\u8fd9\u79cd\u5185\u90e8\u9501\u5f53\u6210\u5e94\u7528\u7a0b\u5e8f\u540c\u6b65\u534f\u8bae，\u5e94\u5728\u9700\u8981\u65f6\u4f7f\u7528 threading.Lock \u7b49\u540c\u6b65\u673a\u5236\u3002<\/p>\n

\u4f8b\u5982：<\/p>\n

if<\/span> key not<\/span> in<\/span> cache:<\/span>
\n cache[<\/span>key]<\/span> =<\/span> expensive_load(<\/span>key)<\/span><\/p>\n

\u5373\u4f7f：<\/p>\n

key in<\/span> cache<\/p>\n

\u548c：<\/p>\n

cache[<\/span>key]<\/span> =<\/span> value<\/p>\n

\u5355\u72ec\u90fd\u4e0d\u4f1a\u628a\u5b57\u5178\u7ed3\u6784\u641e\u574f，\u8fd9\u7ec4\u64cd\u4f5c\u4ecd\u7136\u4e0d\u662f\u4e00\u4e2a\u539f\u5b50\u7684\u4e1a\u52a1\u4e8b\u52a1\u3002<\/p>\n

\u4e24\u4e2a\u7ebf\u7a0b\u53ef\u80fd\u540c\u65f6\u53d1\u73b0 key \u4e0d\u5b58\u5728，\u7136\u540e\u8ba1\u7b97\u4e24\u904d\u3002<\/p>\n

\u771f\u6b63\u5b89\u5168\u7684\u5199\u6cd5\u53ef\u80fd\u662f：<\/p>\n

import<\/span> threading<\/p>\n

cache =<\/span> {<\/span>}<\/span>
\ncache_lock =<\/span> threading.<\/span>Lock(<\/span>)<\/span><\/p>\n

def<\/span> get_or_create<\/span>(<\/span>key)<\/span>:<\/span>
\n with<\/span> cache_lock:<\/span>
\n if<\/span> key in<\/span> cache:<\/span>
\n return<\/span> cache[<\/span>key]<\/span><\/p>\n

value =<\/span> expensive_load(<\/span>key)<\/span>
\n cache[<\/span>key]<\/span> =<\/span> value
\n return<\/span> value<\/p>\n

\u4f46\u65b0\u7684\u95ee\u9898\u53c8\u6765\u4e86：<\/p>\n

\u5982\u679c expensive_load() \u5f88\u6162，\u4f60\u628a\u4e09\u4e2a CPU \u6838\u91cd\u65b0\u4e32\u884c\u5316\u4e86\u3002<\/p>\n

\u6240\u4ee5\u5b9e\u9645\u5de5\u7a0b\u4e2d，\u6211\u66f4\u63a8\u8350：<\/p>\n

Shared everything
\n \u2193
\n\u5c3d\u91cf\u907f\u514d<\/p>\n

Shared immutable data
\n \u2193
\n\u5f88\u597d<\/p>\n

Partitioned mutable state
\n \u2193
\n\u6700\u597d<\/p>\n

\u4f8b\u5982：<\/p>\n

Loop 0 -> cache shard 0
\nLoop 1 -> cache shard 1
\nLoop 2 -> cache shard 2<\/p>\n

\u6839\u636e：<\/p>\n

shard_id =<\/span> hash<\/span>(<\/span>key)<\/span> %<\/span> worker_count<\/p>\n

\u786e\u5b9a\u67d0\u4e2a key \u7684\u552f\u4e00 owner\u3002<\/p>\n

\u8fd9\u6837\u7cfb\u7edf\u4ece：<\/p>\n

shared-memory concurrency<\/p>\n

\u9010\u6e10\u53d8\u6210：<\/p>\n

ownership-based concurrency<\/p>\n

\u800c ownership \u5f80\u5f80\u6bd4 lock \u66f4\u5bb9\u6613\u63a8\u7406\u3002<\/p>\n

\n

\u56db\u3001Thread Safety：\u8981\u540c\u65f6\u7406\u89e3 threading \u4e0e asyncio \u4e24\u5957\u540c\u6b65\u6a21\u578b<\/h2>\n

\u591a loop \u67b6\u6784\u91cc\u6700\u5371\u9669\u7684\u4e00\u4ef6\u4e8b，\u662f\u6df7\u6dc6：<\/p>\n

asyncio.<\/span>Lock<\/p>\n

\u548c：<\/p>\n

threading.<\/span>Lock<\/p>\n

\u5b83\u4eec\u89e3\u51b3\u7684\u4e0d\u662f\u540c\u4e00\u4e2a\u95ee\u9898\u3002<\/p>\n

asyncio.Lock \u7528\u4e8e：<\/p>\n

\u540c\u4e00\u4e2a Event Loop
\n\u4e0d\u540c coroutine<\/p>\n

\u800c\u8de8\u7ebf\u7a0b\u5171\u4eab\u72b6\u6001\u5e94\u8be5\u4f7f\u7528：<\/p>\n

threading.Lock
\nthreading.RLock
\nqueue.Queue
\nSemaphore
\nCondition<\/p>\n

\u66f4\u91cd\u8981\u7684\u662f：<\/p>\n

asyncio \u4e2d\u7edd\u5927\u591a\u6570\u5bf9\u8c61\u4e0d\u662f thread-safe \u7684\u3002<\/p>\n

\u5b98\u65b9\u5efa\u8bae，\u4ece\u5176\u4ed6 OS \u7ebf\u7a0b\u64cd\u4f5c\u4e00\u4e2a event loop \u65f6，\u5e94\u4f7f\u7528：<\/p>\n

loop.<\/span>call_soon_threadsafe(<\/span>.<\/span>.<\/span>.<\/span>)<\/span><\/p>\n

\u6216\u8005：<\/p>\n

asyncio.<\/span>run_coroutine_threadsafe(<\/span>.<\/span>.<\/span>.<\/span>)<\/span><\/p>\n

\u800c\u4e0d\u662f\u76f4\u63a5\u64cd\u4f5c\u5c5e\u4e8e\u53e6\u4e00\u4e2a loop \u7684 Future\u3001Task\u3001Queue \u7b49\u5bf9\u8c61\u3002<\/p>\n

\u4f8b\u5982\u4e0d\u8981\u8fd9\u6837：<\/p>\n

# Thread A<\/span>
\nworker_loop.<\/span>create_task(<\/span>coro(<\/span>)<\/span>)<\/span><\/p>\n

\u5e94\u8be5：<\/p>\n

asyncio.<\/span>run_coroutine_threadsafe(<\/span>
\n coro(<\/span>)<\/span>,<\/span>
\n worker_loop,<\/span>
\n)<\/span><\/p>\n

\u8fd8\u6709\u4e00\u4e2a\u975e\u5e38\u5b9e\u7528\u7684\u539f\u5219：<\/p>\n

\u4e0d\u8981\u62ff\u7740 threading.Lock \u53bb await\u3002<\/p>\n

\u4f8b\u5982：<\/p>\n

with<\/span> shared_lock:<\/span>
\n await<\/span> query_database(<\/span>)<\/span><\/p>\n

\u610f\u5473\u7740\u8fd9\u628a\u8de8\u7ebf\u7a0b\u9501\u53ef\u80fd\u5728\u4e00\u6b21\u7f51\u7edc\u7b49\u5f85\u671f\u95f4\u4e00\u76f4\u88ab\u5360\u7528\u3002<\/p>\n

\u66f4\u5408\u7406\u7684\u662f：<\/p>\n

with<\/span> shared_lock:<\/span>
\n snapshot =<\/span> shared_state.<\/span>copy(<\/span>)<\/span><\/p>\n

result =<\/span> await<\/span> query_database(<\/span>snapshot)<\/span><\/p>\n

\u628a\u540c\u6b65\u533a\u57df\u7f29\u5230\u6700\u5c0f\u3002<\/p>\n

\n

\u4e94\u3001Loop Ownership：\u8c01\u521b\u5efa，\u8c01\u4f7f\u7528，\u8c01\u9500\u6bc1<\/h2>\n

\u5728\u591a event loop \u7cfb\u7edf\u91cc，\u6211\u8ba4\u4e3a\u6700\u91cd\u8981\u7684\u8bbe\u8ba1\u539f\u5219\u53ea\u6709\u4e00\u53e5：<\/p>\n

Loop-bound resource belongs to one loop for its entire lifetime.<\/p>\n

\u6bd4\u5982：<\/p>\n

asyncio.Future
\nasyncio.Queue
\nasyncio.Lock
\nHTTP ClientSession
\nDB AsyncEngine
\nConnection Pool
\nTransport
\nTask<\/p>\n

\u4e0d\u8981：<\/p>\n

Main Thread \u521b\u5efa
\n \u2193
\nThread 2 \u4f7f\u7528
\n \u2193
\nThread 3 close<\/p>\n

\u800c\u5e94\u8be5：<\/p>\n

Thread 2
\n\u2514\u2500\u2500 Event Loop 2
\n \u251c\u2500\u2500 create pool
\n \u251c\u2500\u2500 use pool
\n \u2514\u2500\u2500 close pool<\/p>\n

\u4e00\u4e2a\u7b80\u5355\u7684 worker \u53ef\u4ee5\u8fd9\u6837\u5199：<\/p>\n

import<\/span> asyncio
\nimport<\/span> threading
\nfrom<\/span> concurrent.<\/span>futures import<\/span> Future<\/p>\n

class<\/span> LoopWorker<\/span>:<\/span>
\n def<\/span> __init__<\/span>(<\/span>self,<\/span> worker_id:<\/span> int<\/span>)<\/span>:<\/span>
\n self.<\/span>worker_id =<\/span> worker_id
\n self.<\/span>_ready =<\/span> threading.<\/span>Event(<\/span>)<\/span><\/p>\n

self.<\/span>_thread =<\/span> threading.<\/span>Thread(<\/span>
\n target=<\/span>self.<\/span>_thread_main,<\/span>
\n name=<\/span>f"loop-worker-<\/span>{<\/span>worker_id}<\/span><\/span>"<\/span><\/span>,<\/span>
\n )<\/span><\/p>\n

self.<\/span>loop =<\/span> None<\/span>
\n self.<\/span>_stop =<\/span> None<\/span><\/p>\n

def<\/span> start<\/span>(<\/span>self)<\/span>:<\/span>
\n self.<\/span>_thread.<\/span>start(<\/span>)<\/span>
\n self.<\/span>_ready.<\/span>wait(<\/span>)<\/span><\/p>\n

def<\/span> _thread_main<\/span>(<\/span>self)<\/span>:<\/span>
\n asyncio.<\/span>run(<\/span>self.<\/span>_main(<\/span>)<\/span>)<\/span><\/p>\n

async<\/span> def<\/span> _main<\/span>(<\/span>self)<\/span>:<\/span>
\n self.<\/span>loop =<\/span> asyncio.<\/span>get_running_loop(<\/span>)<\/span>
\n self.<\/span>_stop =<\/span> asyncio.<\/span>Event(<\/span>)<\/span><\/p>\n

# \u5728\u8fd9\u91cc\u521b\u5efa\u5c5e\u4e8e\u5f53\u524d loop \u7684\u8d44\u6e90<\/span>
\n # self.http = aiohttp.ClientSession(…)<\/span>
\n # self.db = create_async_engine(…)<\/span><\/p>\n

self.<\/span>_ready.<\/span>set<\/span>(<\/span>)<\/span><\/p>\n

try<\/span>:<\/span>
\n await<\/span> self.<\/span>_stop.<\/span>wait(<\/span>)<\/span>
\n finally<\/span>:<\/span>
\n # \u5728\u5b83\u6240\u5c5e\u7684 loop \u4e2d\u91ca\u653e<\/span>
\n # await self.http.close()<\/span>
\n # await self.db.dispose()<\/span>
\n pass<\/span><\/p>\n

def<\/span> submit<\/span>(<\/span>self,<\/span> coro)<\/span> –<\/span>><\/span> Future:<\/span>
\n return<\/span> asyncio.<\/span>run_coroutine_threadsafe(<\/span>
\n coro,<\/span>
\n self.<\/span>loop,<\/span>
\n )<\/span><\/p>\n

def<\/span> stop<\/span>(<\/span>self)<\/span>:<\/span>
\n self.<\/span>loop.<\/span>call_soon_threadsafe(<\/span>
\n self.<\/span>_stop.<\/span>set<\/span>
\n )<\/span>
\n self.<\/span>_thread.<\/span>join(<\/span>)<\/span><\/p>\n

\u542f\u52a8\u4e09\u4e2a worker：<\/p>\n

workers =<\/span> [<\/span>
\n LoopWorker(<\/span>i)<\/span>
\n for<\/span> i in<\/span> range<\/span>(<\/span>3<\/span>)<\/span>
\n]<\/span><\/p>\n

for<\/span> worker in<\/span> workers:<\/span>
\n worker.<\/span>start(<\/span>)<\/span><\/p>\n

\u7136\u540e\u8fdb\u884c hash routing：<\/p>\n

def<\/span> choose_worker<\/span>(<\/span>key)<\/span>:<\/span>
\n return<\/span> workers[<\/span>hash<\/span>(<\/span>key)<\/span> %<\/span> len<\/span>(<\/span>workers)<\/span>]<\/span><\/p>\n

future =<\/span> choose_worker(<\/span>user_id)<\/span>.<\/span>submit(<\/span>
\n handle_request(<\/span>user_id)<\/span>
\n)<\/span><\/p>\n

result =<\/span> future.<\/span>result(<\/span>)<\/span><\/p>\n

\u8fd9\u4e2a\u6a21\u578b\u6700\u91cd\u8981\u7684\u4e0d\u662f\u201c\u4e09\u4e2a loop\u201d\u3002<\/p>\n

\u800c\u662f：<\/p>\n

key
\n \u2193
\n\u56fa\u5b9a shard
\n \u2193
\n\u56fa\u5b9a Thread
\n \u2193
\n\u56fa\u5b9a Loop
\n \u2193
\n\u56fa\u5b9a Resource Set<\/p>\n

\u8fd9\u6837\u7cfb\u7edf\u4f1a\u5bb9\u6613\u7406\u89e3\u5f88\u591a\u3002<\/p>\n

\n

\u516d\u3001Connection Pools：\u6700\u5bb9\u6613\u8e29\u5751\u7684\u5730\u65b9<\/h2>\n

\u591a event loop \u540e，\u8fde\u63a5\u6c60\u4e0d\u80fd\u518d\u60f3\u5f53\u7136\u5730\u5168\u5c40\u5171\u4eab\u3002<\/p>\n

\u5047\u8bbe：<\/p>\n

Loop 1 \u2500\u2500\u2510
\nLoop 2 \u2500\u2500\u253c\u2500\u2500 Global Async DB Pool
\nLoop 3 \u2500\u2500\u2518<\/p>\n

\u8fd9\u5f80\u5f80\u4e0d\u662f\u4e00\u4e2a\u597d\u8bbe\u8ba1\u3002<\/p>\n

\u4ee5 SQLAlchemy \u7684\u5f02\u6b65\u5f15\u64ce\u4e3a\u4f8b，\u5b98\u65b9\u660e\u786e\u8bf4\u660e：\u4f7f\u7528\u9ed8\u8ba4 pool \u65f6，\u4e0d\u5e94\u628a\u540c\u4e00\u4e2a AsyncEngine \u5728\u4e0d\u540c event loop \u4e4b\u95f4\u5171\u4eab；\u5426\u5219\u53ef\u80fd\u51fa\u73b0\u5178\u578b\u7684：<\/p>\n

Future attached to a different loop<\/p>\n

\u9519\u8bef\u3002\u5982\u679c\u786e\u5b9e\u8981\u8de8 loop \u8f6c\u79fb，\u9700\u8981\u5148 dispose()；\u5982\u679c\u5fc5\u987b\u8de8 loop \u4f7f\u7528，\u5219\u53ef\u8003\u8651 NullPool，\u4e0d\u8fc7\u8fd9\u901a\u5e38\u610f\u5473\u7740\u653e\u5f03\u8fde\u63a5\u590d\u7528\u3002<\/p>\n

\u56e0\u6b64\u66f4\u63a8\u8350：<\/p>\n

Loop 1 -> DB Pool 1
\nLoop 2 -> DB Pool 2
\nLoop 3 -> DB Pool 3<\/p>\n

HTTP client \u4e5f\u662f\u7c7b\u4f3c\u601d\u8def\u3002aiohttp.ClientSession \u81ea\u8eab\u7ef4\u62a4 connection pool，\u5e94\u8ba9 session \u4e0e\u521b\u5efa\u3001\u4f7f\u7528\u5b83\u7684\u5f02\u6b65\u8fd0\u884c\u73af\u5883\u4fdd\u6301\u4e00\u81f4\u3002<\/p>\n

\u4f46\u8fd9\u91cc\u8fd8\u6709\u7b2c\u4e8c\u4e2a\u5751：<\/p>\n

\u5047\u8bbe\u6570\u636e\u5e93\u6700\u591a\u5141\u8bb8：<\/p>\n

100 connections<\/p>\n

\u539f\u6765：<\/p>\n

1 loop \u00d7 pool_size 60
\n= 60<\/p>\n

\u6539\u6210：<\/p>\n

4 loops \u00d7 pool_size 60
\n= 240<\/p>\n

\u6570\u636e\u5e93\u53ef\u80fd\u5148\u88ab\u4f60\u6253\u6b7b\u3002<\/p>\n

\u6240\u4ee5\u591a loop \u540e\u6240\u6709\u8d44\u6e90\u9650\u5236\u90fd\u5fc5\u987b\u91cd\u65b0\u7406\u89e3\u6210：<\/p>\n

Process Budget
\n \u2193
\nShard Budget<\/p>\n

\u4f8b\u5982：<\/p>\n

DB max = 100<\/p>\n

4 loops<\/p>\n

\u6bcf\u4e2a pool \u2248 20
\n\u9884\u7559 \u2248 20<\/p>\n

HTTP concurrency\u3001rate limiter\u3001Redis connection\u3001Kafka producer \u7b49\u8d44\u6e90\u90fd\u8981\u9075\u5faa\u540c\u6837\u539f\u5219\u3002<\/p>\n

\n

\u4e03\u3001Signal Handling：\u5fc5\u987b\u8bbe\u8ba1\u4e00\u4e2a Control Plane<\/h2>\n

\u591a\u7ebf\u7a0b\u591a loop \u67b6\u6784\u53e6\u4e00\u4e2a\u5f88\u5bb9\u6613\u88ab\u5ffd\u7565\u7684\u95ee\u9898\u662f signal\u3002<\/p>\n

Python \u7684 signal handler \u7531 main Python thread of the main interpreter \u6267\u884c，\u800c\u4e14\u53ea\u6709\u4e3b\u7ebf\u7a0b\u53ef\u4ee5\u5b89\u88c5\u65b0\u7684 signal handler；asyncio \u6587\u6863\u4e5f\u660e\u786e\u6307\u51fa，\u5904\u7406 signals \u7684 event loop \u5fc5\u987b\u8fd0\u884c\u5728\u4e3b\u7ebf\u7a0b\u3002<\/p>\n

\u56e0\u6b64\u4e0d\u8981\u8ba9\u4e09\u4e2a worker \u90fd\u5c1d\u8bd5：<\/p>\n

loop.<\/span>add_signal_handler(<\/span>.<\/span>.<\/span>.<\/span>)<\/span><\/p>\n

\u66f4\u597d\u7684\u8bbe\u8ba1\u662f：<\/p>\n

Main Thread
\n Control Plane
\n SIGTERM\/SIGINT
\n \u2502
\n \u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u253c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510
\n \u2193 \u2193 \u2193
\n Loop 1 Loop 2 Loop 3<\/p>\n

\u4e3b\u7ebf\u7a0b\u6536\u5230：<\/p>\n

SIGTERM<\/p>\n

\u4e4b\u540e\u8c03\u7528：<\/p>\n

worker.<\/span>loop.<\/span>call_soon_threadsafe(<\/span>
\n worker.<\/span>stop_event.<\/span>set<\/span>
\n)<\/span><\/p>\n

\u901a\u77e5\u6bcf\u4e2a owner loop \u81ea\u5df1：<\/p>\n

\u505c\u6b62\u63a5\u53d7\u65b0\u4efb\u52a1
\n\u2192 drain existing tasks
\n\u2192 close HTTP pool
\n\u2192 close DB pool
\n\u2192 shutdown async generators
\n\u2192 stop loop<\/p>\n

\u4e0d\u8981\u7531\u4e3b\u7ebf\u7a0b\u76f4\u63a5\u5173\u95ed worker \u7684\u5f02\u6b65\u5bf9\u8c61\u3002<\/p>\n

\u4ecd\u7136\u9075\u5faa\u90a3\u6761\u539f\u5219：<\/p>\n

\u8c01\u62e5\u6709\u8d44\u6e90，\u8c01\u8d1f\u8d23 teardown\u3002<\/p>\n

\n

\u516b\u3001Debugging Complexity：\u6027\u80fd\u63d0\u9ad8\u4e86，\u8ba4\u77e5\u6210\u672c\u4e5f\u63d0\u9ad8\u4e86<\/h2>\n

\u4e00\u4e2a event loop \u51fa\u95ee\u9898\u65f6，\u6211\u4eec\u901a\u5e38\u53ea\u9700\u8981\u7406\u89e3：<\/p>\n

Task A
\nTask B
\nTask C<\/p>\n

\u4e09\u4e2a loop \u540e\u53d8\u6210：<\/p>\n

Thread
\n \u251c\u2500\u2500 Loop
\n \u2502 \u251c\u2500\u2500 Task
\n \u2502 \u251c\u2500\u2500 Future
\n \u2502 \u2514\u2500\u2500 Connection
\n \u2502
\nShared State
\n \u2502
\nLock
\n \u2502
\nCross-loop message<\/p>\n

Race condition \u4e5f\u53ef\u80fd\u4ece：<\/p>\n

\u6bcf\u6b21\u90fd\u53d1\u751f<\/p>\n

\u53d8\u6210：<\/p>\n

\u751f\u4ea7\u73af\u5883\u4e00\u5929\u4e00\u6b21\u3002<\/p>\n

\u8fd9\u624d\u662f\u771f\u6b63\u53ef\u6015\u7684 bug\u3002<\/p>\n

\u56e0\u6b64\u5efa\u8bae\u4ece\u7b2c\u4e00\u5929\u5c31\u8ba9\u65e5\u5fd7\u5305\u542b：<\/p>\n

timestamp
\nthread_name
\nworker_id
\nloop_id
\ntask_name
\nrequest_id
\ntrace_id<\/p>\n

\u4f8b\u5982：<\/p>\n

threading.<\/span>current_thread(<\/span>)<\/span>.<\/span>name<\/p>\n

\u53ef\u4ee5\u76f4\u63a5\u544a\u8bc9\u4f60\u4ee3\u7801\u6b63\u5728\u54ea\u4e2a worker \u4e2d\u8fd0\u884c\u3002<\/p>\n

\u540c\u65f6\u5f3a\u70c8\u5efa\u8bae\u5728\u5f00\u53d1\u73af\u5883\u5f00\u542f：<\/p>\n

PYTHONASYNCIODEBUG<\/span>=<\/span>1<\/span><\/p>\n

asyncio debug mode \u53ef\u4ee5\u68c0\u6d4b\u90e8\u5206\u9519\u8bef\u7ebf\u7a0b\u8c03\u7528，\u5e76\u8bb0\u5f55\u6162 callback；\u5f53\u524d\u6587\u6863\u4e2d\u9ed8\u8ba4\u4f1a\u62a5\u544a\u6267\u884c\u65f6\u95f4\u8d85\u8fc7 100ms \u7684 callback，\u4e5f\u53ef\u4ee5\u901a\u8fc7 loop.slow_callback_duration \u8c03\u6574\u3002<\/p>\n

\u6bd4\u5982：<\/p>\n

loop =<\/span> asyncio.<\/span>get_running_loop(<\/span>)<\/span><\/p>\n

loop.<\/span>set_debug(<\/span>True<\/span>)<\/span>
\nloop.<\/span>slow_callback_duration =<\/span> 0.02<\/span><\/p>\n

\u628a 20ms \u4ee5\u4e0a\u7684 callback \u66b4\u9732\u51fa\u6765\u3002<\/p>\n

\u5728\u4f4e\u5ef6\u8fdf\u7cfb\u7edf\u4e2d，\u8fd9\u79cd\u6307\u6807\u751a\u81f3\u5e94\u8be5\u76f4\u63a5\u7eb3\u5165\u76d1\u63a7：<\/p>\n

loop lag p50
\nloop lag p95
\nloop lag p99<\/p>\n

\u56e0\u4e3a CPU \u5229\u7528\u7387 60% \u5e76\u4e0d\u4ee3\u8868 event loop \u5065\u5eb7\u3002<\/p>\n

\u4e00\u4e2a worker \u5982\u679c\u6267\u884c\u4e86 200ms \u4e0d\u5e26 await \u7684 Python \u4ee3\u7801：<\/p>\n

async<\/span> def<\/span> handler<\/span>(<\/span>)<\/span>:<\/span>
\n result =<\/span> expensive_python_work(<\/span>)<\/span><\/p>\n

\u5176\u4ed6 CPU \u4e0a\u7684 loop \u53ef\u4ee5\u7ee7\u7eed\u8fd0\u884c，\u4f46 \u8fd9\u4e2a worker \u81ea\u5df1\u7684\u6240\u6709 I\/O \u90fd\u4f1a\u505c 200ms\u3002<\/p>\n

Free-threaded \u89e3\u51b3\u7684\u662f：<\/p>\n

\u8de8\u7ebf\u7a0b\u5e76\u884c<\/p>\n

\u5b83\u6ca1\u6709\u89e3\u51b3：<\/p>\n

\u5355 event loop cooperative scheduling<\/p>\n

\u8fd9\u4e00\u533a\u522b\u975e\u5e38\u91cd\u8981\u3002<\/p>\n

\n

\u4e5d\u3001\u4e00\u4e2a\u66f4\u6210\u719f\u7684\u67b6\u6784：Shared-Nothing-ish，\u800c\u4e0d\u662f Shared-Everything<\/h2>\n

\u5982\u679c\u8ba9\u6211\u771f\u6b63\u8bbe\u8ba1\u8fd9\u6837\u7684\u751f\u4ea7\u7cfb\u7edf，\u6211\u4f1a\u503e\u5411\u4e8e：<\/p>\n

Main Thread
\n Control Plane
\n \u2502
\n \u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u253c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510
\n \u2193 \u2193 \u2193
\n Thread 0 Thread 1 Thread 2
\n \u2502 \u2502 \u2502
\n Loop 0 Loop 1 Loop 2
\n \u2502 \u2502 \u2502
\n HTTP-0 HTTP-1 HTTP-2
\n DB-0 DB-1 DB-2
\n Cache-0 Cache-1 Cache-2
\n \u2502 \u2502 \u2502
\n Shard 0 Shard 1 Shard 2<\/p>\n

\u8fdb\u7a0b\u91cc\u771f\u6b63\u5171\u4eab\u7684\u4e1c\u897f\u5c3d\u91cf\u53ea\u5269\u4e0b：<\/p>\n

Immutable Config
\nMetrics
\nAtomic-ish counters \/ protected state
\nThread-safe message channels<\/p>\n

\u4e1a\u52a1\u72b6\u6001\u5c3d\u53ef\u80fd shard\u3002<\/p>\n

\u8fd9\u6837\u867d\u7136\u8fd0\u884c\u5728 shared-memory process \u4e2d，\u7f16\u7a0b\u6a21\u578b\u5374\u66f4\u63a5\u8fd1 actor：<\/p>\n

\u4e00\u4e2a\u72b6\u6001
\n\u53ea\u6709\u4e00\u4e2a owner
\n\u5176\u4ed6\u7ebf\u7a0b\u5411 owner \u53d1\u6d88\u606f<\/p>\n

\u8fd9\u5f80\u5f80\u6bd4\u201c\u5927\u5bb6\u5171\u4eab\u4e00\u4e2a dict，\u7136\u540e\u5230\u5904\u52a0\u9501\u201d\u7a33\u5b9a\u5f97\u591a\u3002<\/p>\n

\n

\u5341\u3001\u4ec0\u4e48\u65f6\u5019\u4e0d\u8981\u4f7f\u7528\u8fd9\u79cd\u6a21\u578b？<\/h2>\n

\u4e00\u4e2a\u975e\u5e38\u91cd\u8981\u7684\u5de5\u7a0b\u80fd\u529b，\u662f\u77e5\u9053\u4ec0\u4e48\u65f6\u5019\u4e0d\u5e94\u8be5\u4f7f\u7528\u65b0\u6280\u672f\u3002<\/p>\n

\u6211\u4f1a\u628a\u5224\u65ad\u6807\u51c6\u538b\u7f29\u6210\u4e0b\u9762\u8fd9\u5f20\u8868：<\/p>\n\n\u573a\u666f\u5efa\u8bae<\/tr>\n\n\n\n\n\n\n\n\n\n\n
\u5927\u91cf I\/O，CPU \u5360\u7528\u5f88\u4f4e<\/td>\n\u5355 event loop<\/td>\n<\/tr>\n
\u5355 loop \u5df2\u7ecf\u8dd1\u6ee1\u4e00\u4e2a CPU<\/td>\n\u8003\u8651\u591a loop<\/td>\n<\/tr>\n
I\/O + \u5927\u91cf Python callback\/\u89e3\u6790<\/td>\n\u5f88\u503c\u5f97 benchmark<\/td>\n<\/tr>\n
\u5b8c\u5168 CPU batch<\/td>\n\u5148\u8003\u8651\u666e\u901a\u7ebf\u7a0b\u6c60<\/td>\n<\/tr>\n
\u9700\u8981\u8fdb\u7a0b\u7ea7\u6545\u969c\u9694\u79bb<\/td>\nmultiprocessing<\/td>\n<\/tr>\n
\u5927\u91cf mutable global state<\/td>\n\u8c28\u614e\u4f7f\u7528<\/td>\n<\/tr>\n
\u7b2c\u4e09\u65b9 C \u6269\u5c55\u5c1a\u672a\u652f\u6301 free-threading<\/td>\n\u6682\u7f13<\/td>\n<\/tr>\n
workload \u53ef\u4ee5\u5929\u7136 shard<\/td>\n\u975e\u5e38\u9002\u5408<\/td>\n<\/tr>\n
\u56e2\u961f\u7f3a\u4e4f\u5e76\u53d1\u8c03\u8bd5\u7ecf\u9a8c<\/td>\n\u4ece\u7b80\u5355\u6a21\u578b\u5f00\u59cb<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

\u5c24\u5176\u8981\u6ce8\u610f\u6700\u540e\u4e00\u4e2a\u73b0\u5b9e\u95ee\u9898：<\/p>\n

\u67d0\u4e9b\u7b2c\u4e09\u65b9 C \u6269\u5c55\u5982\u679c\u6ca1\u6709\u58f0\u660e\u652f\u6301 free-threading，\u5bfc\u5165\u540e\u53ef\u80fd\u91cd\u65b0\u542f\u7528 GIL\u3002Python \u5b98\u65b9\u6587\u6863\u660e\u786e\u8bf4\u660e\u4e86\u8fd9\u4e00\u884c\u4e3a\u3002<\/p>\n

\u6240\u4ee5\u542f\u52a8\u670d\u52a1\u65f6\u6700\u597d\u68c0\u67e5\u8fd0\u884c\u73af\u5883：<\/p>\n

import<\/span> sys
\nimport<\/span> sysconfig<\/p>\n

print<\/span>(<\/span>
\n "free-threaded build:"<\/span>,<\/span>
\n sysconfig.<\/span>get_config_var(<\/span>"Py_GIL_DISABLED"<\/span>)<\/span>,<\/span>
\n)<\/span><\/p>\n

if<\/span> hasattr<\/span>(<\/span>sys,<\/span> "_is_gil_enabled"<\/span>)<\/span>:<\/span>
\n print<\/span>(<\/span>
\n "GIL enabled:"<\/span>,<\/span>
\n sys.<\/span>_is_gil_enabled(<\/span>)<\/span>,<\/span>
\n )<\/span><\/p>\n

\u4e0d\u8981\u4ee5\u4e3a\u4f60\u5b89\u88c5\u4e86 free-threaded Python，\u5c31\u610f\u5473\u7740\u5b9e\u9645 workload \u59cb\u7ec8\u5904\u4e8e\u65e0 GIL \u72b6\u6001\u3002<\/p>\n

\n

\u5341\u4e00\u3001\u600e\u6837\u8bc1\u660e\u5b83\u771f\u7684\u503c\u5f97？<\/h2>\n

\u4e0d\u8981\u901a\u8fc7\u201c\u7406\u8bba\u4e0a\u4e09\u9897 CPU\u201d\u51b3\u5b9a\u67b6\u6784\u3002<\/p>\n

\u901a\u8fc7 benchmark\u3002<\/p>\n

\u5efa\u8bae\u81f3\u5c11\u540c\u65f6\u89c2\u5bdf：<\/p>\n

throughput
\nrequests \/ second<\/p>\n

latency
\np50 \/ p95 \/ p99 \/ max<\/p>\n

event-loop lag<\/p>\n

CPU utilization
\nper-core utilization<\/p>\n

memory \/ RSS<\/p>\n

DB connections<\/p>\n

lock contention<\/p>\n

context switches<\/p>\n

error rate<\/p>\n

\u6d4b\u8bd5：<\/p>\n

A:
\n1 process
\n1 thread
\n1 event loop<\/p>\n

B:
\n1 process
\n2 threads
\n2 event loops<\/p>\n

C:
\n1 process
\n4 threads
\n4 event loops<\/p>\n

D:
\n4 processes
\n1 loop\/process<\/p>\n

\u7136\u540e\u6bd4\u8f83\u3002<\/p>\n

\u771f\u6b63\u503c\u5f97\u5173\u6ce8\u7684\u4e0d\u662f：<\/p>\n

\u5e73\u5747\u541e\u5410\u63d0\u5347 20%<\/p>\n

\u800c\u662f：<\/p>\n

p99 \u662f\u5426\u4e0b\u964d？<\/p>\n

CPU \u662f\u5426\u771f\u6b63\u5747\u5300\u4f7f\u7528？<\/p>\n

connection \u6570\u91cf\u662f\u5426\u5931\u63a7？<\/p>\n

\u9501\u7ade\u4e89\u662f\u5426\u62b5\u6d88\u4e86\u5e76\u884c\u6536\u76ca？<\/p>\n

RSS \u76f8\u6bd4\u591a\u8fdb\u7a0b\u8282\u7701\u591a\u5c11？<\/p>\n

\u6545\u969c\u548c shutdown \u662f\u5426\u4ecd\u7136\u53ef\u63a7？<\/p>\n

\u5982\u679c\u56db\u4e2a loop \u6700\u7ec8\u90fd\u5728\u62a2：<\/p>\n

global_cache_lock<\/p>\n

\u90a3\u4e48\u4f60\u53ea\u662f\u628a\u8fc7\u53bb\u7684：<\/p>\n

GIL<\/p>\n

\u6362\u6210\u4e86：<\/p>\n

MyOwnGIL\u2122<\/p>\n

\u67b6\u6784\u5e76\u6ca1\u6709\u771f\u6b63\u8fdb\u6b65\u3002<\/p>\n

\n

\u5341\u4e8c\u3001Free-Threaded Python \u771f\u6b63\u5e26\u6765\u7684，\u4e0d\u53ea\u662f\u201cPython \u53ef\u4ee5\u591a\u7ebf\u7a0b\u201d<\/h2>\n

\u6211\u8ba4\u4e3a free-threaded Python \u6700\u6709\u4ef7\u503c\u7684\u610f\u4e49，\u5e76\u4e0d\u662f\u7b80\u5355\u5730\u628a：<\/p>\n

ThreadPoolExecutor(<\/span>max_workers=<\/span>8<\/span>)<\/span><\/p>\n

\u6539\u6210 CPU \u5e76\u884c\u3002<\/p>\n

\u66f4\u6709\u610f\u601d\u7684\u662f，\u5b83\u91cd\u65b0\u6253\u5f00\u4e86 Python \u670d\u52a1\u7aef\u67b6\u6784\u7684\u8bbe\u8ba1\u7a7a\u95f4\u3002<\/p>\n

\u4ee5\u524d：<\/p>\n

asyncio
\n\u2248 \u5355\u7ebf\u7a0b\u4e8b\u4ef6\u9a71\u52a8<\/p>\n

threading
\n\u2248 I\/O concurrency<\/p>\n

multiprocessing
\n\u2248 CPU parallelism<\/p>\n

\u73b0\u5728\u8fb9\u754c\u5f00\u59cb\u53d8\u6210：<\/p>\n

asyncio
\n+
\nfree-threaded threading
\n+
\nshared memory
\n+
\nownership \/ sharding<\/p>\n

\u6211\u4eec\u53ef\u4ee5\u6784\u5efa\u4e00\u79cd\u4ecb\u4e8e\u4f20\u7edf asyncio \u4e0e multiprocessing \u4e4b\u95f4\u7684\u6a21\u578b：<\/p>\n

\u5355\u8fdb\u7a0b
\n\u591a\u6838
\n\u591a reactor
\n\u5171\u4eab\u5730\u5740\u7a7a\u95f4
\n\u5c40\u90e8\u72b6\u6001\u9694\u79bb
\n\u4f4e\u6210\u672c\u7ebf\u7a0b\u95f4\u901a\u4fe1<\/p>\n

\u8fd9\u5bf9\u5b9e\u65f6\u7cfb\u7edf\u3001\u7f51\u7edc\u7f51\u5173\u3001\u6570\u636e\u6d41\u7cfb\u7edf\u548c\u9ad8\u6027\u80fd Python \u670d\u52a1\u5c24\u5176\u503c\u5f97\u63a2\u7d22\u3002<\/p>\n

\u4f46\u65b0\u7684\u81ea\u7531\u4e5f\u610f\u5473\u7740\u65b0\u7684\u8d23\u4efb\u3002<\/p>\n

GIL \u66fe\u7ecf\u65e0\u610f\u4e2d\u66ff\u5927\u91cf Python \u7a0b\u5e8f\u9690\u85cf\u4e86\u4e00\u90e8\u5206\u5e76\u53d1\u95ee\u9898\u3002<\/p>\n

\u5f53\u5b83\u79bb\u5f00\u4ee5\u540e，\u6211\u4eec\u5fc5\u987b\u91cd\u65b0\u8ba4\u771f\u5b66\u4e60：<\/p>\n

ownership
\nsynchronization
\nmemory sharing
\nresource lifecycle
\nbackpressure
\nsharding
\nobservability<\/p>\n

\u6240\u4ee5\u771f\u6b63\u7684\u95ee\u9898\u5e76\u4e0d\u662f：<\/p>\n

\u201cFree-threaded Python \u80fd\u4e0d\u80fd\u8fd0\u884c\u4e09\u4e2a asyncio loop？\u201d<\/p>\n

\u5f53\u7136\u80fd\u3002<\/p>\n

\u66f4\u503c\u5f97\u95ee\u7684\u662f：<\/p>\n

\u4f60\u7684\u7cfb\u7edf\u80fd\u5426\u88ab\u6e05\u6670\u5730\u5212\u5206\u4e3a\u4e09\u4e2a ownership domain？<\/p>\n

\u5982\u679c\u7b54\u6848\u662f\u80af\u5b9a\u7684：<\/p>\n

Thread 1 -> Loop 1 -> Resource Set 1 -> Shard 1
\nThread 2 -> Loop 2 -> Resource Set 2 -> Shard 2
\nThread 3 -> Loop 3 -> Resource Set 3 -> Shard 3<\/p>\n

\u90a3\u4e48 free-threaded Python + asyncio \u5f88\u53ef\u80fd\u6210\u4e3a\u4e00\u79cd\u76f8\u5f53\u6709\u5438\u5f15\u529b\u7684\u65b0\u67b6\u6784\u3002<\/p>\n

\u5982\u679c\u7b54\u6848\u662f：<\/p>\n

\u4e09\u4e2a\u7ebf\u7a0b
\n\u4e09\u4e2a loop
\n\u4e00\u4e2a\u5de8\u5927\u7684 mutable global state
\n\u4e00\u4e2a\u5168\u5c40 DB pool
\n\u51e0\u5341\u628a\u4e92\u76f8\u5d4c\u5957\u7684\u9501<\/p>\n

\u90a3\u4e48\u5efa\u8bae\u5c3d\u65e9\u56de\u5934\u3002<\/p>\n

\u56e0\u4e3a\u771f\u6b63\u4f18\u79c0\u7684\u5e76\u53d1\u67b6\u6784，\u4ece\u6765\u4e0d\u662f：<\/p>\n

\u8ba9\u66f4\u591a\u4e1c\u897f\u540c\u65f6\u8fd0\u884c\u3002<\/p>\n

\u800c\u662f：<\/p>\n

\u8ba9\u6bcf\u4e00\u4efd\u72b6\u6001\u7684\u5f52\u5c5e\u66f4\u52a0\u6e05\u695a\u3002<\/p>\n

\n

\u5b9e\u6218\u68c0\u67e5\u6e05\u5355<\/h3>\n

\u5728\u628a Multi-Loop Free-Threaded \u67b6\u6784\u6295\u5165\u751f\u4ea7\u4e4b\u524d，\u6211\u81f3\u5c11\u4f1a\u786e\u8ba4：\u2460 \u5f53\u524d\u89e3\u91ca\u5668\u786e\u5b9e\u662f free-threaded build，\u5e76\u68c0\u67e5\u8fd0\u884c\u65f6 GIL \u72b6\u6001；\u2461 \u6240\u6709\u5173\u952e C \u6269\u5c55\u5df2\u7ecf\u652f\u6301 free-threading；\u2462 \u6bcf\u4e2a\u5f02\u6b65\u8fde\u63a5\u6c60\u90fd\u6709\u660e\u786e\u7684 loop owner；\u2463 asyncio \u5bf9\u8c61\u7edd\u4e0d\u76f4\u63a5\u8de8\u7ebf\u7a0b\u4f7f\u7528；\u2464 mutable shared state \u8981\u4e48 shard，\u8981\u4e48\u660e\u786e\u540c\u6b65；\u2465 asyncio.Lock \u4e0e threading.Lock \u7684\u804c\u8d23\u4e25\u683c\u5206\u79bb；\u2466 \u4e3b\u7ebf\u7a0b\u7edf\u4e00\u8d1f\u8d23 SIGINT\/SIGTERM；\u2467 shutdown \u5728\u8d44\u6e90\u6240\u5c5e loop \u4e2d\u6267\u884c；\u2468 \u76d1\u63a7\u5305\u542b thread\u3001worker\u3001loop\u3001task\u3001trace \u4fe1\u606f；\u2469 \u7528 1-loop\u3001N-loop \u548c multiprocess \u505a\u771f\u5b9e workload benchmark，\u800c\u4e0d\u662f\u4f9d\u8d56\u5fae\u57fa\u51c6\u505a\u67b6\u6784\u51b3\u7b56\u3002<\/p>\n

\n

\u53c2\u8003\u8d44\u6599<\/h3>\n

Python Free-Threading \u5b98\u65b9\u6307\u5357：Python support for free threading<\/p>\n

PEP 703：Making the Global Interpreter Lock Optional in CPython<\/p>\n

PEP 779：Criteria for supported status for free-threaded Python<\/p>\n

asyncio \u591a\u7ebf\u7a0b\u5f00\u53d1\u6307\u5357：Developing with asyncio<\/p>\n

asyncio Event Loop \u6587\u6863：Python asyncio Event Loop<\/p>\n

SQLAlchemy \u591a Event Loop \u6307\u5357：SQLAlchemy AsyncIO Documentation<\/p>\n","protected":false},"excerpt":{"rendered":"

\u5f53 GIL \u4e0d\u518d\u662f\u5929\u82b1\u677f：Free-Threaded Python \u00d7 asyncio \u7684\u591a\u4e8b\u4ef6\u5faa\u73af\u67b6\u6784\u5b9e\u8df5
\n\u8fc7\u53bb\u8c08 Python \u9ad8\u5e76\u53d1，\u6211\u4eec\u5f88\u5bb9\u6613\u5f62\u6210\u4e00\u79cd\u56fa\u5b9a\u601d\u7ef4： I\/O \u5bc6\u96c6\u578b\u7528 asyncio，CPU \u5bc6\u96c6\u578b\u7528\u591a\u8fdb\u7a0b\u3002 \u8fd9\u5957\u7ecf\u9a8c\u957f\u671f\u4ee5\u6765\u6ca1\u6709\u9519，\u4f46 Free-Threaded CPython \u6b63\u5728\u6539\u53d8\u5b83\u80cc\u540e\u7684\u524d\u63d0\u3002
\nPEP 703 \u8ba9 CPython \u53ef\u4ee5\u5728\u5173\u95ed GIL \u7684\u6a21\u5f0f\u4e0b\u8fd0\u884c\u3002Pytho<\/p>\n","protected":false},"author":2,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[11653,81,190],"topic":[],"class_list":["post-106721","post","type-post","status-publish","format-standard","hentry","category-server","tag-11653","tag-python","tag-190"],"yoast_head":"\n\u5f53 GIL \u4e0d\u518d\u662f\u5929\u82b1\u677f\uff1aFree-Threaded Python \u00d7 asyncio \u7684\u591a\u4e8b\u4ef6\u5faa\u73af\u67b6\u6784\u5b9e\u8df5 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/www.wsisp.com\/helps\/106721.html\" \/>\n<meta property=\"og:locale\" content=\"zh_CN\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"\u5f53 GIL \u4e0d\u518d\u662f\u5929\u82b1\u677f\uff1aFree-Threaded Python \u00d7 asyncio \u7684\u591a\u4e8b\u4ef6\u5faa\u73af\u67b6\u6784\u5b9e\u8df5 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\" \/>\n<meta property=\"og:description\" content=\"\u5f53 GIL \u4e0d\u518d\u662f\u5929\u82b1\u677f：Free-Threaded Python \u00d7 asyncio \u7684\u591a\u4e8b\u4ef6\u5faa\u73af\u67b6\u6784\u5b9e\u8df5 \u8fc7\u53bb\u8c08 Python \u9ad8\u5e76\u53d1，\u6211\u4eec\u5f88\u5bb9\u6613\u5f62\u6210\u4e00\u79cd\u56fa\u5b9a\u601d\u7ef4： I\/O \u5bc6\u96c6\u578b\u7528 asyncio，CPU \u5bc6\u96c6\u578b\u7528\u591a\u8fdb\u7a0b\u3002 \u8fd9\u5957\u7ecf\u9a8c\u957f\u671f\u4ee5\u6765\u6ca1\u6709\u9519，\u4f46 Free-Threaded CPython \u6b63\u5728\u6539\u53d8\u5b83\u80cc\u540e\u7684\u524d\u63d0\u3002 PEP 703 \u8ba9 CPython \u53ef\u4ee5\u5728\u5173\u95ed GIL \u7684\u6a21\u5f0f\u4e0b\u8fd0\u884c\u3002Pytho\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.wsisp.com\/helps\/106721.html\" \/>\n<meta property=\"og:site_name\" content=\"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\" \/>\n<meta property=\"article:published_time\" content=\"2026-09-17T23:32:33+00:00\" \/>\n<meta name=\"author\" content=\"admin\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"\u4f5c\u8005\" \/>\n\t<meta name=\"twitter:data1\" content=\"admin\" \/>\n\t<meta name=\"twitter:label2\" content=\"\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4\" \/>\n\t<meta name=\"twitter:data2\" content=\"8 \u5206\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/106721.html\",\"url\":\"https:\/\/www.wsisp.com\/helps\/106721.html\",\"name\":\"\u5f53 GIL \u4e0d\u518d\u662f\u5929\u82b1\u677f\uff1aFree-Threaded Python \u00d7 asyncio \u7684\u591a\u4e8b\u4ef6\u5faa\u73af\u67b6\u6784\u5b9e\u8df5 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\",\"isPartOf\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/#website\"},\"datePublished\":\"2026-09-17T23:32:33+00:00\",\"dateModified\":\"2026-09-17T23:32:33+00:00\",\"author\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41\"},\"breadcrumb\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/106721.html#breadcrumb\"},\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/www.wsisp.com\/helps\/106721.html\"]}]},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/106721.html#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\u9996\u9875\",\"item\":\"https:\/\/www.wsisp.com\/helps\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"\u5f53 GIL \u4e0d\u518d\u662f\u5929\u82b1\u677f\uff1aFree-Threaded Python \u00d7 asyncio \u7684\u591a\u4e8b\u4ef6\u5faa\u73af\u67b6\u6784\u5b9e\u8df5\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#website\",\"url\":\"https:\/\/www.wsisp.com\/helps\/\",\"name\":\"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\",\"description\":\"\u9999\u6e2f\u670d\u52a1\u5668_\u9999\u6e2f\u4e91\u670d\u52a1\u5668\u8d44\u8baf_\u670d\u52a1\u5668\u5e2e\u52a9\u6587\u6863_\u670d\u52a1\u5668\u6559\u7a0b\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/www.wsisp.com\/helps\/?s={search_term_string}\"},\"query-input\":\"required name=search_term_string\"}],\"inLanguage\":\"zh-Hans\"},{\"@type\":\"Person\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41\",\"name\":\"admin\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery\",\"contentUrl\":\"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery\",\"caption\":\"admin\"},\"sameAs\":[\"http:\/\/wp.wsisp.com\"],\"url\":\"https:\/\/www.wsisp.com\/helps\/author\/admin\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"\u5f53 GIL \u4e0d\u518d\u662f\u5929\u82b1\u677f\uff1aFree-Threaded Python \u00d7 asyncio \u7684\u591a\u4e8b\u4ef6\u5faa\u73af\u67b6\u6784\u5b9e\u8df5 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/www.wsisp.com\/helps\/106721.html","og_locale":"zh_CN","og_type":"article","og_title":"\u5f53 GIL \u4e0d\u518d\u662f\u5929\u82b1\u677f\uff1aFree-Threaded Python \u00d7 asyncio \u7684\u591a\u4e8b\u4ef6\u5faa\u73af\u67b6\u6784\u5b9e\u8df5 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","og_description":"\u5f53 GIL \u4e0d\u518d\u662f\u5929\u82b1\u677f：Free-Threaded Python \u00d7 asyncio \u7684\u591a\u4e8b\u4ef6\u5faa\u73af\u67b6\u6784\u5b9e\u8df5 \u8fc7\u53bb\u8c08 Python \u9ad8\u5e76\u53d1，\u6211\u4eec\u5f88\u5bb9\u6613\u5f62\u6210\u4e00\u79cd\u56fa\u5b9a\u601d\u7ef4： I\/O \u5bc6\u96c6\u578b\u7528 asyncio，CPU \u5bc6\u96c6\u578b\u7528\u591a\u8fdb\u7a0b\u3002 \u8fd9\u5957\u7ecf\u9a8c\u957f\u671f\u4ee5\u6765\u6ca1\u6709\u9519，\u4f46 Free-Threaded CPython \u6b63\u5728\u6539\u53d8\u5b83\u80cc\u540e\u7684\u524d\u63d0\u3002 PEP 703 \u8ba9 CPython \u53ef\u4ee5\u5728\u5173\u95ed GIL \u7684\u6a21\u5f0f\u4e0b\u8fd0\u884c\u3002Pytho","og_url":"https:\/\/www.wsisp.com\/helps\/106721.html","og_site_name":"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","article_published_time":"2026-09-17T23:32:33+00:00","author":"admin","twitter_card":"summary_large_image","twitter_misc":{"\u4f5c\u8005":"admin","\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4":"8 \u5206"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/www.wsisp.com\/helps\/106721.html","url":"https:\/\/www.wsisp.com\/helps\/106721.html","name":"\u5f53 GIL \u4e0d\u518d\u662f\u5929\u82b1\u677f\uff1aFree-Threaded Python \u00d7 asyncio \u7684\u591a\u4e8b\u4ef6\u5faa\u73af\u67b6\u6784\u5b9e\u8df5 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","isPartOf":{"@id":"https:\/\/www.wsisp.com\/helps\/#website"},"datePublished":"2026-09-17T23:32:33+00:00","dateModified":"2026-09-17T23:32:33+00:00","author":{"@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41"},"breadcrumb":{"@id":"https:\/\/www.wsisp.com\/helps\/106721.html#breadcrumb"},"inLanguage":"zh-Hans","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.wsisp.com\/helps\/106721.html"]}]},{"@type":"BreadcrumbList","@id":"https:\/\/www.wsisp.com\/helps\/106721.html#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u9996\u9875","item":"https:\/\/www.wsisp.com\/helps"},{"@type":"ListItem","position":2,"name":"\u5f53 GIL \u4e0d\u518d\u662f\u5929\u82b1\u677f\uff1aFree-Threaded Python \u00d7 asyncio \u7684\u591a\u4e8b\u4ef6\u5faa\u73af\u67b6\u6784\u5b9e\u8df5"}]},{"@type":"WebSite","@id":"https:\/\/www.wsisp.com\/helps\/#website","url":"https:\/\/www.wsisp.com\/helps\/","name":"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","description":"\u9999\u6e2f\u670d\u52a1\u5668_\u9999\u6e2f\u4e91\u670d\u52a1\u5668\u8d44\u8baf_\u670d\u52a1\u5668\u5e2e\u52a9\u6587\u6863_\u670d\u52a1\u5668\u6559\u7a0b","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/www.wsisp.com\/helps\/?s={search_term_string}"},"query-input":"required name=search_term_string"}],"inLanguage":"zh-Hans"},{"@type":"Person","@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41","name":"admin","image":{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/image\/","url":"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery","contentUrl":"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery","caption":"admin"},"sameAs":["http:\/\/wp.wsisp.com"],"url":"https:\/\/www.wsisp.com\/helps\/author\/admin"}]}},"_links":{"self":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts\/106721","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/comments?post=106721"}],"version-history":[{"count":0,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts\/106721\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/media?parent=106721"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/categories?post=106721"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/tags?post=106721"},{"taxonomy":"topic","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/topic?post=106721"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}