DRF 节流行为极其不一致？

Question

我一直在我的 DRF API 中使用非常基本的用户节流率，当我尝试运行基准测试时，我得到的结果只是……荒谬的结果。

例如，当我将限制设置为每天 15 次时，我的 python 基准测试程序能够在实际被限制之前发出 29 (???) 请求。

这是我的基本设置：

# Django Rest Framework
REST_FRAMEWORK = { 
...
    'DEFAULT_THROTTLE_CLASSES': [
        'rest_framework.throttling.UserRateThrottle',
    ],
    'DEFAULT_THROTTLE_RATES': {
        'user': '15/day'
    }
}

这是我的基准测试程序：

#!/usr/bin/env python3
# -*- coding: iso-8859-15 -*-
import requests, time, datetime
from pprint import pprint
TOKEN = [redacted]
CSDB_NEXT_URL = "http://localhost:8000/api/siteroles?format=json"

response = requests.get(CSDB_NEXT_URL, headers={'Authorization': 'Token {}'.format(TOKEN)})

startTime = datetime.datetime.now()
for i in range(0, 52):
    data = requests.get(CSDB_NEXT_URL, headers={'Authorization': 'Token {}'.format(TOKEN)}).json()
    if "Request was throttled" in str(data):
        break
endTime = datetime.datetime.now()
print(data, '\n', i, " requests made in ", endTime-startTime, " seconds.").

产生了

[Running] /usr/bin/env python3 "/Users/jgossett/Documents/gateway3/utils/benchmarks/test"
{'detail': 'Request was throttled. Expected available in 86389 seconds.'}
29 requests made in 0:00:12.437853 seconds.

[Done] exited with code=0 in 13.248 seconds

现在这里的数字29特别麻烦，因为它大于15？值得注意的是？有谁知道为什么会这样？

Answer 1

我相信我找到了不一致的原因：我的 Django 服务器正在使用默认的 LocMemCache，其 multi-threaded 缓存系统导致查询记录不一致 as was documented here

DRF 节流行为极其不一致？

DRF Throttling behaving extremely inconsistently?

throttling

django-authentication

django-rest-framework