循环调用Bybit API获取K线时遭遇429错误（Http状态码非200）

我用以下代码从Bybit批量导出K线数据：

for i in range(len(pairName)):
    coinKline = session.query_kline(symbol=pairName[i], interval=str(sinceTime), from_time=since)['result']
    time.sleep(0.05)

原本运行正常，但批量获取多个交易对时频繁抛出429错误，错误触发的循环位置不固定，调整time.sleep时长（甚至到1秒）也无法解决。错误信息如下：

Traceback (most recent call last):
  File "c:\Users\XXXXXXXX\OneDrive\algoTrading\coinbaseData\bybit\multiple_coin_check.py", line 95, in <module>
    coinKline = session.query_kline(symbol=pairName[i], interval=str(sinceTime), from_time=since)['result']
                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\XXXXXXXX\AppData\Local\Programs\Python\Python311\Lib\pybit\usdt_perpetual.py", line 30, in query_kline
    return self._submit_request(
           ^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\XXXXXXXX\AppData\Local\Programs\Python\Python311\Lib\pybit\_http_manager.py", line 307, in _submit_request
    raise FailedRequestError(
pybit.exceptions.FailedRequestError: Http status code is not 200. (ErrCode: 429) (ErrTime: 14:10:00).
Request → GET https://api.bybit.com/public/linear/kline: {'symbol': 'BAKEUSDT', 'interval': '1', 'from': 1705402146}.

可行的解决方法 #

1. 实现指数退避重试机制 #

固定sleep时长应对限流不够灵活，指数退避能在触发限流后逐步延长等待时间，避免持续触发限流。手动实现示例：

import time
from pybit.exceptions import FailedRequestError

for i in range(len(pairName)):
    retries = 3
    delay = 0.5  # 初始延迟
    success = False
    while retries > 0 and not success:
        try:
            coinKline = session.query_kline(symbol=pairName[i], interval=str(sinceTime), from_time=since)['result']
            success = True
        except FailedRequestError as e:
            if "ErrCode: 429" in str(e):
                retries -= 1
                time.sleep(delay)
                delay *= 2  # 每次重试延迟翻倍
            else:
                raise e
    if not success:
        raise Exception(f"获取{pairName[i]}数据重试耗尽")
    time.sleep(0.1)  # 基础请求间隔

2. 严格遵守Bybit API限流规则 #

Bybit公开API的线性合约K线接口（public/linear/kline）限流为：每秒最多10次请求，每分钟最多600次请求。可以用计数器控制请求频率：

import time
from collections import deque
from pybit.exceptions import FailedRequestError

# 存储最近请求的时间戳，控制频率
request_timestamps = deque(maxlen=600)

for i in range(len(pairName)):
    now = time.time()
    
    # 控制每分钟请求数不超过600
    while request_timestamps and now - request_timestamps[0] < 60:
        time.sleep(0.1)
        now = time.time()
    
    # 控制每秒请求数不超过10
    if len(request_timestamps) >= 10 and now - request_timestamps[-10] < 1:
        time.sleep(0.1)
    
    try:
        coinKline = session.query_kline(symbol=pairName[i], interval=str(sinceTime), from_time=since)['result']
        request_timestamps.append(time.time())
    except FailedRequestError as e:
        if "ErrCode: 429" in str(e):
            time.sleep(1)
            # 重试一次
            coinKline = session.query_kline(symbol=pairName[i], interval=str(sinceTime), from_time=since)['result']
            request_timestamps.append(time.time())
        else:
            raise e

3. 配置pybit自动重试 #

pybit库支持针对特定错误码的自动重试，初始化session时可以配置：

from pybit.usdt_perpetual import HTTP

session = HTTP(
    endpoint="https://api.bybit.com",
    retry_codes=[429],  # 针对429错误重试
    retry_times=3,      # 重试次数
    retry_delay=0.5     # 初始重试延迟
)

4. 避免重复请求相同数据 #

如果脚本重复请求同一时间段的K线数据，可能会被判定为恶意请求，加剧限流。确保from_time参数每次请求都是正确的，不要重复获取已拉取过的数据。

内容的提问来源于stack exchange，提问作者mmm bbb