接口方式实现业务指标统计CDN数据

mujingluo

已于 2023-03-31 11:55:47 修改

阅读量306

点赞数 1

文章标签： python 运维开发 Powered by 金山文档

于 2023-03-13 15:05:49 首次发布

本文链接：https://blog.csdn.net/mujingluo/article/details/129494304

版权

需求

按天粒度分析阿里云所有域名的CDN日志，按文件后缀区分业务，统计流量占比、总流量、PV。

业务分类如下：

音频：

mp3、.wav等

视频：

mp4、mpg等

图片：

png、jepg等

构思

第一确认：

接口采用GET请求方式，根据URL中的日期参数，返回JSON格式的统计数据

第二确认：

阿里云CDN服务不支持自定义归类，只能利用CDN日志进行统计，通过对uri字段后缀名判断进行业务分类，对response_size字段进行汇总得出流量大小

注：response_size统计的大小不完全等于CDN流量的实际大小，业内一般差额在7%～13%，阿里云采用的是10%差额

第三确认：

原设想用GO语言，但通过学习阿里云日志服务接口文档，发现GO的文档太少，而Python的文档则很丰富，最终选择使用Python

第四确认：

最核心的问题，如何统计日志。之前使用方法是每日将日志下载到服务本地，在本地计算，果断弃掉这个方法，太重了。通过学习阿里云SDK文档，发现支持SQL，确定利用阿里云在线检索，但海量的日志检索，必然存在返回时间过长的问题，通过需求沟通，对时间没有特殊要求，只要能够返回数据就行。

示例代码：

from time import time
from aliyun.log import GetLogsRequest
request = GetLogsRequest("project1", "logstore1", fromTime=int(time()-3600), toTime=int(time()), topic='', query="*", line=100, offset=0, reverse=False)
# 或者
request = GetLogsRequest("project1", "logstore1", fromTime="2018-1-1 10:10:10", toTime="2018-1-1 10:20:10", topic='', query="*", line=100, offset=0, reverse=False)


res = client.get_logs(request)

附链接：https://aliyun-log-python-sdk.readthedocs.io/README_CN.html#get

遇到的问题：

在测试过程中，发现将检索时间大于1小时后，就会出现检索超时或者返回数据不准确的问题。原因是日志量太大并且SQL语句判断太多，虽然早有预期执行SQL会消耗不少时间，但超时和检索结果不准的情况还是有点出乎意料，与阿里云客服沟通，检索量触发了阿里云的流量限制。根据客服的建议之一，使用SQL增强功能，依然无改善。

尝试客服建议的另一个方案——Scheduled SQL，大致原理是设置定时任务定期执行SQL，并将执行结果储存到单独的日志库内。

附链接：https://help.aliyun.com/document_detail/215936.html

最终方案

创建多个Scheduled SQL作业，每隔一个小时执行SQL，结果以新的字段名储存到新的日志库中，这边代码只需汇总新的日志库的字段即可。经过测试，接口结果返回不仅仅准确，检索时间没有限制，而且时间是秒级。

完整代码

from fastapi import FastAPI
from aliyun.log import LogClient
# 定义阿里云SDK访问地址、key
endpoint = 'cn-xxxx.log.aliyuncs.com'
accessKeyId = 'xxxxxxxxxxxxxxxx'
accessKey = 'xxxxxxxxxxxxxxx'
client = LogClient(endpoint, accessKeyId, accessKey)

app = FastAPI()
# 统计总流量PV和Flow
def total (begintime,endtime):
    pv = client.get_log("kaishucdnlogintime", "cdnscheduled", begintime,endtime,query="*|select sum(total_pv) as pv")
    flow = client.get_log("kaishucdnlogintime", "cdnscheduled", begintime,endtime,query="*|select sum(total_flow) as flow")
    total_pv = int(pv.body[0]['pv'])
    total_flow = int(flow.body[0]['flow'])
    return total_pv,total_flow
# 统计image PV和Flow
def image (begintime,endtime):
    pv = client.get_log("kaishucdnlogintime", "cdnscheduled", begintime,endtime,query="*|select sum(image_pv) as pv")
    flow = client.get_log("kaishucdnlogintime", "cdnscheduled", begintime,endtime,query="*|select sum(image_flow) as flow")
    image_pv = int(pv.body[0]['pv'])
    image_flow = int(flow.body[0]['flow'])
    return image_pv,image_flow
# 统计video PV和Flow
def video (begintime,endtime):
    pv = client.get_log("kaishucdnlogintime", "cdnscheduled", begintime,endtime,query="*|select sum(video_pv) as pv")
    flow = client.get_log("kaishucdnlogintime", "cdnscheduled", begintime,endtime,query="*|select sum(video_flow) as flow")
    video_pv = int(pv.body[0]['pv'])
    video_flow = int(flow.body[0]['flow'])
    return video_pv,video_flow
# 统计audio PV和Flow
def audio (begintime,endtime):
    pv = client.get_log("kaishucdnlogintime", "cdnscheduled", begintime,endtime,query="*|select sum(audio_pv) as pv")
    flow = client.get_log("kaishucdnlogintime", "cdnscheduled", begintime,endtime,query="*|select sum(audio_flow) as flow")
    audio_pv = int(pv.body[0]['pv'])
    audio_flow = int(flow.body[0]['flow'])
    return audio_pv,audio_flow

@app.get('/aliyun/beginDate={fromtime}&endDate={totime}')
def calculate(fromtime: str=None, totime: str=None):
    # 调整时间格式
    fromtime = list(fromtime)
    totime = list(totime)
    f_time = fromtime[0] + fromtime[1] + fromtime[2] + fromtime[3] + "-" + fromtime[4] + fromtime[5] + "-" + fromtime[6] + fromtime[7] + " 0:0:0"
    t_time = totime[0] + totime[1] + totime[2] + totime[3] + "-" + totime[4] + totime[5] + "-" + totime[6] + totime[7] + " 23:59:59"
    # 获取 audio、image、video及totalpv数据
    total_pv = total(f_time,t_time)[0]
    audio_pv = audio(f_time,t_time)[0]
    image_pv = image(f_time,t_time)[0]
    video_pv = video(f_time,t_time)[0]
    other_pv = total_pv - audio_pv - video_pv
    # 获取 audio、image、video及tatal流量数据
    total_flow = round(total(f_time,t_time)[1]/1024/1024/1024,3)
    audio_flow = round(audio(f_time,t_time)[1]/1024/1024/1024,3)
    image_flow = round(image(f_time,t_time)[1]/1024/1024/1024,3)
    video_flow = round(video(f_time,t_time)[1]/1024/1024/1024,3)
    other_flow = total_flow - audio_flow - video_flow - image_flow
    # 计算 audio、image、video百分比
    audio_flow_ratio = round(audio_flow/total_flow*100,3)
    image_flow_ratio = round(image_flow/total_flow*100,3)
    video_flow_ratio = round(video_flow/total_flow*100,3)
    other_flow_ratio = round(other_flow/total_flow*100,3)
    res = {"data":{"total":[{"totalFlow":image_flow,"flowRatio":image_flow_ratio,"pv":image_pv,"type":"image"},{"totalFlow":audio_flow,"flowRatio":audio_flow_ratio,"pv":audio_pv,"type":"audio"},{"totalFlow":video_flow,"flowRatio":video_flow_ratio,"pv":video_pv,"type":"video"},{"totalFlow":other_flow,"flowRatio":other_flow_ratio,"pv":other_pv,"type":"other"}]}}
    return res

if __name__ == '__main__':
    import uvicorn
    uvicorn.run(app=app,
                host="0.0.0.0",
                port=8080,
                workers=1)

仓库地址：https://github.com/huoji1990/aliyun-cdn