linux grpc测试基于ghz工具

1.测试前提，参考先熟悉grpc测试demo:

参考官网：python 实现grpc client以及service :

https://grpc.io/docs/quickstart/python/

java实现client 和service:https://grpc.io/docs/quickstart/java/

细节我就不讲了：

主要说下.proto文件，他是实现grpc的基础，根据他产生契约helloworld_pb2，以及客户端服务端调用类：helloworld_pb2_grpc（核心调用client编写就靠他的Stub 类）：

生成契约和grpc client和 service 类

$ python -m grpc_tools.protoc   -I     ../../protos    --python_out=.      --grpc_python_out=.      ../../protos/helloworld.proto

依赖文件：

grpcio==1.28.1
grpcio-tools==1.28.1
protobuf==3.11.3
coverage>=4.0
cython>=0.29.8
enum34>=1.0.4
#protobuf>=3.5.0.post1
six>=1.10
wheel>=0.29
setuptools==46.1.3

2.linux 环境搭建安装python3：

虚拟环境：基于virtualenv 

安装依赖

3.安装ghz工具，安装go：

安装go环境linux教程：https://www.runoob.com/go/go-environment.html

ghza安装包下载：https://github.com/bojand/ghz/releases 

wget  https://github.com/bojand/ghz/releases/download/v0.52.0/ghz_0.52.0_Linux_x86_64.tar.gz

ghz 帮助文档：https://ghz.sh/docs/options

配置ghz和go环境变量vi/etc/profile：

export PATH=$PATH:/usr/local/go/bin
export PATH=$PATH:/project

source /etc/profile:

验证golang ,ghz 环境：

go version 

ghz -h 

4.启动grpc server:

 编写一个简单的服务端grpc_server和对应的client 先mock下server可用，然后进行ghz工具压测server ：

[root@vm_master helloword]# cat grpc_client.py 
import logging
import time
import grpc
import helloworld_pb2
import helloworld_pb2_grpc
import json

def run():
    with grpc.insecure_channel('localhost:50051') as channel:
        stub = helloworld_pb2_grpc.GreeterStub(channel)
        req={"name":"jack","id":1001}
　　　　 # transfer json  object to bytes to send grpc request 
        body=json.dumps(req).encode("utf-8")
        response = stub.SayHello(helloworld_pb2.HelloRequest(name=body))
        # print(type(response.message))
    print(response.message)



if __name__ == '__main__':
    logging.basicConfig()
    run()

　grpc_server.py:

from concurrent import futures
import logging

import grpc

import helloworld_pb2
import helloworld_pb2_grpc


class Greeter(helloworld_pb2_grpc.GreeterServicer):

    def SayHello(self, request, context):
        return helloworld_pb2.HelloReply(message= request.name)


def serve():
    server = grpc.server(futures.ThreadPoolExecutor(max_workers=10))
    helloworld_pb2_grpc.add_GreeterServicer_to_server(Greeter(), server)
    server.add_insecure_port('[::]:50051')
    server.start()
    server.wait_for_termination()


if __name__ == '__main__':
    logging.basicConfig()
    serve()

　启动服务端 python grpc_server.py,在运行client测试，查看结果：

C:Python37python.exe D:/workspace/demos/python/helloword/grpc_client.py
{"name": "jack", "id": 1001}

Process finished with exit code 0

，接着进行压测：

前面已经安装好了ghz ，接着进行压力测试：

鉴于ghz 命令基于linux所以打算借助脚本调用 ghz  压力测试命令，以及请求参数构造：

脚本：

import  subprocess
import json

def exe_cmd(args):


    sub=subprocess.Popen(args,shell=False,stdout=subprocess.PIPE,stderr=subprocess.STDOUT)
    out,err=sub.communicate()

    return out.decode("utf-8")

def demo():
    bytes_data={"name":"zhangsan","id":"10001"}
    bytes_data=json.dumps(bytes_data).encode("utf-8")
    args=["ghz","--skipTLS","--insecure","--proto","../../protos/helloworld.proto",
          "--call","helloworld.Greeter.SayHello","--metadata",bytes_data,"-c","5","--qps","100","-z","1m","--connections","5","-t","20s","0.0.0.0:50051"]
    print(exe_cmd(args))

if __name__ == '__main__':
    demo()

这里输出默认是命令行展示结果：

[root@vm_master helloword]# python deso.py

Summary:
Count: 29761
Total: 60.01 s
Slowest: 71.66 ms
Fastest: 0.52 ms
Average: 3.63 ms
Requests/sec: 495.95

Response time histogram:
0.515 [1] |
7.630 [29219] |∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎
14.744 [384] |∎
21.859 [133] |
28.973 [8] |
36.088 [2] |
43.202 [0] |
50.317 [3] |
57.431 [2] |
64.546 [4] |
71.660 [1] |

Latency distribution:
10 % in 1.67 ms
25 % in 2.03 ms
50 % in 3.36 ms
75 % in 4.82 ms
90 % in 5.76 ms
95 % in 6.37 ms
99 % in 12.27 ms

Status code distribution:
[OK] 29757 responses
[Unavailable] 2 responses
[Canceled] 2 responses

Error distribution:
[2] rpc error: code = Unavailable desc = transport is closing
[2] rpc error: code = Canceled desc = grpc: the client connection is closing

 为了方便阅读，可以保存为html 只需添加参数 -o ./result.html -O html既可

args=["ghz","--skipTLS","--insecure","--proto","../../protos/helloworld.proto",
"--call","helloworld.Greeter.SayHello","--metadata",bytes_data,"-o","./result.html","-O","html","-c","5","-n","10","0.0.0.0:50051"]

html参考效果：https://ghz.sh/sample.html

output指标解释：https://ghz.sh/docs/output

注意事项：

def demo():
bytes_data={"name":"zhangsan","id":"10001"}
bytes_data=json.dumps(bytes_data).encode("utf-8")
args=["ghz","--skipTLS","--insecure","--proto","../../protos/helloworld.proto",
"--call","helloworld.Greeter.SayHello","--metadata",bytes_data,"-c","5","--qps","100","-z","1m","--connections","5","-t","20s","0.0.0.0:50051"]
print(exe_cmd(args))

这里bytes_data=里整数按字符串填写，否则go 转换出错，其次 一般 -d {"key":"value"}需要在proto文件定义type,这里为了通用传输字节，我直接按metadata元数据传输他会自动处理

 技术群答疑群：