shell写一个压测脚本

ab命令

ab是apache下面的一个性能压测工具

yum install -y httpd-tools
ab -n 1000 -c 10 http://xxxx
# -n 请求数 -c 并发数

基于ab写一个压测脚本

 #!/bin/bash
 
 requests=(1000 2000 10000)
 concurrency=(10 100)
 path="http://localhost:3000"
 
 benchmark() {
   if [ $3 -eq 0 ];then
     ab -n $1 -c $2 $path 2>&1 | awk 'BEGIN {printf "%-10s %-15s %-8s %-8s %-8s\n", "Requests", "Concurrency", "Min(ms)", "Max(ms)", "Mean(ms)";printf "------------------------------------------------------ \n"} /Total:/ {printf "%-10s %-15s %-8s %-8s %-8s\n", "'$1'", "'$2'", $2, $6, $3}'
   else
     ab -n $1 -c $2 $path 2>&1 | awk '/Total:/ {printf "%-10s %-15s %-8s %-8s %-8s\n", "'$1'", "'$2'", $2, $6, $3}'
   fi  
 }
 
 index=0
 for request in ${requests[@]}
 do
   for con in ${concurrency[@]} 
   do  
     benchmark $request $con $index 
     let 'index++'
   done
 done 

压测结果：

Requests   Concurrency     Min(ms)  Max(ms)  Mean(ms)
------------------------------------------------------
1000       10              4        23       7       
1000       100             9        52       39      
2000       10              3        63       4       
2000       100             9        42       31      
10000      10              2        409      5       
10000      100             9        1980     49    

几个细节

1. shell遍历数组 for request in ${requests[@]}
2. awk中/Total:/必须放在BEGIN{printf ''}后面
3. 管道符默认只传递标准输出，所以需要2>&1
4. awk中printf中%-8s，负号表示左对齐，8表示占8位字符，s表示字符串
5. shell中使用let进行数学计算