Process when do performance test by load runner in linux system for website

Secne:

The process when do performance testing by load runner in linux system for website

Process:

1. 同步所有服务器的时间，包括Apache/Tomcat/Database servers　　
2. 开启监测所有服务器 CPU/Memory/Disk/Network/Process queue
3. Using Load Runner to do the test
4. Get all "CPU/Memory/Disk/Network/Process queue" datas from step 2
5. Move datas from servers to local
6. Draw "CPU/Memory/Disk/Network/Process queue" charts by python according to the datas
7. Analysis the final result

Commands:

date

(sudo) ntpdate ntp1.aliyun.com

sar -o /tmp/load20170306 3 >/dev/null 2>&1 &

export LC_TIME=*POSIX*

sar -u -f /tmp/load20170306 >/tmp/load20170306_cpu.txt

sar -r -f /tmp/load20170306 >/tmp/load20170306_memory.txt

sar -q -f /tmp/load20170306 >/tmp/load20170306_queue.txt

sar -n DEV -f /tmp/load20170306 >/tmp/load20170306_network.txt

sar -b -f /tmp/load20170306 >/tmp/load20170306_disk.txt

sz /tmp/load20170306.*.txt

kill %1

service tomcat restart

service httpd restart

 

Experience about load testing for Allocaiton DT GA5.5.14.0

Process

1. Meeting with QA members to decide(time, resources, notice, scenarios for load testing, goal for the load testing)
2. Using Load Runner Vugen to record script
3. Edit script(transaction, 同步点, think time, parametrize)
4. Make sure all testing users can pass the script
5. Start to build scenario in the Load Runner Controller
6. Set in the Controller: run time setting, step download timeout, synchronous, result dir
7. Set Site Scope to choose servers and monitors
8. Connect site scope with load runner controller
9. If the load testing is in linux servers, need to monitor each servers' CPU, Memory, Disk, Network, IO by sar commands
10. Start to run all scenarios
11. Analysis the final results to know the maximum limit on concurrent users. And find out the bottle-necks.

Lessons:

1. The goal is not clear in my final report. The goals are:

   • To find out the maxmium limit concurrent users in Daily Task our system supported
   • To track the bottle-necks if concurrent users number is up to maxmium

2. Analysis charts are not very clear

   • This could follow "Z:QAQA MaterialAllocationLoad Testing20160122"

3. Not quite understand the hardware resources like cpu, memory... And cannot response to QA if they have any questions.

   • memory full does mean 性能瓶颈 necessarily

4. Resources setting in Site Scope are not quite understand.

5. 提前get date, edit cells, rows, 确认没有被release等

6. If DB memored is 64G and used 95%. It's ok.

7. Why used memory didn't release after test done. (Cann't release at a short time)

Secret to success:

Just practice more~