jmeter 脚本常用配置方法

  • Job name \({__jexl3("gpu-test"+"_"+"\){__threadNum}”)}

  • 随机数 ${__Random(1,1000)}

\({__jexl3(200+\){__threadNum})}

  • csv 文件读取

   resources_folder: ${__P(resources, ../resources)}
   ${resources_folder}/accounts.csv
   utf-8
   username,pureKey,passwd
   ,   # 分隔符

  • 设置全局变量 \({__setProperty(auth, \){token},)}

  • 子控制过程引用全局变量 Cookie： token=${__P(auth,)}

  • 当前线程ID ${__threadNum}

  • 提取返回值 Json 中的id array userIdArray $.[*].id -1

    • 可直接批量引用 [\({rolePageIdArray_1},\){rolePageIdArray_2},\({rolePageIdArray_3},\){rolePageIdArray_4},${rolePageIdArray_5},] userIdArray rolePageIdArray_1=1 rolePageIdArray_10=69 rolePageIdArray_2=2 rolePageIdArray_3=62 rolePageIdArray_4=63 rolePageIdArray_5=64 rolePageIdArray_6=65 rolePageIdArray_7=66 rolePageIdArray_8=67 rolePageIdArray_9=68

  • foreach controllor 逻辑块参数设置

MyJobIds iJobId

  • if controllor 逻辑块参数设置 \({__jexl3("\){jobStatus}” != “killed”)}

  • Loop controller 中的index JMeter will expose the looping index as a variable named jm__idx. So for example, if your Loop Controller is named LC, then you can access the looping index through ${__jm__LC__idx}. Index starts at 0

示例： Loop controller 重命名为：LoopSubmit 多用户批量执行Job name: \({__jexl3("cpu_test"+"_"+"\){username}”+”_”+”${__jm__LoopSubmit__idx}”)}

  • CPU job

    • cmd: sleep 10m

  1. 单机

{"userName":"${username}","jobType":"training","gpuType":"nvidia","vcName":"platform","containerUserId":0,"jobName":"${__jexl3("cpu_test"+"_"+"${username}"+"_"+"${__jm__LoopSubmit__idx}")}","jobtrainingtype":"RegularJob","preemptionAllowed":"${preemptionAllowed}","image":"apulistech/mindspore:0.2.0","cmd":"sleep ${jobSleep}m","workPath":"./","enableworkpath":true,"dataPath":"./","enabledatapath":true,"jobPath":"./","enablejobpath":true,"env":[],"hostNetwork":false,"isPrivileged":false,"interactivePorts":"40000,45000","plugins":{"blobfuse":[{"accountName":"","accountKey":"","containerName":"","mountPath":"","mountOptions":""}],"imagePull":[{"registry":"","username":"","password":""}]},"resourcegpu":0}

  1. 分布式

{"userName":"${username}","jobType":"training","gpuType":"nvidia","vcName":"baseline","containerUserId":0,"jobName":"${__jexl3("cpu_distirbuted_test"+"_"+"${username}"+"_"+"${__jm__LoopSubmit__idx}")}","jobtrainingtype":"RegularJob","preemptionAllowed":"${preemptionAllowed}","image":"ubuntu:16.04","cmd":"sleep ${jobSleep}m","workPath":"./","enableworkpath":true,"dataPath":"./","enabledatapath":true,"jobPath":"./","enablejobpath":true,"env":[],"hostNetwork":false,"isPrivileged":false,"interactivePorts":"","plugins":{"blobfuse":[{"accountName":"","accountKey":"","containerName":"","mountPath":"","mountOptions":""}],"imagePull":[{"registry":"","username":"","password":""}]},"resourcegpu":0}

  • GPU job

    • cmd: sudo bash -c "source /root/anaconda3/bin/deactivate && conda activate open-mmlab && cd /root/pytorch_samples/mnist/ && python main.py --epochs 10002"

  1. 单机

{"userName":"${username}","jobType":"training","gpuType":"nvidia","vcName":"platform","containerUserId":0,"jobName":"${__jexl3("gpu_test"+"_"+"${username}"+"_"+"${__jm__LoopSubmit__idx}")}","jobtrainingtype":"RegularJob","preemptionAllowed":"${preemptionAllowed}","image":"apulistech/cuda:10.0","cmd":"sudo bash -c \"source /root/anaconda3/bin/deactivate && conda activate open-mmlab && cd /root/pytorch_samples/mnist/ && python main.py --epochs ${jobSleep}\"","workPath":"./","enableworkpath":true,"dataPath":"./","enabledatapath":true,"jobPath":"./","enablejobpath":true,"env":[],"hostNetwork":false,"isPrivileged":false,"interactivePorts":"40000,45000","plugins":{"blobfuse":[{"accountName":"","accountKey":"","containerName":"","mountPath":"","mountOptions":""}],"imagePull":[{"registry":"","username":"","password":""}]},"resourcegpu":${resourcegpu}}

  1. 分布式

{"userName":"${username}","jobType":"training","gpuType":"${gpuType}","vcName":"${vcName}","containerUserId":0,"jobName":"${__jexl3("gpu_distirbuted_test"+"_"+"${username}"+"_"+"${__jm__LoopSubmit__idx}")}","jobtrainingtype":"PSDistJob","preemptionAllowed":"${preemptionAllowed}","image":"apulistech/cuda:10.0","cmd":"sudo bash -c \"source /root/anaconda3/bin/deactivate && conda activate open-mmlab && cd /root/pytorch_samples/mnist/ && python main.py --epochs ${jobSleep}\"","workPath":"./","enableworkpath":true,"dataPath":"./","enabledatapath":true,"jobPath":"./","enablejobpath":true,"env":[],"hostNetwork":true,"isPrivileged":true,"interactivePorts":"","plugins":{"blobfuse":[{"accountName":"","accountKey":"","containerName":"","mountPath":"","mountOptions":""}],"imagePull":[{"registry":"","username":"","password":""}]},"numps":1,"resourcegpu":${resourcegpu},"numpsworker":${numpsworker}}

{“userName”:”sample0”,”jobType”:”training”,”gpuType”:”nvidia_gpu_amd64”,”vcName”:”atlas01”,”containerUserId”:0,”jobName”:”test_gpu_2”,”jobtrainingtype”:”PSDistJob”,”preemptionAllowed”:”False”,”image”:”apulistech/cuda:10.0”,”cmd”:”sudo bash -c “source /root/anaconda3/bin/deactivate && conda activate open-mmlab && python /data/code/main.py –epochs 10002””,”workPath”:”./”,”enableworkpath”:true,”dataPath”:”./”,”enabledatapath”:true,”jobPath”:”./”,”enablejobpath”:true,”env”:[],”hostNetwork”:true,”isPrivileged”:true,”interactivePorts”:””,”plugins”:{“blobfuse”:[{“accountName”:””,”accountKey”:””,”containerName”:””,”mountPath”:””,”mountOptions”:””}],”imagePull”:[{“registry”:””,”username”:””,”password”:””}]},”numps”:1,”resourcegpu”:8,”numpsworker”:2}

  • 设置 endpoint {“endpoints”:[{“name”:”port-40000”,”podPort”:40000},{“name”:”port-45000”,”podPort”:45000},”ssh”,”ipython”]}

https://china-gpu02.sigsus.cn/sign-in

https://china-gpu02.sigsus.cn/api/v2/clusters/china-gpu02/teams/platform/jobs?limit=999 /api/v2/clusters/${host}/teams/baseline/jobs?limit=999

  • 创建VC 1.VC_Name \({__jexl3("VC"+"_"+"\){VCType}”+”_”+”${__threadNum}”}

//sandbox2-master.sigsus.cn:52080/api/sandbox02-gpu01/addVc/\({__jexl3("VC"+"_"+"\){VCType}”+”_”+”\({__threadNum}"}/%7B%22\){VCType}%22:${VCNu

  • 随机取Array 元素

${__RandomFromMultipleVars(userIdArray)}