Thanks for your report. Change ML to hdfs-dev. Please first to choose common-dev, hdfs-dev or yarn-dev to discuss the issues. This ML is for security issues. Thanks.
Best Regards, - He Xiaoqiao On Sun, Nov 10, 2024 at 2:26 AM 徐子康 <xuzik...@stu.xmu.edu.cn> wrote: > Hello! > > I am currently conducting a study to detect potential performance failures > by comparing the performance metrics of two adjacent Hadoop versions under > the same load. > > I deployed pseudo distributed HDFS on a single node and executed the same > load multiple times in each version. The data collected from each test was > used for hypothesis testing. > > I have noticed that some metrics seem to be inferior in the new version > compared to the old version, such as > Hadoop_DataNode_GcTimeMillis、Hadoop_DataNode_ThreadsTimedWaiting > and Hadoop_NameNode_GcTotalExtraSleepTime in version 3.3.6. > > In version 3.4.0, there are also suspicious metrics such as > Hadoop_DataNode_MemHeapUsedM、Hadoop_DataNode_MemNonHeapCommittedM、Hadoop_NameNode_ElapsedTime、Hadoop_NameNode_LoadingFsImageElapsedTime > and so on. > > I have attached the box plots of these indicators, and I would like to > know if there are any possible performance issues or if it is simply an > error caused by personal research methods. > > Best wishes! > > --------------------------------------------------------------------- > To unsubscribe, e-mail: security-unsubscr...@hadoop.apache.org > For additional commands, e-mail: security-h...@hadoop.apache.org
