background

With the soaring business volume , The company's system needs high-speed iteration for business in the early stage , From the perspective of lean thinking , Rapid development of , Left a lot of historical baggage （ Or performance bad code）, Lead to frequent performance problems , This article , Mainly sort out some commonly used performance tools , Some general ideas for troubleshooting performance problems will be updated later .

Linux Common performance tools

Generally speaking , My business is deployed in linux The server , I have to consider it here , Performance troubleshooting at the operating system level .
Linux Tool set for system analysis ：
Brendan Gregg http://www.brendangregg.com/linuxperf.html

You can check the website in detail . What is mentioned in the above figure is linux All levels ： application layer 、 System level 、 Driver layer 、 The network layer , And the corresponding troubleshooting commands , I won't cover all of them here , Click on the link , Make sure you receive a lot of goods .

Java Built in performance troubleshooting tool

1. jstack Grab the current stack information , It is mainly used to view java What are the threads doing in , It is very suitable for troubleshooting the slow response of the system 、CPU High , Whether there are a series of problems such as hot lock . Everyone may only know jstack, But I suggest you also take a look jstack Some auxiliary parameters of ,jstack By default, you can only see Java Stack , and jstack -m You can see the thread Java Stack and native Stack .
   Usually we go through linux command top -H and jstack To locate some cpu Execute higher questions . Two more are recommended linux The order of
   pstack and perf pstack You can use it to see Java Process native Stack ,perf linux Performance analysis super weapon .
2. Memory related commands ：
   jmap obtain heap Information , He also has different parameters , Dealing with different scenarios such as jmap -dump and jmap -histo:live

   # Extract process memory information , Used to analyze OOM Cause 
    jmap -dump:format=b,file=xxx.hprof  <pid> 
    [ Export the whole JVM  Memory information in ]
    # Output heap information 
    jmap -heap <PID>
    # Output permanent generation information 
      jmap －permstat  <PID>   

jstat View live gc situation , However, it is recommended to check for complex problems gclog

    Generally, the running command is as follows ：
   jstat -gc 12538 5000 
    Every time 5 The process number is displayed once per second 12538 Of java Progressive GC situation ,

   jps  see java In the process of id Information 
   jhat  see dump Information , However, it is not commonly used at present , More advanced tools are described below .

java Built in performance troubleshooting tool , It is very fast in troubleshooting some obvious problems , On the other hand, it is also a quick snapshot of online system status , Convenient for further analysis with sharp tools .

Further sharp weapon

1. gc Log analysis tool ：

     GCViewer   and   gceasy.io , Or online , You can import logs directly , Visual analysis gc situation .
   

2. Heap memory analysis tool ：
MAT Also known as Eclipse Memory Analyzer

3. Thread analysis tool ：
IBM JCA Tools （IBM Thread and Monitor Dump Analyze for Java）
A permissive in java The positioning thread in the thread is suspended / Deadlock / Tools for resource competition and bottlenecks .

4. Other performance tools ：

  JProfiler： Set CPU、 Memory and thread performance analysis in one , Can be used to analyze performance bottlenecks 、 Memory leak 、CPU Load and solve thread related problems , Support local and remote direct connection JVM, It is recommended not to use during peak hours online .
  VisualVM： ditto