The JVM analysis tool that comes with the JDK

Memory overflow, proper famous scene;

1. Business Background

For file services in a distributed architecture, due to the large number of IO stream operations involved, it is easy to cause JVM-related exceptions, especially the problem of memory overflow;

In the latest version iteration, the real business processing scenario is almost the same as the above. Due to the addition of batch processing actions in the file service, the bug that has been hidden for a long time is directly awakened, which is the most common memory overflow;

The cause of the problem: After the word document completes the content recognition, it is converted into a pdf file, and then the page is divided into a set of pictures. In this complex and long process, there is an array container that is not destroyed;

Solution: analyze the dump file of the JVM, locate the root cause of the OOM problem, combine the abnormal log analysis of the file service, and add the release action of the resource to solve the problem;

2. Jdk-Bin directory

For a considerable number of novices, seeing the problems of the JVM are all ignorant bugs. In fact, this kind of mentality is not necessary. From the development experience of several years in the workplace, the problems of the JVM can be roughly divided into two types:

• Easy solution for development: You can upgrade memory resources or adjust allocation, or optimize programs to manage and release related resources, which is the most commonly used method;
• Easy solution to development: Due to lack of experience, major bugs in the program lead to JVM exceptions, which in turn cause a series of chain reactions. This kind of rebound will not be Jedi, but only chicken feathers;

When solving conventional JVM exceptions, you usually rely on the basic tools in the JDK to complete the problem location, so as to analyze and solve the problem, but these require skilled use of the basic tools, and many JDK's own capabilities are often ignored. ;

In the bin directory of jdk, there are many built-in tools that can be used to analyze the JVM;

The above is a directory based on jdk1.8 , and there are many commands that are often used in development. Let's focus on the startup and operation of a microservice, and let's take a look at the usage of JVM tools based on JDK;

3. Command Line Tools

1. jps command

jps : virtual machine process status tool, this command is often used in Java environment deployment and service startup viewing. First, start a facade microservice locally, and then execute the query on the command line;

• jps: The default output of the command is the process ID and the name of the main class of the application;
• -l: Output the process ID and the full path of the application's main class;
• -v：输出向jvm传递的参数，此处展示为idea中显式配置的VM-options参数，其他内容自行查看即可；
• -m：输出向main方法传递的参数，服务启动前可以在idea的Program-arguments配置；

$ jps
1281 FacadeApp

$ jps -l
1281 com.explore.facade.FacadeApp

$ jps -v
1281 FacadeApp -Xms128m -Xmx256m -XX:MaxNewSize=256m -XX:MaxPermSize=256m

$ jps -m
1281 FacadeApp hello,main-method
复制代码

2、jinfo命令

jinfo：在命令后面带pid进程号，可以输出指定进程的配置信息，在应用启动时通常不会指定过多的配置参数，就可以使用该命令查询很多参数的默认值；该命令还可以在运行时动态调整部分参数，只是很少被使用；

$ jinfo 1281            # 只粘贴个别参数
Java System Properties: # 系统参数
    java.runtime.version=1.8.0_144-b01
    file.encoding=UTF-8
    sun.java.command=com.explore.facade.FacadeApp hello,main-method
    
VM Flags:               # 虚拟机参数
    -XX:InitialHeapSize=134217728 -XX:MaxHeapSize=268435456 -XX:MaxNewSize=267911168
    
VM Arguments:           # 运行时参数
    jvm_args: -Xms128m -Xmx256m -XX:MaxNewSize=256m -XX:MaxPermSize=256m
    java_command: com.explore.facade.FacadeApp hello,main-method

$ jinfo -sysprops 1281  # 只输出【Java System Properties】参数

$ jinfo -flags 1281     # 只输出【VM Flags】参数
复制代码

3、jstat命令

jstat：以指定的频率输出JVM的监控指标，下述命令输出内存占用和GC相关信息，每隔3秒输出一次，连续打印5次；由于这里只是启动一个简单的微服务，没有执行业务逻辑，所以各项指标比较平稳；

$ jstat -gcutil 1281 3000 5
S0     S1     E      O      M     CCS    YGC     YGCT    FGC    FGCT    CGC    CGCT     GCT   
0.00   0.00  57.97  64.16  92.82  88.75    3    0.028     9    0.516     -        -    0.544
0.00   0.00  57.97  64.16  92.82  88.75    3    0.028     9    0.516     -        -    0.544
复制代码

该命令是比较常用的，这里各项指标的统计逻辑，在tools.jar包中有jstat_options参考文档，相对路径sun/tools/jstat/resources/目录下；

option gcutil {
  column {
    header "^S0^"	/* Survivor 0 Space - Percent Used */
    data (1-((sun.gc.generation.0.space.1.capacity - sun.gc.generation.0.space.1.used)/sun.gc.generation.0.space.1.capacity)) * 100
  }
  column {
    header "^S1^"	/* Survivor 1 Space - Percent Used */
    data (1-((sun.gc.generation.0.space.2.capacity - sun.gc.generation.0.space.2.used)/sun.gc.generation.0.space.2.capacity)) * 100
  }
  ......
}
复制代码

4、jstack命令

jstack：输出指定进程当前时刻在JVM中的线程信息，为了清楚的展示其效果，在服务启动时创建线程死锁，然后通过该命令就会把发生死锁的线程打印出来，通过输出可以发现两条互相等待的线程信息；

$ jstack 1281
Found one Java-level deadlock:
=============================
"test-thread-02":
  waiting for ownable synchronizer 0x00000007b00a35d0, (a java.util.concurrent.locks.ReentrantLock$NonfairSync),
  which is held by "test-thread-01"
"test-thread-01":
  waiting for ownable synchronizer 0x00000007b00a35a0, (a java.util.concurrent.locks.ReentrantLock$NonfairSync),
  which is held by "test-thread-02"

Java stack information for the threads listed above:
===================================================
"test-thread-02":
    at sun.misc.Unsafe.park(Native Method)
    - parking to wait for  <0x00000007b00a35d0> (a java.util.concurrent.locks.ReentrantLock$NonfairSync)

"test-thread-01":
    at sun.misc.Unsafe.park(Native Method)
    - parking to wait for  <0x00000007b00a35a0> (a java.util.concurrent.locks.ReentrantLock$NonfairSync)

Found 1 deadlock.
复制代码

5、jmap命令

jmap：可以输出指定进程的内存中对象映射信息，或者堆的关键信息、内存的使用统计、GC算法、配置、类的实例信息及内存占用等，该命令在解决JVM问题时也经常使用；

$ jmap 1281

$ jmap -heap 1281
Heap Configuration:
   MinHeapFreeRatio         = 0
   MaxHeapFreeRatio         = 100
   MaxHeapSize              = 536870912 (512.0MB)
   
Heap Usage:
    PS Young Generation
    Eden Space:
    From Space:
    To Space:
    PS Old Generation

$ jmap -histo:live 1281
 num     #instances         #bytes  class name
----------------------------------------------
1311:             1             32  com.explore.facade.FacadeApp$$EnhancerBySpringCGLIB$$313d9e3
复制代码

四、可视化工具

1、jconsole

Java内置的JVM性能监控工具，在熟悉上述的命令行工具之后，对于该可视化工具的使用不会太陌生，在命令中可以查看到的默认参数或者应用自定义配置，在该工具中也可以找到，并且以图形化的方式呈现；

$ jconsole  # 通过该命令会唤起jconsole界面
复制代码

这里选择线程一栏，可以直观的看到线程数量的变化曲线，也可以在下方查看某个线程的具体信息，并且可以通过检测死锁功能，发现在服务中创建的test-thread-01和test-thread-02两条线程；

2、visualvm

VisualVM作为解决JVM问题的常用工具，集成的功能丰富且强大，此处通过Idea插件的方式启动FacadeApp微服务，在进程主页可以看到自定义的配置，线程一栏因为检测到死锁直接给到异常提示；

在监视一栏可以通过【堆dump】查看详细的信息，可以查看类的实例数和大小，并且完成了结果排序和占用统计；此处信息在定位和解决JVM问题时非常重要；

对于JVM的监控工具来说，其能力与常用的命令行语法差异很小，并且这些命令在jdk中tools.jar包也可以找到其对应的类，对于一些更高级的监控平台来说，例如Kuboard、Skywalking等，也对这些底层能力做了集成，其原理应该也是大同小异。

五、参考源码

编程文档：
https://gitee.com/cicadasmile/butte-java-note

应用仓库：
https://gitee.com/cicadasmile/butte-flyer-parent
复制代码