JVM Basics - > how to troubleshoot JVM problems in your project

How to check your project JVM problem

For those who are still in normal operation ⾏ The system of ：

• Through the results of each command , perhaps jvisualvm etc. ⼯ Come and go ⾏ analysis

  • You can make ⽤jmap Check it out. JVM The of each area in the ⽤ situation

  • Can pass jstack To view the operation of the thread ⾏ situation ,⽐ Such as which threads block 、 Is there a deadlock

  • Can pass jstat Command to view garbage collection , especially fullgc, If you find that fullgc⽐ More frequently , Then you have to go into ⾏ Tuned

• ⾸ First , Initial speculation is frequent fullgc( In the old days, large-scale garbage collection was carried out ) Why

• If you send it frequently ⽣fullgc however ⼜⼀ straight No memory overflow occurred , The system is running well

• So expressed gc It's actually You can recycle many objects at once , So it's best that these objects can younggc( A new generation of garbage collection ) In the process, it is recycled directly , Avoid these objects from entering ⼊ To ⽼ years , Even in the permanent zone

• In this case , It's time to see why

  • Consider these Survival time is not ⻓ Is the object of ⽐ a ⼤, The young generation cannot let go , Direct entry ⼊ here we are ⽼ years , Try to add ⼤ Young generation ⼤ Small
  • Check which thread is using too much memory
  • Check which thread accounts for ⽤CPU That's too much , Positioning to specific ⽅ Law , Optimize this ⽅ The enforcement of law ⾏, See if you can avoid the creation of some objects , from ⽽ Save memory

For those who have sent ⽣ 了 OOM( out of memory ) The system of ：

1. ⼀ like ⽣ The production system will set when the system sends ⽣ 了 OOM when ,⽣ Become the... Of the time dump⽂ Pieces of

   • - XX:+HeapDumpOnOutOfMemoryError -XX:HeapDumpPath=/usr/local/base

2. We can benefit from ⽤jsisualvm( visualization ) etc. ⼯ To analyze dump⽂ Pieces of

3. according to dump⽂ An exception instance object was found , And exception threads （ Occupy ⽤CPU⾼）, Go to the specific code

4. Then go in again. ⾏ Detailed analysis and commissioning

JVM View details graphically

• The graphical interface can go to jdk Of bin Find below jvisualvm.exe -> This is a built-in graphical interface
• There are some easy-to-use plug-ins you can try to install Visual GCbtrace

View threads

View the memory and used by each thread cpu The situation of

Check garbage collection

see dump Tracking information

JVM Command view details

jps -> Java ps: See what's running Java process

• jps: You can list running Java process , And display the name of the virtual machine execution main class and the process id

  C:\>jps
  5932  test Collection
  

• Common options : jps -ljps -v

  • jps -l -> Output the full class name of the main class , If the process is executing Jar package , Output jar The name of the bag

    C:\>jps -l
    5932 rod. aggregate . test Collection
    

  • jps -v -> Specified when the program starts jvm Parameters

    5932  test Collection 
        -agentlib:jdwp=transport=dt_socket,
    	address=127.0.0.1:54731,
    	suspend=y,
    	server=n 
        -javaagent:C:\Users\ Ouhuang xiaodezi \AppData\Local\JetBrains\IntelliJIdea2021.2\captureAgent\debugger-agent.jar 
        -Dfile.encoding=UTF-8
    

jstack -> Java stack: Print thread snapshot

• View a certain Java In progress Status of all threads .

• Generally used The cause of the long pause in the positioning thread , Such as There is a cycle of life and death , Deadlock , Request external resources to wait for a long time etc. ！

• Common options : jstack process idjps -v

  • jstack process id -> The state of all threads in the process , As long as the program is still going , Just print the track

    [C:\~]$ jstack 37476
    2022-02-22 19:17:04
    Full thread dump Java HotSpot(TM) 64-Bit Server VM (25.231-b11 mixed mode):
    
    "DestroyJavaVM" #24 prio=5 os_prio=0 tid=0x0000000002f94000 nid=0x8a40 waiting on condition [0x0000000000000000]
       java.lang.Thread.State: RUNNABLE
    
    "myThreadB" #23 prio=5 os_prio=0 tid=0x0000000025314000 nid=0x5018 waiting for monitor entry [0x00000000266bf000]
       java.lang.Thread.State: BLOCKED (on object monitor)
    	at rod.TestMain.lambda$main$1(TestMain.java:35)
    	- waiting to lock <0x0000000743a6a820> (a java.lang.Object)
    	- locked <0x0000000743a6a830> (a java.lang.Object)
    	at rod.TestMain$$Lambda$2/1349414238.run(Unknown Source)
    	at java.lang.Thread.run(Thread.java:748)
    
    "myThreadA" #22 prio=5 os_prio=0 tid=0x0000000025313000 nid=0x2794 waiting for monitor entry [0x00000000265bf000]
       java.lang.Thread.State: BLOCKED (on object monitor)
    	at rod.TestMain.lambda$main$0(TestMain.java:21)
    	- waiting to lock <0x0000000743a6a830> (a java.lang.Object)
    	- locked <0x0000000743a6a820> (a java.lang.Object)
    	at rod.TestMain$$Lambda$1/1873653341.run(Unknown Source)
    	at java.lang.Thread.run(Thread.java:748)
    
    "Service Thread" #21 daemon prio=9 os_prio=0 tid=0x0000000024ff1000 nid=0x8cc4 runnable [0x0000000000000000]
       java.lang.Thread.State: RUNNABLE
       
    ...
    
    "VM Thread" os_prio=2 tid=0x00000000215f7000 nid=0x8710 runnable 
    
    "GC task thread#0 (ParallelGC)" os_prio=0 tid=0x0000000002faa800 nid=0x9378 runnable 
    ...
     Find a Java Level deadlock :
    =============================
    "myThreadB":
       Wait to lock the monitor  0x0000000021601c08 (object 0x0000000743a6a820, a java.lang.Object),
       This is from  "myThreadA"
    "myThreadA":
       Wait to lock the monitor  0x0000000021604338 (object 0x0000000743a6a830, a java.lang.Object),
       This is from  "myThreadB"
    
     Of the threads listed Java Stack information :
    ===================================================
    "myThreadB":
    	at rod.TestMain.lambda$main$1(TestMain.java:35)
    	- waiting to lock <0x0000000743a6a820> (a java.lang.Object)
    	- locked <0x0000000743a6a830> (a java.lang.Object)
    	at rod.TestMain$$Lambda$2/1349414238.run(Unknown Source)
    	at java.lang.Thread.run(Thread.java:748)
    "myThreadA":
    	at rod.TestMain.lambda$main$0(TestMain.java:21)
    	- waiting to lock <0x0000000743a6a830> (a java.lang.Object)
    	- locked <0x0000000743a6a820> (a java.lang.Object)
    	at rod.TestMain$$Lambda$1/1873653341.run(Unknown Source)
    	at java.lang.Thread.run(Thread.java:748)
    
    

  • You may see some thread tracks :

    • DestroyJavaVM -> Destruction of the thread **(RUNNABLE= Running state )**
    • myThreadB , myThreadA -> We define the thread directly **(BLOCKED= Blocked state )**
    • Service Thread -> There are many other names daemon So is the daemon thread Running state Let's not list
    • GC -> Garbage collection thread **(RUNNABLE= Running state )**

  • And at the end, it indicates the location of the deadlock

jmap -> Java map: Export heap memory image file

• jmap It is mainly used to export Heap memory image file , See if there is a memory leak etc. .

  • out of memory : The memory is full , Fried , You also have to create new objects , Direct explosion -> Overflow is easy to remember , It's full , The reverse is leakage
  • Memory leak : The memory is not full , however There is a lot of garbage that can be recycled but can't be recycled , Standing in the manger

• The production environment will generally Configure the following parameters , Let the virtual machine be in OOM Auto generate after exception dump file

  • Dump File is process Of Memory image . You can put the program's Execution status adopt The debugger Save to dump In file .
  • It is mainly used in the system An exception or crash occurs Time to generate dump file
  • And then use Debugger to debug , In this way, the dump Copy the file to your own development machine
  • Debugging can find the location of the program error .

  // Output error heap Dump Information   route /Users/peng
  -XX:+HeapDumpOnOutOfMemoryError -XX:HeapDumpPath=/Users/peng
  

• Execute the following command Manual acquisition dump file

  jmap -dump:file= file name .dump  process id
  

jstat -> java stat: see jvm Statistics

• jstat Can be displayed Local or remote virtual machine In process Class loading 、 Memory 、 garbage collection 、 JIT( compiler ) compile Etc

• use jstat Take a look at the class loading information . Personally, I rarely use this command , Viewing garbage collection information from the command line is not as convenient as viewing the graphical interface

  [C:\~]$ jstat -class  37476
   Number of loaded classes 	 Number of bytes to load class 	 Number of unloaded classes   The number of bytes to unload the class 	 Time spent 
  Loaded  		Bytes 		Unloaded  	Bytes    	 Time   
     661  		1285.3        0     	0.0      	 0.17
  

give the thumbs-up , handsome young man !!!