GC seems to make java server application unresponsive problem, how to solve it

GC seems to make Java server applications unresponsive, how to solve it

System.gc() “calls” the garbage collector to collect garbage. This is not very rigorous, but in fact it is just a “notification” “That’s all. Whether to recycle or not is determined by the garbage collector’s algorithm. You can develop a garbage collector that does nothing, or a garbage collector that waits until the memory is occupied to exceed a certain proportion before recycling. The finalize() method is a method that will be called when a class object is destroyed. When the garbage collector finds that the class object will no longer be used, it will reclaim the memory, that is, destroy the object, and finalize() is called.

System.gc() of your program is the last sentence. Obviously, the Book object created in the previous sentence will not be used later, so the garbage collection algorithm provided by JDK will call the destruction method of the object to recycle it. .

The java process on the server is still there, but it is not executed. What’s going on?

1. Obviously there is no log. Any program must have corresponding log records. In this way, any problems that arise can be traced to the root cause through the logs.

2. The key to log operation is to output the corresponding information at key locations. It does not have to be comprehensive, which will affect performance.

Let’s try again.

Use jdk common tools to troubleshoot the troubleshooting process

jps locates the process, jstat statistics heap information, jstack locates the problem thread, jmap locates the problem object, jps is used to check which java processes are currently on the server, and when troubleshooting problems , generally use jps to locate the pid first. The -l parameter can print the complete class path or jar package path, and the -v parameter can print startup parameters. Note: ps -ef | grep java can print startup parameters and jar paths, but the class path jstat cannot be printed. Used to count the usage of each area of ​​the heap and other information. In the figure below, the jstat -gc 1 1000 5 command means that the heap information of the process with pid 1 is printed every 1000ms, printing 5 times in total. S0C represents the capacity of S0, which is the total capacity. S0U represents the used space of S0, which is the used space in kb. U/C can get the usage rate, such as MU/MC = meta space usage rate.

E, O, and M represent the Eden area, the old generation, and the metaspace respectively. YGC/YGCT represents the number of GCs/GC time of YGC, and FGC/FGCT represents the number of GCs/GC time of FGC.

-gc parameter displays the specific value, -gcutil display ratio jstack is used to view thread information within a process, and is often used to troubleshoot CPU problems. When using jstack, the general steps are as follows: Step 1, use top -Hp pid Check the cpu ratio of the threads in the pid process and sort them by cpu usage from large to small. What needs to be observed in this step is: threads with a relatively high CPU usage, and threads with a long CPU running time (i.e. TIME+column). There is a question here, how high is the CPU? Suppose there is a 4-core CPU, one core runs at 100%, and the entire CPU runs at a maximum of 400%. If there is a thread infinite loop (while true), the CPU can generally reach 90%+, but it cannot reach 100% because even if it is while true, there will still be time slice rotation. In addition, the CPU running time is also a very important indicator. In the picture above, the thread with pid 71 ran for 2 hours and 18 minutes. There must be something wrong.

Step 2, export the current thread information to the thread dump file. Step 3, according to the pid of the problem thread found in the previous step, convert the pid into hexadecimal, because in the thread dump file, the thread pid is 16 In step 4 of hexadecimal representation, retrieve the hexadecimal pid of the thread pid calculated in the previous step in the thread dump file. Generally, if you look at the last 20 lines, you can see the complete call stack. This thread is a kafka consumer thread. Note : During development, the thread name must be determined and can correspond to the specific business. In this way, when using jstack, you can immediately locate the specific business based on the thread name. In addition, the getAllStackTraces method of the Thread class can obtain the call stacks of all threads of the current virtual machine. If developers are not convenient for production, they can use this API to expose the interface. jmap is used to view object information of a process, often used to troubleshoot memory problems. jmap -dump:live,format=b,file=/root/1.hprof 1 prints the contents of the heap to a file. The live parameter indicates that only live ones are output. object.

Note: The size of the heap dump file is consistent with the current heap usage. If the heap usage is 2G, the heap dump file will be 2G. It takes a long time to export the heap dump file, and the export process may Will affect external services. In addition, heap dump files can only be viewed with professional tools (such as jhat, jvisualvm, mat) and cannot be retrieved with commands such as grep. In summary, this command is generally rarely used in production environments.

jmap -histo:live pid counts the number of class instances and the number of bytes. Generally, use jmap -histo:live pid | grep packge to check whether there is a memory leak. Page is the package name of the java project. jmap -J-D64 -heap 1. Print the summary information of the heap. Follow the above steps. Generally, you can locate the problem, but it is more difficult. , need to do a lot of operations in a short time, therefore, it is recommended to use Alibaba’s arthas.

How to find the cause of Java process zombie

When encountering this situation, you can generally deal with it from the following aspects (not necessarily all): 1. Check the business log. If the system log is If it is good, most of the problems can usually be found through this. 2. Check it with vmstat, iostat, top and other commands.System resource consumption, including CPU, IO, network, etc. 3. Check the Tomat log 4. Check the GC status with jstat -gcutil 5. Check the GC log 6. If it still doesn’t work, check the system log again, such as too many open files The error is usually caused by the exhaustion of the number of file handles.

Why does the JAVA program not respond when I click run as?

Hello, you are using eclipse. Click project -> clean on the menu. Look at your project, select autobuilding, and see if run as application responds.