Shulou(Shulou.com)06/01 Report--

Today, I will talk to you about the example analysis caused by the computer avalanche caused by the OOM test in Java. Many people may not know much about it. In order to make you understand better, the editor has summarized the following for you. I hope you can get something from this article.

The problem is emerging-computer avalanche

The experiment goes like this: want to test the OOM type caused by constantly creating multithreads at a specified stack size (160k)

The code of the experiment is as follows:

Publicclass Test {

Private void dontStop () {

While (true) {

}

}

Public void stackLeakByThread () {

While (true) {

Thread thread = new Thread (new Runnable () {

@ Override public void run () {

DontStop ()

}

});

Thread.start ()

}

}

Public static void main (String [] args) {

Test oom = new Test ()

Oom.stackLeakByThread ()

}

}

After a while, the fan turned wildly, and soon OOM occurred, and then the program did not terminate, even if it could not be terminated with Ctrl + C, it would prompt "the VM may need to be forcibly terminated", as shown in the figure.

The computer is stuck and the mouse and keyboard can't respond at all! I had to restart the computer, and then I first entered the top command in the terminal, and then executed the above program, and found that the load of CPU reached 800%!

In the above description of the problem, there are at least three questions worth thinking about.

Why does the above while (true) cause cpu load 800% after the occurrence of OOM in the main thread, I tried to terminate the execution of the Java process with Ctrl + C at the terminal, but failed, why did the abort signal not take effect? why didn't the Java process stop running after the main thread had OOM?

Look at it one by one

The relationship between while (true) and cpu load

First of all, we have to understand the meaning of% CPU, which refers to the percentage of a core occupied by a process. If a process starts multiple threads, multiple threads will occupy multiple cores, which may exceed 100%, but no more than the number of CPU cores * 100%. How to check the number of logical CPU

You can use cat / proc/cpuinfo under Linux | grep "processor" | wc-l

Mac can use sysctl hw.logicalcpu

My computer is Mac, using the above command to check the logic core and found that it is 8, but the CPU share seen in the experiment is 800%, that is to say, our experimental program is full of 8 logic CPU! Some people say that it is because you are constantly creating threads, and of course you have filled all the CPU. Let's try it again. Only seven threads are created, plus a total of eight main threads. Only one while (true) is executed inside these eight main threads, as shown below.

Publicclass Test {

Privateint threadCount = 0

Private void dontStop () {

While (true) {

}

}

Public void stackLeakByThread () {

While (true) {

/ / create only 7 threads, plus the main thread, there are 8 threads if (threadCount > 7) {

Continue

}

Thread thread = new Thread (new Runnable () {

@ Override public void run () {

DontStop ()

}

});

Thread.start (); threadCount++

}

}

Public static void main (String [] args) {

Test oom = new Test ()

Oom.stackLeakByThread ()

}

}

After execution,% CPU is still close to 800% (you can try it, there is no mapping here), that is to say, eight while (true) fill all eight cores, on average, a while (true) fills one core, so the question is, why does a single thread execute while (true) fill a core? doesn't CPU allocate each process according to time slices?

As shown in the figure: the operating system allocates CPU time to different processes according to the time slice scheduling algorithm. If one process runs out of time slices, it will cede control of CPU to other processes to execute.

First of all, it needs to be pointed out that CPU does assign its control to different processes according to time slices.

However, CPU's allocation strategy for time slices is dynamic and biased, which is simply understood as follows:

It is true that threads in Java will relinquish the right to execute CPU after executing the time slices allocated by the system, but other processes will tell the system that they have nothing to do and do not need so much time, and the system will switch to the next process until it returns to the process of this endless loop, and the Java process will always report something to do whenever it is recycled. The system will allocate as much time as possible to it (the so-called crying baby is fed), and the system will constantly raise the priority of the while (true) thread and increase its CPU time slice, that is to say, the dead cycle while (true) uses up the time saved by other processes and does not allow CPU to have a rest, resulting in excessive CPU load, which is like the Matthew effect, the harder the diligent thread executes. The more time slices other lazy threads are shortened, the less opportunities they get!

Voiceover: there is a function called "Priority Boosting" (can be turned off) in the Windows system, which roughly means that when the system finds that a thread is executing very hard, it may assign execution time to the thread beyond the thread priority.

Why can't Ctrl+C abort the Java process after OOM occurs

As mentioned above, after the occurrence of OOM, because the phenomenon of OOM has been observed, I want to kill the Java process through Ctrl+C, but it does not work, as shown in the figure

Why doesn't Ctrl + C, the general way of kill dropping processes, work? I found the answer of Oracle engineers in the Oracle forum (see the reference link at the end of the article).

The message "Java HotSpot (TM) 64-Bit Server VM warning: Exception java.lang.OutOfMemoryError occurred dispatching signal UNKNOWN to handler- the VM may need to be forcibly terminated" is getting printed by the JVM's native signal handling code. The signal handler itself encountered OOM while making a Java up-call and that's why the JVM didn't get terminated with ctrl+c.

To put it simply, the signal processor in JVM does receive the termination signal from Ctrl + C sent by the terminal, but when it calls the Java process to abort, OOM causes the interrupt to fail, so why does the call occur OOM? I guess it is because the signal processor wants to start a thread to do this termination notification operation, and we know that the thread can no longer be created (unable to create new native thread error has occurred)

Why does the Java process not stop running after OOM occurs in the main thread

The last question is that the Java process did not terminate after the main thread OOM. How to explain this?

Main main thread and other child thread is not a parent-child relationship, but an equal relationship, so although the main thread is dead because of OOM, but other child threads will not stop running, because the child threads execute while (true), so child threads will always exist, since they always exist, then the corresponding Java process will always be running.

Then how can other threads end immediately after the main thread stops running? you can set these child threads as daemon threads, and after creating the Thread thread, you can use thread.setDaemon (true) to set them as daemon threads, so that when the main thread dies, the daemon thread will stop running immediately. The reason is also very simple. Since it is a daemon thread, the guarded thread is dead, so there is no meaning for the daemon thread to exist.

Through an OOM experiment led to three questions worth thinking, I believe you should learn a lot of knowledge points, here or to remind you, when you see the demo in the book, it is best to try it in person, maybe you can make a new discovery! Paper will sleep shallow, never know the matter want to practice! After reading the above, do you have any further understanding of the example analysis caused by the computer avalanche caused by the OOM test in Java? If you want to know more knowledge or related content, please follow the industry information channel, thank you for your support.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.