JVM tuning related commands and explanations

Reference to the original :https://www.zhihu.com/question/427461208/answer/2537026920

jinfo

You can print some current jvm All kinds of parameters , such as jvm Some parameters to start ,jvm Some properties in k-v etc. , Can pass jinfo --help View specific commands .

jmap（ Memory overflow solution ）

This parameter allows you to view JVM Some related data in memory ;

adopt jps obtain java Process PID, adopt jmap Analyze the present java Memory data of the process , Common commands ：

• 1. Heap history ：jmap -histo[:live] <pid> [ > ./xx.log] : You can see the current JVM The number of objects created by all loaded classes in , Occupy memory, etc , You can import files to view ;
• 2. Heap information ：jmap -heap <pid> ： You can see java The proportion of the new generation and the old generation of the program is the usage .
• 3. Pile up dump jmap -dump：format=b,file=xx.hprof <pid> : Sure dump Heap log , See and jmap similar , have access to visualVM see .

adopt jmap You can view the heap usage , For example, locating simple memory leaks , adopt 1,3 You can see that there are many objects of those classes .
You can set parameters to control the recording when a memory overflow occurs xx.hprof journal , This can be used to analyze memory leaks .-XX:+HeapDumpOnOutOfMemoryError -XX:HeapDumpPath=xxx/xxx/xx.hprof

jstack（cpu High occupancy solutions ）

This command can view the stack information of the thread ,jstack You can locate a simple deadlock , The most common way is through jstack location CPU High problem , The specific steps are ：

• 1.top Command to view the current occupancy cpu Highest process pid;
• 2.top -p <pid> View high memory consumption cpu The process of , Press H To get the occupation in the current process cpu The highest thread of tid;
• 3.jstack <pid> | grep -A10 <16 Base number tid> View the specific class information that the thread that occupies the highest thread is executing . Be careful ：tid Of 16 Lower case base .

Through this process, you can directly locate which thread is executing, which takes a lot of cpu, Including which class to execute , Very convenient , This interview often asks .

jstat（FullGC Frequent solutions ）

This command can view the detailed usage of various parts of the heap , Can pass jstat --help view help ;

The most common command jstat -gc <pid> [1000 10] Show view gc situation , Every time 1 Print once per second, a total of 10 Time ( Optional ), You can view the total size and size of each belt , The key can also be seen youngGC, and FullGC Respective total times , Total time spent and time spent together , about jvm The optimization of is to optimize it FullGC frequency ,FullGC The less, the better. , It's better to control FullGC Once every few hours or even days , It depends on the business situation .

FullGC Analyze the process frequently ：

FullGC We can't directly guess why , It needs some analysis , To determine what caused it , Generally, we need to use the above command to regularly observe the heap .

• 1. Observe Eden The object of the zone grows , For example, how many objects are created per second ;
• 2. happen youngGC The interval time and time consumption of each time , Every time youngGC How many objects enter the new generation and how many enter the old generation ;
• 3. Trigger fullGC Time interval and time consuming , Every time FullGC How much space will be released .

Build the model ：
How many threads in the system will generate per second M The object of ==> ① Get into Eden;② Big target into the old generation ？ ==> n The second is full Eden District ==> Conduct youngGC ==> ① Number of objects entering the elderly generation （ The reasons causing , Dynamic age determination mechanism ？ Age threshold reached ？） ==> FullGC Number of objects remaining after

To analyze problems ：

Trigger FullGC The reason is that the old generation can't fit any more , So we need to see why we put objects frequently in the old age ...

• 1. to JVM The allocated heap memory is too small , Can't meet the requirements , In the old days, every time GC Only a small amount of space can be released , You can increase the size of the heap .
• 2. Large object setting threshold is too small , A temporary large object may be generated in the system , For example, a lot of data is loaded into memory at one time , You can modify the threshold, but it's best to optimize the code .
• 3. The judgment mechanism of dynamic age leads to too many objects entering the elderly generation , If we judge that a large number of business objects are living and dying, we can appropriately increase the proportion of the new generation and the old generation , Avoid dynamic age determination, which leads to many garbage objects entering the elderly generation .
• 4. The old space guarantee mechanism may lead to every youngerGC It's going to be done before FullGC, Whether the space guarantee mechanism has been opened , If it is not turned on and a large number of objects are generated each time, it will result in one or several times youngGC To carry out FullGC