Oom happened. Do you know the reason and how to solve it?

Preface

Lu java Code students , More or less, you will encounter memory overflow （OOM） Scene , But it causes OOM There is more than one reason . Let's summarize the common OOM The reasons and solutions .

1、 Out of heap memory （Java heap space）

reason ：

1. There may be large object assignments in the code , Usually a big tree group .
2. Possible memory leak , Lead to many times GC after , Still can't find a block of memory large enough to hold the current object , Commonly used in the use of File Wait for resources not to be recycled .
3. More than expected visits 、 Data volume , Usually the upstream system request traffic soars , Common in all kinds of promotions 、 Seckill activity , You can check whether there is a sharp peak in combination with business traffic indicators .

Solution ：

1. For most cases , Usually it's just through -Xmx Turn up the parameters JVM Heap memory is enough . If it's still not solved , You can refer to the following situations for handling .
2. If it's a super object , Check its rationality , For example, whether all the results of the database are queried at one time , And there's no limit to the number of results .
3. If it's peak business pressure , Consider adding machine resources , Or do current limiting and degrading operation .
4. If it's a memory leak , You need to find the holding object , Modify the code design , Like closing a connection that has not been released .

2、 Permanent generation space / Meta space （Permgen space/ metaspace）

reason ：

1. Permanent generation is HotSot The specific implementation of virtual machine to method area , Store class information loaded by virtual machine 、 Constant pool 、 Static variables ,JIT Compiled code, etc
2. Perngen Usage and load into memory class The number of / Size is positively correlated

Solution ：

1. Check whether the space is permanently replaced （ Launch parameters ：-XX:MaxPermSize） Or meta space （ Launch parameters ：-XX:MaxMetaspaceSize） Set too small .
2. Check whether there are a lot of reflection operations in the code
3. dump After through mat Check if there are a large number of proxy classes generated by reflection
4. Application deployment times error , It is likely that the application has not been restarted , Results in multiple copies of class Information , restart JVM Can solve

3、GC overhead limit exceeded

reason ：

1. When java Process cost 98% Do it in the above time GC And it can't be recycled 2% This exception will be thrown in the heap memory of .
2. Heap memory is too small

Solution ：

1. Check whether there are a lot of dead loops or code using large memory in the project , Optimize the code
2. dump Memory , Check for memory leaks , without , Increase memory .

4、 Method stack overflow （Unable to create new native thread）

reason ：

1. The number of threads exceeds the maximum number of threads of the operating system ulimit Limit
2. The number of threads exceeds kernal.pid_max
3. native Out of memory

Solution ：

1. Upgrade configuration , Provide more memory for the machine
2. Reduce Java heap space size
3. Fix the thread leak problem of the application
4. Limit thread pool size
5. Use -Xss Parameter reduce the thread stack size
6. Raise height OS The maximum number of threads in the layer ： perform ulimit -a View the maximum number of threads limit , Use ulimit -u xx Adjust the maximum number of threads limit

5、swap Area overflow （Out of swap space）

This error indicates that all available virtual memory has been exhausted . Virtual memory （Virtual Memory） By physical memory （Physical） And swap space （Swap Space） Two parts . When the virtual memory requested by the runtime program overflows, it will report Out of swap space error .

reason ：

1. Not enough address space
2. Physical memory has run out
3. Local memory leak of application , For example, keep applying for local memory , Without releasing
4. perform jmap -histo:live command , Enforcement Full GC; If the memory drops significantly after several executions , It is basically confirmed as Direct ByteBuffer problem

Solution ：

1. Upgrade the address space to 64bit
2. Use Arthas Check if it is Inflater/Deflater Decompression problem , If it is , Show call end Method
3. Direct ByteBuffer The problem can be solved by starting parameters –XX:MaxDirectMemorySize Lower the threshold
4. enlarge swap Partition size or increase the machine memory size
5. Isolated deployment , Avoid fighting

6、 Allocate large arrays （Requested array size exceeds VM limit）

reason ：

This situation is generally caused by unreasonable array allocation requests , Before allocating memory for an array ,JVM An inspection will be carried out . Whether the array to be allocated can be addressed in the platform , If you can't address, you will throw this error .

Solution ：

Check whether there is a place to create super large arrays in the code

7、Direct buffer memory

Java Allow applications to pass through Direct ByteBuffer Direct access to out of heap memory , Many high-performance programs pass Direct ByteBuffer Combined with memory mapping file （Memory MappedFile） High speed IO

reason ：

Direct ByteBuffer The default size is 64MB, Once the use exceeds the limit , Will throw Direct buffer memory error .

Solution ：

Java Only through ByteBuffer.allocateDirect Methods use DirectByteBuffer, therefore , Can pass Arthas And other online diagnostic tools intercept this method for troubleshooting .

1. Check for direct or indirect use of NIO, Such as netty,jetty etc.
2. Through the startup parameters -XX:MaxDirectMemorySize adjustment Direct ByteBuffer Upper limit value
3. Check JVM Whether the parameter has -XX:+DisableExplicitGC Options , If there is one, remove it , Because this parameter makes System.gc() invalid
4. Check the code of off heap memory , Confirm whether there is memory leak ; Or call through reflection sun.misc.Cleaner Of clean() Methods to actively release Direct ByteBuffer Memory space held .
5. The memory capacity is really not enough , Upgrade configuration

8、Kill process or sacrifice child

There is a kernel job （Kernel Job） be known as Out of Memory Killer, It will be in the case of very low available memory “ Kill ”（kill） Some processes .OOM Killer All processes will be scored , Then the lower scoring process “ Kill ”, Specific scoring rules can be referred to Surviving the Linux OOM Killer.

Different from other OOM error ,Kill processorsacrifice child Mistakes are not caused by JVM Level triggered , It's triggered at the operating system level .

reason ：

By default ,Linux The kernel allows processes to request more memory than the system can use , Through this kind of “ Peak shifting multiplexing ” In this way, system resources can be used more effectively .

However , This way will inevitably bring some “ Oversold ” risk . For example, some processes continue to occupy system memory , And then cause other processes to have no available memory . here , The system will automatically activate OOM Killer, Look for low scoring processes , And its “ Kill ”, Free up memory resources .

Solution ：

1. Upgrade server configuration / Isolated deployment , Avoid contention .
2. OOM Killer tuning .

Conclusion

Okay , I'll share it with you today . If this article helps you , Welcome to thumb up & Collection & Share , This is for me to continue to share & It's very important to create quality articles . thank

-------------- end ----------------

WeChat official account ： Below scan QR code or Search for Laugh at Fengyun Road Focus on , Get more big data dry goods ！

​