Understanding the OutOfMemoryError

August 7, 2013 by Nikita Salnikov-Tarnovski

Whenever you find yourself staring a stacktrace with OutOfMemoryError in it, it should all be crystal clear. The program has got no more elbow room and is dying just because of lack of it. From the 10,000 feet or executive chair this might already contain too much information. But those of you who have to build or maintain the applications and figure out why a particular error is created – we can share a bit more insight into the issue.

In this post we will take a look what do different OutOfMemoryError messages actually mean. We start with the most common cases and move forward to the more interesting situations.

  • java.lang.OutOfMemoryError: Java heap space
  • java.lang.OutOfMemoryError: PermGen space
  • java.lang.OutOfMemoryError: GC overhead limit exceeded
  • java.lang.OutOfMemoryError: unable to create new native thread
  • java.lang.OutOfMemoryError: nativeGetNewTLA
  • java.lang.OutOfMemoryError: Requested array size exceeds VM limit
  • java.lang.OutOfMemoryError: request <size> bytes for <reason>. Out of swap space?
  • java.lang.OutOfMemoryError: <reason> <stack trace> (Native method)

 java.lang.OutOfMemoryError: Java heap space. We start with the one you have all seen more than you actually would like. This is the Java Virtual Machine’s way to announce you that there is no more room in the virtual machine heap area. You are trying to create a new object, but the amount of memory this newly created structure is about to consume is more than the JVM has in the heap. The JVM has tried to free the memory by calling full GC before throwing in the towel, but without any success. The fastest way to get rid of the symptoms is to increase the heap via -Xmx parameter. Note that this nor the other the recommendations in the article should be taken with a grain of salt. More often than not you just end up hiding the symptoms of the underlying problem.

Next suspect is also quite common. I guess most of you have seen the java.lang.OutOfMemoryError: PermGen space during redeploys. It is pretty much the same message as the first one, but instead of the heap you are now trying to allocate memory in Permanent Generation area. And again, you do not have enough room, so the JVM native code is kind enough to let you know about it. This message tends to disappear (for awhile) if you increase the -XX:MaxPermSize parameter.

Third one – the java.lang.OutOfMemoryError: GC overhead limit exceeded – is a bit different beast. Instead of the missing heap / permgen the JVM is signaling that your application is spending too much time in the garbage collection with little to show for it. By default the JVM is configured to throw this error if you are spending more than 98% of the total time in GC and after the GC less than 2% of the heap is recovered. Sounds like a perfectly good place to have the “fail fast” safeguard at place. On rare cases where it makes sense to disable it, add -XX:-UseGCOverheadLimit to your startup scripts.

The first three OutOfMemoryError messages conduct up to 98% of the cases we are solving using Plumbr. So there is a strong chance that the remaining trio is somewhat unknown to you.

java.lang.OutOfMemoryError: unable to create new native thread is a message you will receive if the JVM is asking a new thread from the OS and the underlying OS cannot allocate a new thread anymore. This limit is very platform dependent, so if you are curious to find out your limitations then run your own little experiment using the following code snippet. On my 64bit MacOS X running a latest JDK 7 I run into troubles when creating thread #2032.

    new Thread(new Runnable(){
        public void run() {
            try {
            } catch(InterruptedException e) { }        

java.lang.OutOfMemoryError: nativeGetNewTLA is a symptom where the JVM cannot allocate new Thread Local Area. This is something you only encounter on jRockit virtual machine. If you recall, the Thread Local Area is the buffer used to efficiently allocate memory in multi-threaded application. Each thread has its own pre-allocated buffer where all the objects instantiated by this thread are born. You will run into problems when you are creating vast amount of objects in heavily multi-threaded application, in case of which you might turn to the -XXtlaSize parameter tweaking.

java.lang.OutOfMemoryError: Requested array size exceeds VM limit is a message you find yourself staring at when you are trying to create an array larger than your VM limitations allow. On my 64bit Mac OS X with a recent JDK 7 build I find myself acknowledging the fact that arrays with Integer.MAX_INT-2 elements are OK, but if just one more straw, namely to Integer.MAX_INT-1, breaks the camel’s back. On older 32-bit machines it had its benefits limiting the array sizes to fit into the tiny heaps available back then. On modern 64bit machines it seems to create more confusion than to actually help solving anything.

java.lang.OutOfMemoryError: request <size> bytes for <reason>. Out of swap space? This error message is thrown when the JVM fails to allocate native memory from the OS. Note that it is completely different from the standard cases where you have exhausted the heap or permgen spaces. This message tends to be displayed when you are operating close to the platform limits. As the message itself is stating you might have exceeded the amount of physical and virtual memory available. As the latter is often implemented via swapping the memory to the disk, then the first thing you might think of as a quick fix would be to increase the size of the swap file. But I am yet to see an application which would behave normally while swapping, so most likely this quick fix won’t help you much.

java.lang.OutOfMemoryError: <reason> <stack trace> (Native method) Now it is time to beg help from your fellow C developers. As the message states you are facing problems from the native code, but as opposed to the last case the allocation failure was detected in a JNI or native method instead of the JVM code.

All the recommendations in the article should be taken with a grain of salt. More often than not you just end up hiding the symptoms of the underlying problem. In case you want to be sure those messages were not caused by a memory leak, download and try out Plumbr for free.

Can't figure out what causes your OutOfMemoryError? Read more




Thank you for the comprehensive list, I wasn’t aware of “Requested array size exceeds VM limit”.

There’s one you haven’t mentioned, Non-Heap area too small or leaks in that area which results in java.lang.OutOfMemoryError (no message). We’ve seen that in case of leaking classes and classloaders, for example stubs/proxies from jaxws.

And regarding Out of swap space, my scenario was a bit different, it wasn’t really a leak. Here’s what I’ve recorded then: This may happen if you are making lots of heavy JNI calls, but the JavaHeap objects occupy little space. In that scenario the GC might not feel the urge to cleanup JavaHeap, while the JNI Heap keeps on increasing till it goes out of memory. If you use java NIO packages, watch out for this issue. DirectBuffer allocation uses the native heap. The NativeHeap can be increasded by -XX:MaxDirectMemorySize=256M (default is 128)

Bogdan Tanasa

Thank you for your comment. The issue of DirectBuffers is certainly worth a dedicated mention of its own.

Shouldn’t leaking classes end in OOM: Perm Gen ? Can you explain your case in more details?

Nikita Salnikov-Tarnovski

Can't figure out what causes your OutOfMemoryError? Read more

You cannot predict the way you die
When debugging a situation where systems are failing due to the lack of resources, you can no longer count on anything. Seemingly unrelated changes can trigger completely different messages and control flows within the JVM. Read more
Tuning GC - it does not have to be that hard
Solving GC pauses is a complex task. If you do not believe our words, check out the recent LinkedIn experience in garbage collection optimization. It is a complex and tedious task, so we are glad to report we have a whole lot simpler solution in mind Read more
Building a nirvana
We have invested a lot into our continuous integration / delivery infrastructure. As of now we can say that the Jenkins-orchestrated gang consisting of Ansible, Vagrant, Gradle, LiveRebel and TestNG is something an engineer can call a nirvana. Read more
Creative way to handle OutOfMemoryError
Wish to spend a day troubleshooting? Or make enemies among sysops? Registering pkill java to OutOfMemoryError events is one darn good way to achieve those goals. Read more