寫在前面
上周，同事寫了一段ConcurrentHashMap的測試代碼，說往map里放了32個元素就內存溢出了，我大致看了一下他的代碼及運行的jvm參數，覺得很奇怪，於是就自己搗鼓了一下。首先上一段代碼：
public class MapTest { public static void main(String[] args) { System.out.println("Before allocate map, free memory is " + Runtime.getRuntime().freeMemory()/(1024*1024) + "M"); Map<String, String> map = new ConcurrentHashMap<String, String>(2000000000); System.out.println("After allocate map, free memory is " + Runtime.getRuntime().freeMemory()/(1024*1024) + "M"); int i = 0; try { while (i < 1000000) { System.out.println("Before put the " + (i + 1) + " element, free memory is " + Runtime.getRuntime().freeMemory()/(1024*1024) + "M"); map.put(String.valueOf(i), String.valueOf(i)); System.out.println("After put the " + (i + 1) + " element, free memory is " + Runtime.getRuntime().freeMemory()/(1024*1024) + "M"); i++; } } catch (Exception e) { e.printStackTrace(); } catch (Throwable t) { t.printStackTrace(); } finally { System.out.println("map size is " + map.size()); } } }
執行時加上jvm執行參數 -Xms512m -Xmx512m ，執行結果：
Before allocate map, free memory is 120M After allocate map, free memory is 121M Before put the 1 element, free memory is 121M After put the 1 element, free memory is 121M Before put the 2 element, free memory is 121M After put the 2 element, free memory is 122M Before put the 3 element, free memory is 122M After put the 3 element, free memory is 122M Before put the 4 element, free memory is 122M After put the 4 element, free memory is 122M Before put the 5 element, free memory is 122M After put the 5 element, free memory is 114M Before put the 6 element, free memory is 114M java.lang.OutOfMemoryError: Java heap space map size is 5 at java.util.concurrent.ConcurrentHashMap.ensureSegment(Unknown Source) at java.util.concurrent.ConcurrentHashMap.put(Unknown Source) at com.j.u.c.tj.MapTest.main(MapTest.java:17)
最開始的代碼是沒有加入一些日志打印的，當時就很奇怪，為什么只往map里放一個元素就報了OutOfMemoryError了。
於是就加了上述打印日志，發現在創建map的時候已經占用了二百多兆的空間，然后往里面put一個元素，put前都還有二百多兆，put時就報了OutOfMemoryError， 那就更奇怪了，初始化map的時候會占用一定空間，可以理解，但是只是往里面put一個很小的元素，為什么就會OutOfMemoryError呢？
排查過程
1、第一步，將第一段代碼中的Map<String, String> map = new ConcurrentHashMap<String, String>(2000000000);修改為Map<String, String> map = new ConcurrentHashMap<String, String>();，這次運行正常。（沒有找到問題根因，但是以后使用ConcurrentHashMap得注意：1、盡量不初始化；2、如果需要初始化，盡量給一個比較合適的值）
2、第二步，執行時加上jvm參數-Xms20124m -Xmx1024m 。發現還是同樣出現問題。
3、第三步，分析ConcurrentHashMap源代碼，首先，了解一下ConcurrentHashMap的結構，它是由多個Segment組成（每個Segment擁有一把鎖，也是ConcurrentHashMap線程安全的保證，不是本文討論的重點），每個Segment由一個HashEntry數組組成，出問題就在這個HashEntry上面
4、第四步，查看ConcurrentHashMap的初始化方法，可以看出，初始化了Segment[0]的HashEntry數組，數組的長度為cap值，這個值為67108864
cap的計算過程（可以針對於初始化過程進行調試）
1）initialCapacity為2,000,000,000，而MAXIMUM_CAPACITY的默認值（也即ConcurrentHashMap支持的最大值是1<<30，即230=1,073,741,824），initialCapacity的值大於MAXIMUM_CAPACITY，即initialCapacity=1,073,741,824
2）c的值計算為 initialCapacity/ssize=67108864
3）cap為 大於等於c的第一個2的n次方數 也即67108864
public ConcurrentHashMap(int initialCapacity, float loadFactor, int concurrencyLevel) { if (!(loadFactor > 0) || initialCapacity < 0 || concurrencyLevel <= 0) throw new IllegalArgumentException(); if (concurrencyLevel > MAX_SEGMENTS) concurrencyLevel = MAX_SEGMENTS; // Find power-of-two sizes best matching arguments int sshift = 0; int ssize = 1; while (ssize < concurrencyLevel) { ++sshift; ssize <<= 1; } this.segmentShift = 32 - sshift; this.segmentMask = ssize - 1; if (initialCapacity > MAXIMUM_CAPACITY) initialCapacity = MAXIMUM_CAPACITY; int c = initialCapacity / ssize; if (c * ssize < initialCapacity) ++c; int cap = MIN_SEGMENT_TABLE_CAPACITY; while (cap < c) cap <<= 1; // create segments and segments[0] Segment<K,V> s0 = new Segment<K,V>(loadFactor, (int)(cap * loadFactor), (HashEntry<K,V>[])new HashEntry[cap]); Segment<K,V>[] ss = (Segment<K,V>[])new Segment[ssize]; UNSAFE.putOrderedObject(ss, SBASE, s0); // ordered write of segments[0] this.segments = ss; }
5、第五步，查看ConcurrentHashMap的put方法，可以看出在put一個元素的時候，
1）計算當前key的hash值，查看當前key所在的semgment有沒有初始化，若果有，則執行后續的put操作。
2）如果沒有去執行ensureSement()方法，而在ensureSement()方法中又會初始化了一個HashEntry數組，數組的長度和第一個初始化的Segment的HashEntry的長度一致。
public V put(K key, V value) { Segment<K,V> s; if (value == null) throw new NullPointerException(); int hash = hash(key); int j = (hash >>> segmentShift) & segmentMask; if ((s = (Segment<K,V>)UNSAFE.getObject // nonvolatile; recheck (segments, (j << SSHIFT) + SBASE)) == null) // in ensureSegment s = ensureSegment(j); return s.put(key, hash, value, false); } private Segment<K,V> ensureSegment(int k) { final Segment<K,V>[] ss = this.segments; long u = (k << SSHIFT) + SBASE; // raw offset Segment<K,V> seg; if ((seg = (Segment<K,V>)UNSAFE.getObjectVolatile(ss, u)) == null) { Segment<K,V> proto = ss[0]; // use segment 0 as prototype int cap = proto.table.length; float lf = proto.loadFactor; int threshold = (int)(cap * lf); HashEntry<K,V>[] tab = (HashEntry<K,V>[])new HashEntry[cap]; if ((seg = (Segment<K,V>)UNSAFE.getObjectVolatile(ss, u)) == null) { // recheck Segment<K,V> s = new Segment<K,V>(lf, threshold, tab); while ((seg = (Segment<K,V>)UNSAFE.getObjectVolatile(ss, u)) == null) { if (UNSAFE.compareAndSwapObject(ss, u, null, seg = s)) break; } } } return seg; }
6、這個時候能定位到是因為put一個元素的時候創建了一個長度為67103684的HashEntry數組，而這個HashEntry數組會占用67108864*4byte=256M，和上面的測試結果能對應起來。為什么數組會占用這么大的空間，很多同學可能會有疑問，那來看一下數組的初始化 而數組初始化會在堆內存創建一個HashEntry引用的數組，且長度為67103684，而每個HashEntry引用（所引用的對象都為null）都占用4個字節。
問題總結
1、ConcurrentHashMap初始化時要指定合理的初始化參數（當然本人做了一個小測試，指定初始化參數暫時沒有發現性能上的提升）