本人最近在学习OpenJDK1.7 的HotSpot VM中遇到了一些问题，麻烦哪位大侠不吝指点下迷津。
我写了一个触发promotion failed的例子，后来分析GC日志时，发现了一些问题，CMS GC并没有按照我设置的参数发生。
JVM 参数设置和程序如下：
-Xmx20M -Xms20M -Xmn10M -XX:+PrintGCDetails -XX:+PrintGCDateStamps -XX:SurvivorRatio=8 -verbose:gc -XX:+PrintHeapAtGC -XX:+UseConcMarkSweepGC -XX:MaxTenuringThreshold=1 -XX:CMSInitiatingOccupancyFraction=80 -XX:+UseCMSInitiatingOccupancyOnly -XX:+UseCMSCompactAtFullCollection -XX:CMSFullGCsBeforeCompaction=10
public static void main(String[] args) throws Exception {
byte[] alloc1,alloc2,alloc3,alloc4, alloc11, alloc22, alloc33, alloc44;
alloc1 = new byte[1*1024*1024];
alloc2 = new byte[1*1024*1024];
alloc3 = new byte[1*1024*1024];
alloc4 = new byte[1*1024*1024];
alloc11 = new byte[1*1024*1024];
Thread.sleep(1000);
System.out.println("*********************1111" + new Date());
alloc22 = new byte[2*1024*1024];
Thread.sleep(1000);
System.out.println("*********************2222" + new Date());
alloc33 = new byte[1*1024*1024];
alloc44 = new byte[1*1024*1024];
byte[] alloc5,alloc6,alloc7,alloc8, alloc9;
alloc5 = new byte[3*1024*1024];
alloc1 = null;
alloc4 = null;
Thread.sleep(2000);
System.out.println("*********************3333" + new Date());
alloc6 = new byte[3*1024*1024];
alloc33 = null;
alloc7 = new byte[3*1024*1024];
Thread.sleep(2000);
System.out.println("*********************4444" + new Date());
alloc8 = new byte[1*1024*1024];
alloc4 = null;
}
打印出来的GC日志是
*********************1111Thu Aug 14 18:57:26 CST 2014
{Heap before GC invocations=0 (full 0):
par new generation total 9216K, used 6827K [0x00000007f9a00000, 0x00000007fa400000, 0x00000007fa400000)
eden space 8192K, 83% used [0x00000007f9a00000, 0x00000007fa0aad18, 0x00000007fa200000)
from space 1024K, 0% used [0x00000007fa200000, 0x00000007fa200000, 0x00000007fa300000)
to space 1024K, 0% used [0x00000007fa300000, 0x00000007fa300000, 0x00000007fa400000)
concurrent mark-sweep generation total 10240K, used 0K [0x00000007fa400000, 0x00000007fae00000, 0x00000007fae00000)
concurrent-mark-sweep perm gen total 21248K, used 3632K [0x00000007fae00000, 0x00000007fc2c0000, 0x0000000800000000)
2014-08-14T18:57:26.451-0800: [GC [ParNew: 6827K->685K(9216K), 0.0043720 secs] 6827K->5807K(19456K), 0.0044000 secs] [Times: user=0.03 sys=0.01, real=0.00 secs]
Heap after GC invocations=1 (full 0):
par new generation total 9216K, used 685K [0x00000007f9a00000, 0x00000007fa400000, 0x00000007fa400000)
eden space 8192K, 0% used [0x00000007f9a00000, 0x00000007f9a00000, 0x00000007fa200000)
from space 1024K, 66% used [0x00000007fa300000, 0x00000007fa3ab770, 0x00000007fa400000)
to space 1024K, 0% used [0x00000007fa200000, 0x00000007fa200000, 0x00000007fa300000)
concurrent mark-sweep generation total 10240K, used 5122K [0x00000007fa400000, 0x00000007fae00000, 0x00000007fae00000)
concurrent-mark-sweep perm gen total 21248K, used 3632K [0x00000007fae00000, 0x00000007fc2c0000, 0x0000000800000000)
}
*********************2222Thu Aug 14 18:57:27 CST 2014
2014-08-14T18:57:28.457-0800: [GC [1 CMS-initial-mark: 5122K(10240K)] 13239K(19456K), 0.0010080 secs] [Times: user=0.00 sys=0.00, real=0.00 secs]
2014-08-14T18:57:28.460-0800: [CMS-concurrent-mark: 0.002/0.002 secs] [Times: user=0.01 sys=0.00, real=0.01 secs]
2014-08-14T18:57:28.460-0800: [CMS-concurrent-preclean: 0.000/0.000 secs] [Times: user=0.00 sys=0.00, real=0.00 secs]
2014-08-14T18:57:28.460-0800: [CMS-concurrent-abortable-preclean: 0.000/0.000 secs] [Times: user=0.00 sys=0.00, real=0.00 secs]
2014-08-14T18:57:28.460-0800: [GC[YG occupancy: 8117 K (9216 K)][Rescan (parallel) , 0.0005060 secs][weak refs processing, 0.0000060 secs][scrub string table, 0.0000370 secs] [1 CMS-remark: 5122K(10240K)] 13239K(19456K), 0.0005850 secs] [Times: user=0.00 sys=0.00, real=0.00 secs]
2014-08-14T18:57:28.461-0800: [CMS-concurrent-sweep: 0.000/0.000 secs] [Times: user=0.00 sys=0.00, real=0.00 secs]
2014-08-14T18:57:28.462-0800: [CMS-concurrent-reset: 0.001/0.001 secs] [Times: user=0.00 sys=0.00, real=0.00 secs]
*********************3333Thu Aug 14 18:57:29 CST 2014
{Heap before GC invocations=1 (full 1):
par new generation total 9216K, used 8117K [0x00000007f9a00000, 0x00000007fa400000, 0x00000007fa400000)
eden space 8192K, 90% used [0x00000007f9a00000, 0x00000007fa141eb8, 0x00000007fa200000)
from space 1024K, 66% used [0x00000007fa300000, 0x00000007fa3ab770, 0x00000007fa400000)
to space 1024K, 0% used [0x00000007fa200000, 0x00000007fa200000, 0x00000007fa300000)
concurrent mark-sweep generation total 10240K, used 3074K [0x00000007fa400000, 0x00000007fae00000, 0x00000007fae00000)
concurrent-mark-sweep perm gen total 21248K, used 3632K [0x00000007fae00000, 0x00000007fc2c0000, 0x0000000800000000)
2014-08-14T18:57:29.459-0800: [GC [ParNew (promotion failed): 8117K->7496K(9216K), 0.0051800 secs][CMS: 7765K->7748K(10240K), 0.0102620 secs] 11191K->10821K(19456K), [CMS Perm : 3632K->3630K(21248K)], 0.0154830 secs] [Times: user=0.04 sys=0.00, real=0.02 secs]
Heap after GC invocations=2 (full 2):
par new generation total 9216K, used 3072K [0x00000007f9a00000, 0x00000007fa400000, 0x00000007fa400000)
eden space 8192K, 37% used [0x00000007f9a00000, 0x00000007f9d00158, 0x00000007fa200000)
from space 1024K, 0% used [0x00000007fa200000, 0x00000007fa200000, 0x00000007fa300000)
to space 1024K, 0% used [0x00000007fa300000, 0x00000007fa300000, 0x00000007fa400000)
concurrent mark-sweep generation total 10240K, used 7748K [0x00000007fa400000, 0x00000007fae00000, 0x00000007fae00000)
concurrent-mark-sweep perm gen total 21248K, used 3630K [0x00000007fae00000, 0x00000007fc2c0000, 0x0000000800000000)
}
2014-08-14T18:57:29.475-0800: [GC [1 CMS-initial-mark: 7748K(10240K)] 13893K(19456K), 0.0000680 secs] [Times: user=0.00 sys=0.00, real=0.00 secs]
{Heap before GC invocations=2 (full 3):
par new generation total 9216K, used 6228K [0x00000007f9a00000, 0x00000007fa400000, 0x00000007fa400000)
eden space 8192K, 76% used [0x00000007f9a00000, 0x00000007fa015060, 0x00000007fa200000)
from space 1024K, 0% used [0x00000007fa200000, 0x00000007fa200000, 0x00000007fa300000)
to space 1024K, 0% used [0x00000007fa300000, 0x00000007fa300000, 0x00000007fa400000)
concurrent mark-sweep generation total 10240K, used 7748K [0x00000007fa400000, 0x00000007fae00000, 0x00000007fae00000)
concurrent-mark-sweep perm gen total 21248K, used 3630K [0x00000007fae00000, 0x00000007fc2c0000, 0x0000000800000000)
2014-08-14T18:57:29.475-0800: [Full GC [CMS2014-08-14T18:57:29.478-0800: [CMS-concurrent-mark: 0.002/0.003 secs] [Times: user=0.00 sys=0.00, real=0.00 secs]
(concurrent mode failure): 7748K->9797K(10240K), 0.0096760 secs] 13976K->12869K(19456K), [CMS Perm : 3630K->3630K(21248K)], 0.0096980 secs] [Times: user=0.01 sys=0.00, real=0.01 secs]
Heap after GC invocations=3 (full 4):
par new generation total 9216K, used 3072K [0x00000007f9a00000, 0x00000007fa400000, 0x00000007fa400000)
eden space 8192K, 37% used [0x00000007f9a00000, 0x00000007f9d00078, 0x00000007fa200000)
from space 1024K, 0% used [0x00000007fa200000, 0x00000007fa200000, 0x00000007fa300000)
to space 1024K, 0% used [0x00000007fa300000, 0x00000007fa300000, 0x00000007fa400000)
concurrent mark-sweep generation total 10240K, used 9797K [0x00000007fa400000, 0x00000007fae00000, 0x00000007fae00000)
concurrent-mark-sweep perm gen total 21248K, used 3630K [0x00000007fae00000, 0x00000007fc2c0000, 0x0000000800000000)
}
2014-08-14T18:57:31.486-0800: [GC [1 CMS-initial-mark: 9797K(10240K)] 16006K(19456K), 0.0002190 secs] [Times: user=0.00 sys=0.00, real=0.00 secs]
*********************4444Thu Aug 14 18:57:31 CST 2014
Heap
par new generation total 9216K, used 7397K [0x00000007f9a00000, 0x00000007fa400000, 0x00000007fa400000)
eden space 8192K, 90% used [0x00000007f9a00000, 0x00000007fa1396a8, 0x00000007fa200000)
from space 1024K, 0% used [0x00000007fa200000, 0x00000007fa200000, 0x00000007fa300000)
to space 1024K, 0% used [0x00000007fa300000, 0x00000007fa300000, 0x00000007fa400000)
concurrent mark-sweep generation total 10240K, used 9797K [0x00000007fa400000, 0x00000007fae00000, 0x00000007fae00000)
concurrent-mark-sweep perm gen total 21248K, used 3638K [0x00000007fae00000, 0x00000007fc2c0000, 0x0000000800000000)
这个日志感觉很奇怪。
1、第一次CMS GC发生时，老年代并没有达到设定的80%，这是为什么啊？
2、promotion failed 发生的条件是什么，看gc日志，发生promotion failed时，老年代还有将近7M空间可用，也不够放青年代的8M对象啊，为什么就发生了promotion failed呢，我的理解应该是老年代空间够用，但是有碎片的时候会发生promotion failed？而且full gc后，只是一部分青年代的对象晋级老年代了，另一部分还是留在了青年代，这是什么原理？发生promotion failed时，会触发full gc， 老年代使用什么收集算法，CMS还是serial MSC？
3、from 区里面存放的是什么？因为我构建的时候N兆的大对象，怎么会有几百K的对象进入from区呢？
4、关于CMSFullGCsBeforeCompaction这个参数，我看R大在之前的帖（http://hllvm.group.iteye.com/group/topic/28854）中解释过，我在源码中找答案时，发现_full_gcs_since_conc_gc这个参数也是在CMSCollector::collect方法中自增的，所以发生concurrent mode failure时，还是使用的CMS算法收集老年代，而不是之前说的Serial old?
烦请哪些大牛给解答下，不胜感激！！
