为什么当我提高记忆时我会更频繁地获得GC？

Question

我有一个关于g1gc的问题。

​

​

​

​

这些是堆使用图。

以上是-Xms4g -Xmx4g。

底部是-Xms8g -Xmx8g。

我不知道为什么8g选项会导致g1gc发生得更频繁。其他选项都是默认的。

而服务器规范是40个逻辑进程。

ps。正确的调优选项是什么？

附加问题

内存分配是否更快，因为内存大小->越大，区域大小越大？

gc.log

4G gc.log

2019-05-07T21:03:42.093+0900: 10.280: [GC pause (G1 Evacuation Pause) (young), 0.1785373 secs]
   [Parallel Time: 43.4 ms, GC Workers: 28]
      [GC Worker Start (ms): Min: 10280.0, Avg: 10280.1, Max: 10280.6, Diff: 0.6]
      [Ext Root Scanning (ms): Min: 0.0, Avg: 0.4, Max: 0.8, Diff: 0.8, Sum: 12.0]
      [Update RS (ms): Min: 0.8, Avg: 1.1, Max: 1.6, Diff: 0.8, Sum: 31.5]
         [Processed Buffers: Min: 0, Avg: 2.0, Max: 3, Diff: 3, Sum: 56]
      [Scan RS (ms): Min: 0.0, Avg: 0.4, Max: 0.7, Diff: 0.7, Sum: 10.9]
      [Code Root Scanning (ms): Min: 0.0, Avg: 0.0, Max: 0.0, Diff: 0.0, Sum: 0.0]
      [Object Copy (ms): Min: 38.0, Avg: 38.5, Max: 39.9, Diff: 1.9, Sum: 1079.0]
      [Termination (ms): Min: 1.3, Avg: 2.6, Max: 3.2, Diff: 1.9, Sum: 74.1]
         [Termination Attempts: Min: 413, Avg: 769.6, Max: 855, Diff: 442, Sum: 21549]
      [GC Worker Other (ms): Min: 0.0, Avg: 0.1, Max: 0.2, Diff: 0.1, Sum: 2.0]
      [GC Worker Total (ms): Min: 42.7, Avg: 43.2, Max: 43.4, Diff: 0.7, Sum: 1209.5]
      [GC Worker End (ms): Min: 10323.3, Avg: 10323.3, Max: 10323.4, Diff: 0.1]
   [Code Root Fixup: 0.0 ms]
   [Code Root Purge: 0.0 ms]
   [Clear CT: 0.4 ms]
   [Other: 134.7 ms]
      [Choose CSet: 0.0 ms]
      [Ref Proc: 132.4 ms]
      [Ref Enq: 0.9 ms]
      [Redirty Cards: 0.3 ms]
      [Humongous Register: 0.1 ms]
      [Humongous Reclaim: 0.0 ms]
      [Free CSet: 0.7 ms]
   [Eden: 928.0M(928.0M)->0.0B(828.0M) Survivors: 26.0M->120.0M Heap: 1193.0M(4096.0M)->409.0M(4096.0M)]
Heap after GC invocations=8 (full 0):
 garbage-first heap   total 4194304K, used 418816K [0x00000006c0000000, 0x00000006c0204000, 0x00000007c0000000)
  region size 2048K, 60 young (122880K), 60 survivors (122880K)
 Metaspace       used 28525K, capacity 30824K, committed 31104K, reserved 1077248K
  class space    used 3583K, capacity 4166K, committed 4224K, reserved 1048576K
}
 [Times: user=1.21 sys=0.08, real=0.18 secs]
{Heap before GC invocations=8 (full 0):
 garbage-first heap   total 4194304K, used 744448K [0x00000006c0000000, 0x00000006c0204000, 0x00000007c0000000)
  region size 2048K, 219 young (448512K), 60 survivors (122880K)
 Metaspace       used 28525K, capacity 30824K, committed 31104K, reserved 1077248K
  class space    used 3583K, capacity 4166K, committed 4224K, reserved 1048576K
2019-05-07T21:03:42.895+0900: 11.082: [GC pause (G1 Evacuation Pause) (young), 0.0505563 secs]
   [Parallel Time: 11.6 ms, GC Workers: 28]
      [GC Worker Start (ms): Min: 11082.3, Avg: 11082.6, Max: 11083.6, Diff: 1.3]
      [Ext Root Scanning (ms): Min: 0.0, Avg: 0.4, Max: 0.8, Diff: 0.8, Sum: 9.9]
      [Update RS (ms): Min: 0.4, Avg: 1.0, Max: 1.5, Diff: 1.1, Sum: 29.4]
         [Processed Buffers: Min: 1, Avg: 1.8, Max: 3, Diff: 2, Sum: 50]
      [Scan RS (ms): Min: 0.8, Avg: 1.2, Max: 1.4, Diff: 0.6, Sum: 32.4]
      [Code Root Scanning (ms): Min: 0.0, Avg: 0.0, Max: 0.0, Diff: 0.0, Sum: 0.0]
      [Object Copy (ms): Min: 8.3, Avg: 8.4, Max: 8.6, Diff: 0.2, Sum: 236.3]
      [Termination (ms): Min: 0.0, Avg: 0.1, Max: 0.1, Diff: 0.1, Sum: 2.8]
         [Termination Attempts: Min: 1, Avg: 42.7, Max: 52, Diff: 51, Sum: 1195]
      [GC Worker Other (ms): Min: 0.0, Avg: 0.1, Max: 0.2, Diff: 0.1, Sum: 2.0]
      [GC Worker Total (ms): Min: 10.2, Avg: 11.2, Max: 11.5, Diff: 1.3, Sum: 312.9]
      [GC Worker End (ms): Min: 11093.7, Avg: 11093.8, Max: 11093.8, Diff: 0.1]
   [Code Root Fixup: 0.0 ms]
   [Code Root Purge: 0.0 ms]
   [Clear CT: 0.3 ms]
   [Other: 38.6 ms]
      [Choose CSet: 0.0 ms]
      [Ref Proc: 37.0 ms]
      [Ref Enq: 0.5 ms]
      [Redirty Cards: 0.3 ms]
      [Humongous Register: 0.1 ms]
      [Humongous Reclaim: 0.0 ms]
      [Free CSet: 0.5 ms]
   [Eden: 318.0M(252.0M)->0.0B(1052.0M) Survivors: 120.0M->48.0M Heap: 727.0M(4096.0M)->397.0M(4096.0M)]
Heap after GC invocations=9 (full 0):
 garbage-first heap   total 4194304K, used 406528K [0x00000006c0000000, 0x00000006c0204000, 0x00000007c0000000)
  region size 2048K, 24 young (49152K), 24 survivors (49152K)
 Metaspace       used 28525K, capacity 30824K, committed 31104K, reserved 1077248K
  class space    used 3583K, capacity 4166K, committed 4224K, reserved 1048576K
}
 [Times: user=0.34 sys=0.02, real=0.05 secs]
{Heap before GC invocations=9 (full 0):
 garbage-first heap   total 4194304K, used 912384K [0x00000006c0000000, 0x00000006c0204000, 0x00000007c0000000)
  region size 2048K, 271 young (555008K), 24 survivors (49152K)
 Metaspace       used 29461K, capacity 31868K, committed 32256K, reserved 1077248K
  class space    used 3681K, capacity 4237K, committed 4352K, reserved 1048576K

8G gc.log

2019-05-05T02:39:16.996+0900: 201016.724: [GC pause (G1 Evacuation Pause) (young), 0.0336675 secs]
   [Parallel Time: 12.9 ms, GC Workers: 28]
      [GC Worker Start (ms): Min: 201016724.7, Avg: 201016724.9, Max: 201016725.0, Diff: 0.2]
      [Ext Root Scanning (ms): Min: 0.8, Avg: 1.2, Max: 5.2, Diff: 4.4, Sum: 32.4]
      [Update RS (ms): Min: 0.0, Avg: 3.1, Max: 3.5, Diff: 3.5, Sum: 87.7]
         [Processed Buffers: Min: 0, Avg: 11.1, Max: 30, Diff: 30, Sum: 310]
      [Scan RS (ms): Min: 0.1, Avg: 0.3, Max: 0.3, Diff: 0.2, Sum: 7.3]
      [Code Root Scanning (ms): Min: 0.0, Avg: 0.0, Max: 0.0, Diff: 0.0, Sum: 0.1]
      [Object Copy (ms): Min: 6.9, Avg: 7.5, Max: 7.7, Diff: 0.8, Sum: 211.2]
      [Termination (ms): Min: 0.2, Avg: 0.3, Max: 0.4, Diff: 0.2, Sum: 9.0]
         [Termination Attempts: Min: 105, Avg: 124.7, Max: 146, Diff: 41, Sum: 3491]
      [GC Worker Other (ms): Min: 0.0, Avg: 0.1, Max: 0.2, Diff: 0.2, Sum: 3.2]
      [GC Worker Total (ms): Min: 12.4, Avg: 12.5, Max: 12.7, Diff: 0.4, Sum: 350.8]
      [GC Worker End (ms): Min: 201016737.3, Avg: 201016737.4, Max: 201016737.5, Diff: 0.2]
   [Code Root Fixup: 0.0 ms]
   [Code Root Purge: 0.0 ms]
   [Clear CT: 0.7 ms]
   [Other: 20.0 ms]
      [Choose CSet: 0.0 ms]
      [Ref Proc: 17.2 ms]
      [Ref Enq: 0.2 ms]
      [Redirty Cards: 0.3 ms]
      [Humongous Register: 0.1 ms]
      [Humongous Reclaim: 0.0 ms]
      [Free CSet: 1.7 ms]
   [Eden: 4864.0M(4864.0M)->0.0B(4868.0M) Survivors: 48.0M->44.0M Heap: 5968.1M(8192.0M)->1091.2M(8192.0M)]
Heap after GC invocations=19405 (full 0):
 garbage-first heap   total 8388608K, used 1117388K [0x00000005c0000000, 0x00000005c0404000, 0x00000007c0000000)
  region size 4096K, 11 young (45056K), 11 survivors (45056K)
 Metaspace       used 187853K, capacity 205120K, committed 210176K, reserved 1232896K
  class space    used 22574K, capacity 25471K, committed 26368K, reserved 1048576K
}
 [Times: user=0.39 sys=0.00, real=0.04 secs]
{Heap before GC invocations=19405 (full 0):
 garbage-first heap   total 8388608K, used 6106497K [0x00000005c0000000, 0x00000005c0404000, 0x00000007c0000000)
  region size 4096K, 1228 young (5029888K), 11 survivors (45056K)
 Metaspace       used 187853K, capacity 205120K, committed 210176K, reserved 1232896K
  class space    used 22574K, capacity 25471K, committed 26368K, reserved 1048576K
2019-05-05T02:39:33.830+0900: 201033.558: [GC pause (G1 Evacuation Pause) (young), 0.0373282 secs]
2019-05-05T02:39:33.830+0900: 201033.558: [GC pause (G1 Evacuation Pause) (young), 0.0373282 secs]
   [Parallel Time: 13.9 ms, GC Workers: 28]
      [GC Worker Start (ms): Min: 201033558.4, Avg: 201033558.5, Max: 201033558.6, Diff: 0.2]
      [Ext Root Scanning (ms): Min: 0.8, Avg: 1.2, Max: 4.5, Diff: 3.7, Sum: 32.5]
      [Update RS (ms): Min: 0.0, Avg: 2.8, Max: 3.1, Diff: 3.1, Sum: 77.4]
         [Processed Buffers: Min: 0, Avg: 10.3, Max: 31, Diff: 31, Sum: 289]
      [Scan RS (ms): Min: 0.1, Avg: 0.3, Max: 0.3, Diff: 0.3, Sum: 7.1]
      [Code Root Scanning (ms): Min: 0.0, Avg: 0.0, Max: 0.0, Diff: 0.0, Sum: 0.1]
      [Object Copy (ms): Min: 8.5, Avg: 8.8, Max: 8.9, Diff: 0.4, Sum: 246.0]
      [Termination (ms): Min: 0.3, Avg: 0.4, Max: 0.5, Diff: 0.2, Sum: 12.0]
         [Termination Attempts: Min: 135, Avg: 156.5, Max: 175, Diff: 40, Sum: 4382]
      [GC Worker Other (ms): Min: 0.0, Avg: 0.1, Max: 0.2, Diff: 0.2, Sum: 3.3]
      [GC Worker Total (ms): Min: 13.3, Avg: 13.5, Max: 13.7, Diff: 0.3, Sum: 378.4]
      [GC Worker End (ms): Min: 201033571.9, Avg: 201033572.0, Max: 201033572.1, Diff: 0.2]
   [Code Root Fixup: 0.0 ms]
   [Code Root Purge: 0.0 ms]
   [Clear CT: 0.8 ms]
   [Other: 22.6 ms]
      [Choose CSet: 0.0 ms]
      [Ref Proc: 18.5 ms]
      [Ref Enq: 0.3 ms]
      [Redirty Cards: 1.0 ms]
      [Humongous Register: 0.1 ms]
      [Humongous Reclaim: 0.0 ms]
      [Free CSet: 2.1 ms]
   [Eden: 4868.0M(4868.0M)->0.0B(4880.0M) Survivors: 44.0M->32.0M Heap: 5963.4M(8192.0M)->1082.0M(8192.0M)]
Heap after GC invocations=19406 (full 0):
 garbage-first heap   total 8388608K, used 1107927K [0x00000005c0000000, 0x00000005c0404000, 0x00000007c0000000)
  region size 4096K, 8 young (32768K), 8 survivors (32768K)
 Metaspace       used 187853K, capacity 205120K, committed 210176K, reserved 1232896K
  class space    used 22574K, capacity 25471K, committed 26368K, reserved 1048576K
}
 [Times: user=0.41 sys=0.00, real=0.04 secs]
{Heap before GC invocations=19406 (full 0):
 garbage-first heap   total 8388608K, used 6122963K [0x00000005c0000000, 0x00000005c0404000, 0x00000007c0000000)
  region size 4096K, 1228 young (5029888K), 8 survivors (32768K)
 Metaspace       used 187853K, capacity 205120K, committed 210176K, reserved 1232896K
  class space    used 22574K, capacity 25471K, committed 26368K, reserved 1048576K

Answer 1

使用G1GC建立的JVM将通过构造一个名为region的内存块来启动，而不需要区分New / Survivor / Old物理内存。逻辑上有New / Survivor / Old，但它在物理上不被地址分隔。

对象在任何区域中创建，对象的引用信息存储在一个记忆集中(在整个堆中使用5%的级别)。“记住集”是一种数据结构，它可以很容易地知道哪些区域被分配给具有引用的对象。(追踪该区域的参考资料)

如果要创建一个大于区域大小的对象，它将在多个区域上创建一个对象，这组区域称为巨型区域。此信息也存储在“记忆集”中。

区域大小可以根据堆大小从1MB到32 MB不等。下表显示了在未显式设置区域大小的情况下将根据最小堆大小选择的区域大小。

|---------------------|------------------|
|    Min Heap Size    |   Region Size    |
|---------------------|------------------|
|     heap < 4GB      |       1MB        |
|---------------------|------------------|
|  4GB <= heap < 8GB  |       2MB        |
|---------------------|------------------|
|  8GB <= heap < 16GB |       4MB        |
|---------------------|------------------|
| 16GB <= heap < 32GB |       8MB        |
|---------------------|------------------|
| 32GB <= heap < 64GB |      16MB        |
|---------------------|------------------|
|     64GB < heap     |      32MB        |
|---------------------|------------------|

因此，在您的情况下，区域的大小将被计算不同。此外，根据应用程序的不同，内存分配模式也可能有所不同。为了找到更准确的原因，我们需要一个垃圾收集日志。

您可以将InitiatingheapOccupancyPercent设置为让后台线程启动时间。堆使用率与堆大小的比率。减少值使您能够快速启动后台线程。默认值为45。但是，如果值太小，垃圾回收将运行得太频繁。它需要CPU周期，并可能影响应用程序本身的性能。

如您所知，链锯图是健康的应用程序。因此，即使不做额外的设置，也没有什么大问题。

额外:可以帮助您解决问题的Bug报告。

Bug-8151176中的描述是指为实际计算IHOP而计算旧的gen占用率/堆总大小。

这意味着对年轻一代的占领完全被忽视。也就是说，如果年轻一代的比例大于IHOP，那么并发周期就无法开始。

原因是，如果旧的根占用率超过当前堆容量的固定百分比，则会启动静态IHOP。如果用户或人类工效学决定旧代不能大于触发并发标记的堆容量的那一部分，则标记永远不会启动。

Answer 2

我想知道更频繁的垃圾收集是否真的发生了，因为您能够更快地分配内存。

如果您查看图表及其比例，第一个显示应用程序能够在大约12秒内分配大约2GB的内存，而GC没有运行。相反，第二个图表显示了大约2.5GB在大约6秒内，而GC没有运行。

如果应用程序能够以两倍以上的速度分配，这意味着生成的垃圾数量是其数量的两倍以上。

接下来，比较两个锯齿图的上斜率和下斜率.在第一种情况下，上坡不均匀，比下坡浅.在第二种情况下，斜坡更平滑，上斜坡比下坡陡.

所以有个理论..。

在第一种情况下(较小的堆)，当收集器空闲时，某些东西会更慢分配速率.或者标记。

在第二种情况下(较大的堆)，分配速度更快，这使得GC可以做更多的工作。更多的工作意味着它需要更频繁地运行。如果我们假设这种更快的分配速率也对应于应用程序线程更多的CPU利用率，并且在GC删除对象时保持了更快的分配速率+应用程序线程活动，那么这可能是从GC线程中窃取周期，从而导致回收速率变慢(c.f )。下坡)。

另一个理论(没有任何证据支持！)您的应用程序正在广泛使用软/弱引用来实现缓存。更多的内存可能意味着一个更大的缓存，它可能以“有趣”的方式改变应用程序的行为。例如，它可能导致更快的请求处理，从而加快与请求相关的短期对象的翻转。

这些理论都相当薄弱。实际上，我们需要更多关于您的应用程序的信息，以及更多的度量。

Answer 3

根据您的两个图表，看起来您的服务器在更短的时间内分配了更多的内存。正如您在自己的图片上所看到的，在第一种情况下，当内存使用量达到3G时执行混合集合，在第二种情况下，假设为5G。但这取决于应用程序的内存分配模式。G1 gc尝试在不牺牲吞吐量的情况下满足gc最大暂停(由XX:MaxGCPauseMillis=设置，默认为500 is )。