有人能理解 G1 垃圾收集器的输出吗? [英] Can someone make sense of the G1 garbage collector output?
本文介绍了有人能理解 G1 垃圾收集器的输出吗?的处理方法，对大家解决问题具有一定的参考价值，需要的朋友们下面随着小编来一起学习吧！
问题描述
我正在使用以下选项运行带有 G1 垃圾收集器的 Java 程序:
I am running a Java program with the G1 garbage collector using the following options:
-XX:-UseBiasedLocking
-XX:+UnlockExperimentalVMOptions
-XX:+UseG1GC
-verbose:gc -XX:+PrintGCTimeStamps -XX:+PrintGCDetails -Xloggc:/var/tmp/gclog.out
输出看起来像这样......
The output looks like this...
44900.297: [GC pause (young)44900.386 (initial-mark), 0.08894851 secs]
: [GC concurrent-mark-start]
[Parallel Time: 83.7 ms]
[GC Worker Start Time (ms): 44900297.6 44900297.6 44900297.6 44900297.6 44900297.6 44900297.7 44900297.7 44900297.7 44900297.7 44900297.7 44900297.7 44900297.7 44900297.7
Avg: 44900297.7, Min: 44900297.6, Max: 44900297.7, Diff: 0.1]
[Update RS (ms): 23.5 24.3 25.0 25.0 23.9 24.4 25.2 24.1 25.7 24.7 24.8 24.4 24.7
Avg: 24.6, Min: 23.5, Max: 25.7, Diff: 2.1]
[Processed Buffers : 16 19 19 23 20 24 18 18 18 17 20 16 19
Sum: 247, Avg: 19, Min: 16, Max: 24, Diff: 8]
[Ext Root Scanning (ms): 2.2 2.7 2.2 2.6 3.0 3.1 2.2 1.1 2.3 3.0 2.2 2.4 2.9
Avg: 2.4, Min: 1.1, Max: 3.1, Diff: 2.0]
[Mark Stack Scanning (ms): 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
Avg: 0.0, Min: 0.0, Max: 0.0, Diff: 0.0]
[Scan RS (ms): 14.1 14.6 14.5 14.3 14.6 14.2 14.4 14.5 14.0 13.9 14.6 14.5 14.0
Avg: 14.3, Min: 13.9, Max: 14.6, Diff: 0.8]
[Object Copy (ms): 41.4 39.5 39.4 39.0 39.6 39.5 39.1 41.4 39.0 39.3 39.3 39.8 39.5
Avg: 39.7, Min: 39.0, Max: 41.4, Diff: 2.4]
[Termination (ms): 1.3 1.4 1.5 1.6 1.5 1.4 1.6 1.4 1.5 1.7 1.5 1.4 1.3
Avg: 1.5, Min: 1.3, Max: 1.7, Diff: 0.4]
[Termination Attempts : 1185 1205 1219 1436 1171 1231 1471 1237 1461 1526 1353 1259 1170
Sum: 16924, Avg: 1301, Min: 1170, Max: 1526, Diff: 356]
[GC Worker End Time (ms): 44900380.2 44900380.2 44900380.2 44900380.2 44900380.2 44900380.3 44900380.2 44900380.2 44900380.2 44900380.2 44900380.2 44900380.2 44900380.2
Avg: 44900380.2, Min: 44900380.2, Max: 44900380.3, Diff: 0.1]
[GC Worker Times (ms): 82.6 82.6 82.6 82.6 82.6 82.6 82.5 82.6 82.5 82.5 82.5 82.5 82.5
Avg: 82.6, Min: 82.5, Max: 82.6, Diff: 0.1]
[Other: 1.2 ms]
[Clear CT: 0.5 ms]
[Other: 4.8 ms]
[Choose CSet: 0.0 ms]
[ 1331M->830M(1840M)]
[Times: user=1.07 sys=0.01, real=0.09 secs]
44901.205: [GC concurrent-mark-end, 0.8186002 sec]
44901.205: [GC remark, 0.0258621 secs]
[Times: user=0.02 sys=0.00, real=0.03 secs]
44901.231: [GC concurrent-count-start]
44901.479: [GC concurrent-count-end, 0.2478477]
44901.479: [GC cleanup 940M->931M(1840M), 0.0073079 secs]
任何人都可以理解发生了什么吗?
Can anyone make sense of what is going on?
推荐答案
免责声明
我并不精通垃圾优先垃圾收集器和这个问题激发了我第一次检查它.有机会我的某些信息可能有误.
Disclaimer
I am in no way well versed on the Garbage-First Garbage Collector and this question inspired me to check it out for the first time. There is a chance I may be wrong with some of my information.
要获取有关垃圾优先垃圾收集器 (G1GC) 的更多信息，请访问Garbage-First Garbage Collection 论文(这里,这里，这里 和 这里) 是一个宝贵的资源.您可以找到 G1GC 的介绍这里.HotSpot Glossary 派上用场了理解 JVM 术语.以下论文也有助于进一步了解垃圾收集:
To get more information about the Garbage-First Garbage Collector (G1GC), the Garbage-First Garbage Collection paper (here, here, here and here) is an invaluable resource. You can find an introduction to the G1GC here. The HotSpot Glossary comes in handy to understand JVM terms. The following papers were also helpful in further understanding Garbage Collection:
- A Hard Look at Hard Real-Time Garbage Collection: here and here
- Non-Blocking Garbage Collection for Multiprocessors: here and here
- Non-blocking Root Scanning for Real-Time Garbage Collection: here
- Non-blocking Object Copy for Real-Time Garbage Collection: here and here
有了这些资源和一个 OpenJDK 7 调试 构建，你可以开始了解G1GC日志了.
With these resources and an OpenJDK 7 debug build, you can start to understand the G1GC log.
使用上面引用的论文和网页，这里有一些有用的最常出现的术语的定义:
Using the papers and web pages referenced above, here are some useful definitions of terms that came up most often:
- 并发标记:两者都提供集合完整性"和通过压实疏散识别成熟的回收区域.提供收集器的完整性，而不对区域选择强加任何顺序收藏集.提供实时数据信息，允许区域被以垃圾优先"的顺序收集.
- 堆:JVM用于动态内存分配的内存区域.
- 堆区域:将堆划分为一组大小相等的堆地区.垃圾优先堆被分成大小相等的堆区域，每个连续的虚拟内存范围.
- 标记位图:每个地址包含一个位，可以作为开始一个对象.
- remembered set:一种数据结构，表示堆位置在可能包含指向该区域中对象的指针的区域.每个地区都有一个关联记忆集，表示所有可能包含的位置指向区域内(活动)对象的指针.垃圾优先记忆集记录来自所有区域的指针(有一些例外).当前缓冲区或修改后的卡片序列.记录指针之间的数据结构几代人.
- 根集:一组已知可直接访问的对象.这可以到达所有活动对象的位置.
- concurrent marking: Both provides collection "completeness" and identifies regions ripe for reclamation via compacting evacuation. Provides collector completeness without imposing any order on region choice for collection sets. Provides the live data information that allows regions to be collected in "garbage-first" order.
- heap: The area of memory used by the JVM for dynamic memory allocation.
- heap region: The heap is partitioned into a set of equal-sized heap regions. The Garbage-First heap is divided into equal-sized heap regions, each a contiguous range of virtual memory.
- marking bitmap: Contains one bit for each address that can be the start of an object.
- remembered set: A data structure that indicates heap locations outside the region that may contain pointers to objects in the region. Each region has an associated remembered set, which indicates all locations that might contain pointers to (live) objects within the region. Garbage-First remembered sets record pointers from all regions (with some exceptions). A current buffer or sequence of modified cards. A data structure that records pointers between generations.
- root set: A set of objects which is known to be directly accessible. The locations from which all live objects are reachable.
为了更好地理解 G1GC 日志，我使用了以下 OpenJDK 7源 文件:
To understand the G1GC log better, I used the following OpenJDK 7 source files:
- hotspot/src/share/vm/gc_implementation/g1/concurrentMarkThread.cpp
- hotspot/src/share/vm/gc_implementation/g1/concurrentMark.cpp
- hotspot/src/share/vm/gc_implementation/g1/concurrentMark.hpp
- hotspot/src/share/vm/gc_implementation/g1/g1_specialized_oop_closures.hpp
- hotspot/src/share/vm/gc_implementation/g1/g1CollectedHeap.cpp
- hotspot/src/share/vm/gc_implementation/g1/g1CollectedHeap.hpp
- hotspot/src/share/vm/gc_implementation/g1/g1CollectorPolicy.cpp
- hotspot/src/share/vm/gc_implementation/g1/g1CollectorPolicy.hpp
- hotspot/src/share/vm/gc_implementation/g1/g1OopClosures.hpp
- hotspot/src/share/vm/gc_implementation/g1/g1OopClosures.inline.hpp
- hotspot/src/share/vm/gc_implementation/g1/g1RemSet.cpp
- hotspot/src/share/vm/memory/sharedHeap.cpp
- hotspot/src/share/vm/memory/sharedHeap.hpp
- hotspot/src/share/vm/utilities/taskqueue.hpp
- hotspot/src/share/vm/runtime/timer.cpp
- hotspot/src/share/vm/runtime/timer.hpp
- hotspot/src/share/vm/gc_implementation/g1/vm_operations_g1.cpp
- hotspot/src/share/vm/gc_implementation/g1/vm_operations_g1.hpp
以下是相关 G1GC 日志的注释副本.
Below is an annotated copy of the G1GC log in question.
44900.297: [GC pause (young) (initial-mark), 0.08894851 secs]
^^^^^^^^^^^ Elapsed seconds from JVM start
^^^^^^^^^ Collection pause at safe-point
^^^^^^^^ In full young GC mode
^^^^^^^^^^^^^^^ Last pause included initial mark
^^^^^^^^^^^^^^^^^^ Elapsed seconds in method
44900.386: [GC concurrent-mark-start]
^^^^^^^^^^^ Elapsed seconds from JVM start
^^^^^^^^^^^^^^^^^^^^^^^^^^ Concurrent mark thread started
[Parallel Time: 83.7 ms]
^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Elapsed milliseconds for all GC worker threads to finish
[GC Worker Start Time (ms): 44900297.6 44900297.6 44900297.6 44900297.6 44900297.6 44900297.7 44900297.7 44900297.7 44900297.7 44900297.7 44900297.7 44900297.7 44900297.7
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^... Elapsed seconds from JVM start that GC worker threads were started
Avg: 44900297.7, Min: 44900297.6, Max: 44900297.7, Diff: 0.1]
^^^^^^^^^^^^^^^^^^^^^^ Average GC worker thread start time (elapsed seconds from JVM start)
^^^^^^^^^^^^^^^^^ Minimum GC worker thread start time (elapsed seconds from JVM start)
^^^^^^^^^^^^^^^^^ Maximum GC worker thread start time (elapsed seconds from JVM start)
^^^^^^^^^^^^^^ Total seconds to start all GC worker threads
[Update RS (ms): 23.5 24.3 25.0 25.0 23.9 24.4 25.2 24.1 25.7 24.7 24.8 24.4 24.7
^^^^^^^^^^^^^^^^^^^^^^^^... Elapsed milliseconds for each GC worker thread to update cards in remembered sets during an evacuation pause
Avg: 24.6, Min: 23.5, Max: 25.7, Diff: 2.1]
^^^^^^^^^^^^^^^^^ Average GC worker thread milliseconds to update RS
^^^^^^^^^^^^ Minimum GC worker thread milliseconds to update RS
^^^^^^^^^^^^ Maximum GC worker thread milliseconds to update RS
^^^^^^^^^^^^^^ Minimum/maximum delta of GC worker thread milliseconds to update RS
[Processed Buffers : 16 19 19 23 20 24 18 18 18 17 20 16 19
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^... Total remembered set buffers processed by each GC worker thread
Sum: 247, Avg: 19, Min: 16, Max: 24, Diff: 8]
^^^^^^^^^^... Summary information for total remembered set buffers processed by all GC worker thread
[Ext Root Scanning (ms): 2.2 2.7 2.2 2.6 3.0 3.1 2.2 1.1 2.3 3.0 2.2 2.4 2.9
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^... Elapsed milliseconds for each GC worker thread to process heap roots during an evacuation pause
Avg: 2.4, Min: 1.1, Max: 3.1, Diff: 2.0]
^^^^^^^... Summary information for total elapsed milliseconds for all GC worker thread to process heap roots during an evacuation pause
[Mark Stack Scanning (ms): 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^... Elapsed milliseconds for each GC worker thread to scan strong roots in the mark stack during an evacuation pause
Avg: 0.0, Min: 0.0, Max: 0.0, Diff: 0.0]
^^^^^^^... Summary information for total milliseconds for all GC worker thread to scan strong roots in the mark stack during an evacuation pause
[Scan RS (ms): 14.1 14.6 14.5 14.3 14.6 14.2 14.4 14.5 14.0 13.9 14.6 14.5 14.0
^^^^^^^^^^^^^^^^^^^^^^... Elapsed milliseconds for each GC worker thread to scan for dirty cards in a heap region to update the remembered set
Avg: 14.3, Min: 13.9, Max: 14.6, Diff: 0.8]
^^^^^^^^^^^^^... Summary information for total elapsed milliseconds for all GC worker thread to scan for dirty cards in a heap region to update the remembered set
[Object Copy (ms): 41.4 39.5 39.4 39.0 39.6 39.5 39.1 41.4 39.0 39.3 39.3 39.8 39.5
^^^^^^^^^^^^^^^^^^^^^^^^^^... Elapsed milliseconds for each GC worker thread to scan root sets for pointers to include in this heap region's remembered set during an evacuation pause
Avg: 39.7, Min: 39.0, Max: 41.4, Diff: 2.4]
^^^^^^^^^^^^^... Summary information for total elapsed milliseconds for all GC worker thread to scan root sets for pointers to include in this heap region's remembered set during an evacuation pause
[Termination (ms): 1.3 1.4 1.5 1.6 1.5 1.4 1.6 1.4 1.5 1.7 1.5 1.4 1.3
^^^^^^^^^^^^^^^^^^^^^^^^^^... Elapsed milliseconds for each GC worker thread to terminate scanning root sets for pointers to include in this heap region's remembered set during an evacuation pause
Avg: 1.5, Min: 1.3, Max: 1.7, Diff: 0.4]
^^^^^^^^^^^^^^... Summary information for total elapsed milliseconds for all GC worker thread to terminate scanning root sets for pointers to include in this heap region's remembered set during an evacuation pause
[Termination Attempts : 1185 1205 1219 1436 1171 1231 1471 1237 1461 1526 1353 1259 1170
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^... Total attempts for each GC worker thread to terminate scanning root sets for pointers to include in this heap region's remembered set during an evacuation pause
Sum: 16924, Avg: 1301, Min: 1170, Max: 1526, Diff: 356]
^^^^^^^^^^^^^^^... Summary information for total attempts for all GC worker thread to terminate scanning root sets for pointers to include in this heap region's remembered set during an evacuation pause
[GC Worker End Time (ms): 44900380.2 44900380.2 44900380.2 44900380.2 44900380.2 44900380.3 44900380.2 44900380.2 44900380.2 44900380.2 44900380.2 44900380.2 44900380.2
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^... Elapsed seconds from JVM start that GC worker threads ended
Avg: 44900380.2, Min: 44900380.2, Max: 44900380.3, Diff: 0.1]
^^^^^^^^^^^^... Summary information for total elapsed seconds from JVM start that GC worker threads ended for all GC worker threads
[GC Worker Times (ms): 82.6 82.6 82.6 82.6 82.6 82.6 82.5 82.6 82.5 82.5 82.5 82.5 82.5
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^... Elapsed milliseconds for each GC worker threads
Avg: 82.6, Min: 82.5, Max: 82.6, Diff: 0.1]
^^^^^^^^^^^^^... Summary information for total elapsed milliseconds for all GC worker threads
[Other: 1.2 ms]
^^^^^^^^^^^^^^^^^^^^^^^ Other milliseconds during GC parallel processing; basically, total milliseconds minus: update RS, ext root scanning, mark stack scanning, scan RS, object copy, termination
[Clear CT: 0.5 ms]
^^^^^^^^^^^^^^^^^^^^^^^ Total milliseconds cleaning up the dirty card tables list
[Other: 4.8 ms]
^^^^^^^^^^^^^^^^^^^^ Total milliseconds in this GC pause not spent in parallel and clear CT
[Choose CSet: 0.0 ms]
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Total milliseconds to choose young heap regions to add to the collection set
[ 1331M->830M(1840M)]
^^^^^ Heap size change for this garbage collector
^^^^^ Total heap size before this collection pause
^^^^ Total heap size after this collection pause
^^^^^ Total heap size capacity for this garbage collector
[Times: user=1.07 sys=0.01, real=0.09 secs]
^^^^^^^^ Execution seconds (uses TraceCPUTime class; constructor initializes times to zero, destructor prints times)
^^^^^^^^^^ User execution seconds for all threads
^^^^^^^^^ System execution seconds
^^^^^^^^^^^ Real (wall clock) execution seconds
44901.205: [GC concurrent-mark-end, 0.8186002 sec]
^^^^^^^^^^^ Elapsed seconds from JVM start
^^^^^^^^^^^^^^^^^^^^^^^^ Concurrent mark thread ended
^^^^^^^^^^^^^ Total seconds executing concurrent mark thread
44901.205: [GC remark, 0.0258621 secs]
^^^^^^^^^^^ Elapsed seconds from JVM start
^^^^^^^^^^ Concurrent re-marking of all heap roots, final work
[Times: user=0.02 sys=0.00, real=0.03 secs]
^^^^^^^^ Execution seconds (uses TraceCPUTime class; constructor initializes times to zero, destructor prints times)
^^^^^^^^^^ User execution seconds for all threads
^^^^^^^^^ System execution seconds
^^^^^^^^^^^ Real (wall clock) execution seconds
44901.231: [GC concurrent-count-start]
^^^^^^^^^^^ Elapsed seconds from JVM start
^^^^^^^^^^^^^^^^^^^^^^^^^^^ Start concurrent marking of live objects
44901.479: [GC concurrent-count-end, 0.2478477]
^^^^^^^^^^^ Elapsed seconds from JVM start
^^^^^^^^^^^^^^^^^^^^^^^^^^^ End concurrent marking of live objects
^^^^^^^^^^^^ Total seconds to do concurrent marking of live objects
44901.479: [GC cleanup 940M->931M(1840M), 0.0073079 secs]
^^^^^^^^^^^ Elapsed seconds from JVM start
^^^^^^^^^^^ Concurrent marking cleanup (world is stopped at this checkpoint)
^... Heap size change for this garbage collector
^^^^ Total heap size before this operation
^^^^ Total heap size after this operation
^^^^^ Total heap size capacity for this garbage collector
^^^^^^^^^^^^^^ Total seconds executing concurrent marking cleanup
OpenJDK 7 G1GC
如果可以构建调试副本 的 OpenJDK 7，您将拥有更多可供您使用的 JVM 选项将为您提供有关以下方面的更多信息G1GC.使用以下命令获取所有 JVM 选项的列表:
OpenJDK 7 G1GC
If you can build a debug copy of OpenJDK 7, you'll have a few more JVM options available to you that will give you even further information about G1GC. Use the following command to get a list of all JVM options:
java -XX:+AggressiveOpts -XX:+UnlockDiagnosticVMOptions -XX:+PrintFlagsFinal -XX:+PrintFlagsWithComments -version
输出显示以下可用的 G1GC 选项；这不是全部包含列表，只是我认为有趣的一些.
The output shows the following G1GC options available; this is not an all inclusive list, just some I thought were intersting.
intx G1CardCountCacheExpandThreshold = 16 Expand the card count cache if the number of collisions for a particular entry exceeds this value.
uintx G1ConcMarkForceOverflow = 0 The number of times we'll force an overflow during concurrent marking
double G1ConcMarkStepDurationMillis = 10.000000 {product} Target duration of individual concurrent marking steps in milliseconds.
intx G1ConcRSHotCardLimit = 4 The threshold that defines (>=) a hot card.
intx G1ConcRSLogCacheSize = 10 Log base 2 of the length of conc RS hot-card cache.
bool G1ConcRegionFreeingVerbose = false Enables verboseness during concurrent region freeing
intx G1ConfidencePercent = 50 {product} Confidence level for MMU/pause predictions
bool G1DeferredRSUpdate = true If true, use deferred RS updates
bool G1FixedEdenSize = false When set, G1 will not allocate unused survivor space regions
uintx G1FixedSurvivorSpaceSize = 0 If non-0 is the size of the G1 survivor space, otherwise SurvivorRatio is used to determine the size
bool G1FixedTenuringThreshold = false When set, G1 will not adjust the tenuring threshold
bool G1Gen = true If true, it will enable the generational G1
uintx G1HeapRegionSize = 0 {product} Size of the G1 regions.
intx G1InitYoungSurvRatio = 50 Expected Survival Rate for newly allocated bytes
intx G1MarkRegionStackSize = 1048576 {product} Size of the region stack for concurrent marking.
intx G1MarkingOverheadPercent = 0 Overhead of concurrent marking
intx G1MarkingVerboseLevel = 0 Level (0-4) of verboseness of the marking code
intx G1MaxHotCardCountSizePercent = 25 The maximum size of the hot card count cache as a percentage of the number of cards for the maximum heap.
intx G1MaxVerifyFailures = -1 The maximum number of verification failrues to print. -1 means print all.
intx G1PausesBtwnConcMark = -1 If positive, fixed number of pauses between conc markings
intx G1PolicyVerbose = 0 The verbosity level on G1 policy decisions
bool G1PrintCTFilterStats = false If true, print stats on RS filtering effectiveness
bool G1PrintHeapRegions = false {diagnostic} If set G1 will print information on which regions are being allocated and which are reclaimed.
bool G1PrintOopAppls = false When true, print applications of closures to external locs.
bool G1PrintParCleanupStats = false When true, print extra stats about parallel cleanup.
bool G1PrintReachableAtInitialMark = false Reachable object dump at the initial mark pause
ccstr G1PrintReachableBaseFile = The base file name for the reachable object dumps
bool G1PrintRegionLivenessInfo = false {product} Prints the liveness information for all regions in the heap at the end of a marking cycle.
bool G1RSCountHisto = false If true, print a histogram of RS occupancies after each pause
bool G1RSLogCheckCardTable = false If true, verify that no dirty cards remain after RS log processing.
bool G1RSScrubVerbose = false When true, do RS scrubbing with verbose output.
intx G1RSetRegionEntries = 0 {product} Max number of regions for which we keep bitmaps.Will be set ergonomically by default
intx G1RSetRegionEntriesBase = 256 Max number of regions in a fine-grain table per MB.
uintx G1RSetScanBlockSize = 64 {product} Size of a work unit of cards claimed by a worker threadduring RSet scanning.
intx G1RSetSparseRegionEntries = 0 {product} Max number of entries per region in a sparse table.Will be set ergonomically by default.
intx G1RSetSparseRegionEntriesBase = 4 Max number of entries per region in a sparse table per MB.
intx G1RSetUpdatingPauseTimePercent = 10 {product} A target percentage of time that is allowed to be spend on process RS update buffers during the collection pause.
bool G1RecordHRRSEvents = false When true, record recent calls to rem set operations.
bool G1RecordHRRSOops = false When true, record recent calls to rem set operations.
intx G1RefProcDrainInterval = 10 {product} The number of discovered reference objects to process before draining concurrent marking work queues.
intx G1ReservePercent = 10 {product} It determines the minimum reserve we should have in the heap to minimize the probability of promotion failure.
uintx G1SATBBufferEnqueueingThresholdPercent = 60 {product} Before enqueueing them, each mutator thread tries to do some filtering on the SATB buffers it generates. If post-filtering the percentage of retained entries is over this threshold the buffer will be enqueued for processing. A value of 0 specifies that mutator threads should not do such filtering.
intx G1SATBBufferSize = 1024 {product} Number of entries in an SATB log buffer.
bool G1SATBPrintStubs = false If true, print generated stubs for the SATB barrier
intx G1SATBProcessCompletedThreshold = 20 Number of completed buffers that triggers log processing.
bool G1ScrubRemSets = true When true, do RS scrubbing after cleanup.
uintx G1SecondaryFreeListAppendLength = 5 The number of regions we will add to the secondary free list at every append operation
bool G1StressConcRegionFreeing = false It stresses the concurrent region freeing operation
uintx G1StressConcRegionFreeingDelayMillis = 0 Artificial delay during concurrent region freeing
bool G1SummarizeConcMark = false {diagnostic} Summarize concurrent mark info
bool G1SummarizeRSetStats = false {diagnostic} Summarize remembered set processing info
intx G1SummarizeRSetStatsPeriod = 0 {diagnostic} The period (in number of GCs) at which we will generate update buffer processing info (0 means do not periodically generate this info); it also requires -XX:+G1SummarizeRSetStats
bool G1TraceConcRefinement = false {diagnostic} Trace G1 concurrent refinement
bool G1TraceMarkStackOverflow = false If true, extra debugging code for CM restart for ovflw.
intx G1UpdateBufferSize = 256 {product} Size of an update buffer
bool G1VerifyDuringGCPrintReachable = false If conc mark verification fails, dump reachable objects
intx G1YoungSurvRateNumRegionsSummary = 0 the number of regions for which we'll print a surv rate summary.
bool G1YoungSurvRateVerbose = false print out the survival rate of young regions according to age.
查看使用 OpenJDK 7 调试提供了什么样的 G1GC 信息构建，我使用了以下测试程序:
To see what kind of G1GC information is provided using the OpenJDK 7 debug build, I used the following test program:
import java.lang.ref.WeakReference;
import java.util.ArrayList;
import java.util.Iterator;
import java.util.List;
import java.util.Random;
public class G1GCTest implements Runnable {
private int iterations = 0;
private final Random rnd = new Random();
private final List<Object> young = new ArrayList<Object>(1000);
private final List<WeakReference<Object>> old =
new ArrayList<WeakReference<Object>>(100000);
private void clearOld() {
int clearedCnt = 0;
for (final Iterator<WeakReference<Object>> iter = old.iterator(); iter
.hasNext();)
{
final WeakReference<Object> ref = iter.next();
if (null == ref.get()) {
iter.remove();
clearedCnt++;
}
}
if (0 < clearedCnt) {
System.out.println("Cleared " + clearedCnt
+ " weak references to old objects.");
}
}
public void run() {
if (0 == ++iterations % 10000) {
System.out.println("iterations=" + iterations + ", young.size()="
+ young.size() + ", old.size()=" + old.size());
clearOld();
try {
Thread.sleep(100);
} catch (final Throwable e) {
// Do nothing!
}
}
if (rnd.nextBoolean()) {
young.add(new byte[1000]);
}
if (rnd.nextBoolean() && !young.isEmpty()) {
final int nextInt = Math.abs(rnd.nextInt());
final int idx = nextInt % young.size();
final Object obj = young.remove(idx);
old.add(new WeakReference<Object>(obj));
}
}
public static void main(final String[] args) {
final G1GCTest t = new G1GCTest();
while (1000000 > t.iterations) {
t.run();
}
}
}
并使用:
java -Xms8m -Xmx8m -XX:+UnlockExperimentalVMOptions -XX:+UseG1GC -XX:+PrintGC -XX:+PrintGCTimeStamps -XX:+PrintGCDetails -XX:+G1SummarizeConcMark -XX:+G1SummarizeRSetStats -XX:+G1YoungSurvRateVerbose -XX:G1PolicyVerbose=2 -verbose:gc G1GCTest
其输出非常冗长，但提供了大量信息如果您正在进行一些 GC 调整，可能会很有用.
The output of which was very verbose, but provides a slew of information that could prove useful if you're doing some GC tuning.
这篇关于有人能理解 G1 垃圾收集器的输出吗?的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！
查看全文
