1.目的
排查 java 程序出现内存一直飙高，或者出现 OutOfMemoryError 的问题
2.查看 jvm 内存
2.1 jps 查看进程号
zxm@zxm-pc:~$ jps -l
2643 com.intellij.idea.Main
20068 org.jetbrains.jps.cmdline.Launcher
19990 org.jetbrains.idea.maven.server.RemoteMavenServer36
20231 sun.tools.jps.Jps
20079 org.example.DemoApplication2.2 查看堆内存和对象大小
jmap -heap <pid>：输出堆内存设置和使用情况jmap -histo <pid>：输出heap的直方图，包括类名，对象数量，对象占用大小- arthas 的 memory 命令也查看堆内存情况
jmap -heap
zxm@zxm-pc:~$ jmap -heap 20079
Attaching to process ID 20079, please wait...
Debugger attached successfully.
Server compiler detected.
JVM version is 25.341-b10
using thread-local object allocation.
Parallel GC with 8 thread(s)
Heap Configuration:
MinHeapFreeRatio = 0
MaxHeapFreeRatio = 100
MaxHeapSize = 134217728 (128.0MB)
NewSize = 67108864 (64.0MB)
MaxNewSize = 67108864 (64.0MB)
OldSize = 67108864 (64.0MB)
NewRatio = 2
SurvivorRatio = 8
MetaspaceSize = 67108864 (64.0MB)
CompressedClassSpaceSize = 58720256 (56.0MB)
MaxMetaspaceSize = 67108864 (64.0MB)
G1HeapRegionSize = 0 (0.0MB)
Heap Usage:
PS Young Generation
Eden Space:
capacity = 23068672 (22.0MB)
used = 18431480 (17.57762908935547MB)
free = 4637192 (4.422370910644531MB)
79.89831404252486% used
From Space:
capacity = 22020096 (21.0MB)
used = 0 (0.0MB)
free = 22020096 (21.0MB)
0.0% used
To Space:
capacity = 22020096 (21.0MB)
used = 0 (0.0MB)
free = 22020096 (21.0MB)
0.0% used
PS Old Generation
capacity = 67108864 (64.0MB)
used = 61432632 (58.58672332763672MB)
free = 5676232 (5.413276672363281MB)
91.54175519943237% used
13506 interned Strings occupying 1166688 bytes.jmap -histo
zxm@zxm-pc:~$ jmap -histo 20079
num #instances #bytes class name
----------------------------------------------
1: 2097152 50331648 org.example.bean.Person
2: 6970 18589368 [Ljava.lang.Object;
3: 33332 3143664 [C
4: 3410 990312 [I
5: 33109 794616 java.lang.String
6: 7016 777240 java.lang.Class
7: 7817 687896 java.lang.reflect.Method
8: 1953 537376 [B
9: 13958 446656 java.util.concurrent.ConcurrentHashMap$Node
10: 2583 223504 [Ljava.util.HashMap$Node;
11: 6782 217024 java.util.HashMap$Node
12: 9549 215560 [Ljava.lang.Class;
13: 5172 206880 java.util.LinkedHashMap$Entry
14: 127 155568 [Ljava.util.concurrent.ConcurrentHashMap$Node;
15: 2690 150640 java.util.LinkedHashMap
16: 9117 145872 java.lang.Object
17: 1429 102888 java.lang.reflect.Field
... ... ... ...
2926: 1 16 sun.util.calendar.Gregorian
2927: 1 16 sun.util.locale.InternalLocaleBuilder$CaseInsensitiveChar
2928: 1 16 sun.util.locale.provider.AuxLocaleProviderAdapter$NullProvider
2929: 1 16 sun.util.locale.provider.CalendarDataUtility$CalendarWeekParameterGetter
2930: 1 16 sun.util.locale.provider.SPILocaleProviderAdapter
2931: 1 16 sun.util.locale.provider.TimeZoneNameUtility$TimeZoneNameGetter
2932: 1 16 sun.util.resources.LocaleData
2933: 1 16 sun.util.resources.LocaleData$LocaleDataResourceBundleControl
Total 2310431 79979448arthas 的 memory 命令也查看堆内存情况
[arthas@20079]$ memory
Memory used total max usage
heap 83M 107M 107M 78.49%
ps_eden_space 20M 22M 22M 92.13%
ps_survivor_space 0K 21504K 21504K 0.00%
ps_old_gen 63M 64M 64M 99.56%
nonheap 53M 56M 360M 14.83%
code_cache 6M 6M 240M 2.68%
metaspace 41M 43M 64M 64.66%
compressed_class_space 5M 6M 56M 9.94%
direct 16K 16K - 100.01%
mapped 0K 0K - 0.00% 2.3.查看 gc
jstat -gc <pid>：输出gc信息，包括gc次数和时间，内存使用状况（可带时间和显示条目参数）
zxm@zxm-pc:~$ jstat -gc 20079
S0C S1C S0U S1U EC EU OC OU MC MU CCSC CCSU YGC YGCT FGC FGCT GCT
21504.0 21504.0 0.0 0.0 22528.0 21104.4 65536.0 65248.6 44544.0 42249.4 6144.0 5681.1 7 0.102 10 3.768 3.869- S0C：第一个幸存区的大小
- S1C：第二个幸存区的大小
- S0U：第一个幸存区的使用大小
- S1U：第二个幸存区的使用大小
- EC：伊甸园区的大小
- EU：伊甸园区的使用大小
- OC：老年代大小
- OU：老年代使用大小
- MC：方法区大小
- MU：方法区使用大小
- CCSC:压缩类空间大小
- CCSU:压缩类空间使用大小
- YGC：年轻代垃圾回收次数
- YGCT：年轻代垃圾回收消耗时间
- FGC：老年代垃圾回收次数
- FGCT：老年代垃圾回收消耗时间
- GCT：垃圾回收消耗总时间
arthas 的 dashboard 命令可查看线程、内存、gc的情况
[arthas@20079]$ dashboard
ID NAME GROUP PRIORITY STATE %CPU DELTA_TIME TIME INTERRUPTED DAEMON
-1 GC task thread#0 (ParallelGC) - -1 - 0.0 0.000 0:3.317 false true
-1 GC task thread#5 (ParallelGC) - -1 - 0.0 0.000 0:3.307 false true
-1 GC task thread#2 (ParallelGC) - -1 - 0.0 0.000 0:3.300 false true
-1 GC task thread#3 (ParallelGC) - -1 - 0.0 0.000 0:3.299 false true
-1 GC task thread#6 (ParallelGC) - -1 - 0.0 0.000 0:3.291 false true
-1 GC task thread#4 (ParallelGC) - -1 - 0.0 0.000 0:3.270 false true
-1 GC task thread#1 (ParallelGC) - -1 - 0.0 0.000 0:3.226 false true
-1 GC task thread#7 (ParallelGC) - -1 - 0.0 0.000 0:3.224 false true
40 DestroyJavaVM main 5 RUNNABLE 0.0 0.000 0:2.474 false false
-1 C1 CompilerThread3 - -1 - 0.0 0.000 0:1.515 false true
-1 VM Periodic Task Thread - -1 - 0.0 0.000 0:1.428 false true
-1 VM Thread - -1 - 0.0 0.000 0:1.208 false true
57 arthas-NettyHttpTelnetBootstrap-3-2 system 5 RUNNABLE 0.0 0.000 0:0.223 false true
23 Catalina-utility-2 main 1 WAITING 0.0 0.000 0:0.177 false false
22 Catalina-utility-1 main 1 TIMED_WAITIN 0.0 0.000 0:0.170 false false
26 http-nio-8080-exec-1 main 5 WAITING 0.0 0.000 0:0.160 false true
36 http-nio-8080-ClientPoller main 5 RUNNABLE 0.0 0.000 0:0.114 false true
25 http-nio-8080-BlockPoller main 5 RUNNABLE 0.0 0.000 0:0.079 false true
58 arthas-command-execute system 5 TIMED_WAITIN 0.0 0.000 0:0.067 false true
-1 C2 CompilerThread1 - -1 - 0.0 0.000 0:0.052 false true
-1 C2 CompilerThread2 - -1 - 0.0 0.000 0:0.048 false true
-1 C2 CompilerThread0 - -1 - 0.0 0.000 0:0.048 false true
50 arthas-NettyHttpTelnetBootstrap-3-1 system 5 RUNNABLE 0.0 0.000 0:0.030 false true
27 http-nio-8080-exec-2 main 5 WAITING 0.0 0.000 0:0.030 false true
Memory used total max usage GC
heap 85M 107M 107M 79.88% gc.ps_scavenge.count 7
ps_eden_space 21M 22M 22M 98.87% gc.ps_scavenge.time(ms) 101
ps_survivor_space 0K 21504K 21504K 0.00% gc.ps_marksweep.count 10
ps_old_gen 63M 64M 64M 99.56% gc.ps_marksweep.time(ms) 3767
nonheap 53M 56M 360M 14.86%
code_cache 6M 6M 240M 2.71%
metaspace 41M 43M 64M 64.70%
compressed_class_space 5M 6M 56M 9.94%
direct 16K 16K - 100.01%
mapped 0K 0K - 0.00%
Runtime
os.name Linux
os.version 5.15.0-58-generic
java.version 1.8.0_341
java.home /opt/jdk1.8.0_341/jre
systemload.average 0.60
processors 8
timestamp/uptime Mon Feb 12 20:14:43 CST 2023/1661s 3.生成 dump 文件
- jvm内存溢出OutOfMemoryError自动生成dump内存快照。需配置
-XX:+HeapDumpOnOutOfMemoryError -XX:HeapDumpPath=/tmp/heapdump.hprof - 手动触发生成
jmap -dump:format=b,file=/tmp/heapdump.hprof - arthas 命令
heapdump /tmp/heapdump.hprof
4.mat 分析 dump 文件
MAT 是 Memory Analyzer Tool 的缩写，是一种快速，功能丰富的Java堆分析工具，能帮助你查找内存泄漏和减少内存消耗
什么时候会用到 MAT?
- OutOfMemoryError 的时候，触发 Full GC，但空间却回收不了，引发内存溢出
- java应用服务器系统异常，比如load负载飙高，io异常，或者线程死锁等，都可能通过分析堆中的内存对象来定位原因
4.1 查看对象直方图
- Shallow Heap: 类对象本身占用内存大小，不包含其引用的对象内存。List对象占用内存大小 4k
- Retained Heap: 对象自己占用内存 + 关联引用对象占用大小。List对象占用内存大小 4k + User对象占用内存大小 123k注：如一个ArrayList持有100,000个对象，每一个占用16 bytes，移除这些ArrayList可以释放16 x 100,000 + X，X代表ArrayList的shallow大小。相对于shallow heap，RetainedHeap可以更精确的反映一个对象实际占用的大小（因为如果该对象释放，retained heap都可以被释放）
4.2 对象引用关系
- outgoing references：对象引用的外部对象（注意不包含对象的基本类型属性。基本属性内容可在 inspector 查看）。
- incoming references：直接引用了当前对象的对象，每个对象的 incoming references 可能有 0 到多个。
4.3 发生 oom 线程栈
