diff --git a/README.md b/README.md index 8ea08dcc..cf018670 100644 --- a/README.md +++ b/README.md @@ -104,7 +104,7 @@ JVM 这部分内容主要参考 [JVM 虚拟机规范-Java8 ](https://docs.oracle ### 小技巧 1. [JAD 反编译](docs/java/tips/JAD反编译tricks.md) -2. [手把手教你定位常见 Java 性能问题](./docs/java/tips/手把手教你定位常见Java性能问题.md) +2. [手把手教你定位常见 Java 性能问题](./docs/java/tips/locate-performance-problems/手把手教你定位常见Java性能问题.md) ## 计算机基础 diff --git a/docs/java/tips/locate-performance-problems/images/0605fbf554814a23b80f6351408598be-1.png b/docs/java/tips/locate-performance-problems/images/0605fbf554814a23b80f6351408598be-1.png new file mode 100644 index 00000000..4a2f13c6 Binary files /dev/null and b/docs/java/tips/locate-performance-problems/images/0605fbf554814a23b80f6351408598be-1.png differ diff --git a/docs/java/tips/locate-performance-problems/images/1fb751b0d78b4a3b8d0f528598ae885d-1.png b/docs/java/tips/locate-performance-problems/images/1fb751b0d78b4a3b8d0f528598ae885d-1.png new file mode 100644 index 00000000..37581b61 Binary files /dev/null and b/docs/java/tips/locate-performance-problems/images/1fb751b0d78b4a3b8d0f528598ae885d-1.png differ diff --git a/docs/java/tips/locate-performance-problems/images/392e4090c0094657ae29af030d3646e3-1.png b/docs/java/tips/locate-performance-problems/images/392e4090c0094657ae29af030d3646e3-1.png new file mode 100644 index 00000000..7ac1b402 Binary files /dev/null and b/docs/java/tips/locate-performance-problems/images/392e4090c0094657ae29af030d3646e3-1.png differ diff --git a/docs/java/tips/locate-performance-problems/images/3be5a280b0f5499a80c706c8e5da2a4f-1.png b/docs/java/tips/locate-performance-problems/images/3be5a280b0f5499a80c706c8e5da2a4f-1.png new file mode 100644 index 00000000..68c0b685 Binary files /dev/null and b/docs/java/tips/locate-performance-problems/images/3be5a280b0f5499a80c706c8e5da2a4f-1.png differ diff --git a/docs/java/tips/locate-performance-problems/images/3d8d5ffd3ada43fb86ef54b05408c656-1.png b/docs/java/tips/locate-performance-problems/images/3d8d5ffd3ada43fb86ef54b05408c656-1.png new file mode 100644 index 00000000..8c50d818 Binary files /dev/null and b/docs/java/tips/locate-performance-problems/images/3d8d5ffd3ada43fb86ef54b05408c656-1.png differ diff --git a/docs/java/tips/locate-performance-problems/images/53fd3ee9a1a0448ca1878e865f4e5f96-1.png b/docs/java/tips/locate-performance-problems/images/53fd3ee9a1a0448ca1878e865f4e5f96-1.png new file mode 100644 index 00000000..03907f53 Binary files /dev/null and b/docs/java/tips/locate-performance-problems/images/53fd3ee9a1a0448ca1878e865f4e5f96-1.png differ diff --git a/docs/java/tips/locate-performance-problems/images/ba07b0fee1754ffc943e546a18a3907e-1.png b/docs/java/tips/locate-performance-problems/images/ba07b0fee1754ffc943e546a18a3907e-1.png new file mode 100644 index 00000000..8e8648df Binary files /dev/null and b/docs/java/tips/locate-performance-problems/images/ba07b0fee1754ffc943e546a18a3907e-1.png differ diff --git a/docs/java/tips/locate-performance-problems/images/e9bf831860f442a3a992eef64ebb6a50-1.png b/docs/java/tips/locate-performance-problems/images/e9bf831860f442a3a992eef64ebb6a50-1.png new file mode 100644 index 00000000..e204e10c Binary files /dev/null and b/docs/java/tips/locate-performance-problems/images/e9bf831860f442a3a992eef64ebb6a50-1.png differ diff --git a/docs/java/tips/手把手教你定位常见Java性能问题.md b/docs/java/tips/locate-performance-problems/手把手教你定位常见Java性能问题.md similarity index 96% rename from docs/java/tips/手把手教你定位常见Java性能问题.md rename to docs/java/tips/locate-performance-problems/手把手教你定位常见Java性能问题.md index 2550d8b3..ab60c9a8 100644 --- a/docs/java/tips/手把手教你定位常见Java性能问题.md +++ b/docs/java/tips/locate-performance-problems/手把手教你定位常见Java性能问题.md @@ -1,3 +1,5 @@ +> 本文来自木木匠投稿。 + ## 手把手教你定位常见 Java 性能问题 ## 概述 @@ -127,15 +129,15 @@ MAT(Memory Analyzer Tool)工具是 eclipse 的一个插件(MAT 也可以单独 请求接口地址测试`curl localhost:8080/cpu/loop`,发现 CPU 立马飙升到 100% -![](./images/performance-tuning/java-performance1.png) +![](./images/3be5a280b0f5499a80c706c8e5da2a4f-1.png) 通过执行`top -Hp 32805` 查看 Java 线程情况 -![](./images/performance-tuning/java-performance2.png) +![](./images/3d8d5ffd3ada43fb86ef54b05408c656-1.png) 执行 `printf '%x' 32826` 获取 16 进制的线程 id，用于`dump`信息查询，结果为 `803a`。最后我们执行`jstack 32805 |grep -A 20 803a`来查看下详细的`dump`信息。 -![](./images/performance-tuning/java-performance3.png) +![](./images/1fb751b0d78b4a3b8d0f528598ae885d-1.png) 这里`dump`信息直接定位出了问题方法以及代码行，这就定位出了 CPU 占满的问题。 @@ -158,7 +160,9 @@ MAT(Memory Analyzer Tool)工具是 eclipse 的一个插件(MAT 也可以单独 我们给启动加上堆内存大小限制，同时设置内存溢出的时候输出堆栈快照并输出日志。 -`java -jar -Xms500m -Xmx500m -XX:+HeapDumpOnOutOfMemoryError -XX:HeapDumpPath=/tmp/heapdump.hprof -XX:+PrintGCTimeStamps -XX:+PrintGCDetails -Xloggc:/tmp/heaplog.log analysis-demo-0.0.1-SNAPSHOT.jar` +```bash +java -jar -Xms500m -Xmx500m -XX:+HeapDumpOnOutOfMemoryError -XX:HeapDumpPath=/tmp/heapdump.hprof -XX:+PrintGCTimeStamps -XX:+PrintGCDetails -Xloggc:/tmp/heaplog.log analysis-demo-0.0.1-SNAPSHOT.jar +``` 启动成功后我们循环执行 100 次,`for i in {1..500}; do curl localhost:8080/memory/leak;done`,还没执行完毕，系统已经返回 500 错误了。查看系统日志出现了如下异常： @@ -168,23 +172,23 @@ java.lang.OutOfMemoryError: Java heap space 我们用`jstat -gc pid` 命令来看看程序的 GC 情况。 -![](./images/performance-tuning/java-performance4.png) +![](./images/e9bf831860f442a3a992eef64ebb6a50-1.png) 很明显，内存溢出了，堆内存经过 45 次 Full Gc 之后都没释放出可用内存，这说明当前堆内存中的对象都是存活的，有 GC Roots 引用，无法回收。那是什么原因导致内存溢出呢？是不是我只要加大内存就行了呢？如果是普通的内存溢出也许扩大内存就行了，但是如果是内存泄漏的话，扩大的内存不一会就会被占满，所以我们还需要确定是不是内存泄漏。我们之前保存了堆 Dump 文件，这个时候借助我们的 MAT 工具来分析下。导入工具选择`Leak Suspects Report`，工具直接就会给你列出问题报告。 -![](./images/performance-tuning/java-performance5.png) +![](./images/392e4090c0094657ae29af030d3646e3-1.png) 这里已经列出了可疑的 4 个内存泄漏问题，我们点击其中一个查看详情。 -![](./images/performance-tuning/java-performance6.png) +![](./images/53fd3ee9a1a0448ca1878e865f4e5f96-1.png) 这里已经指出了内存被线程占用了接近 50M 的内存，占用的对象就是 ThreadLocal。如果想详细的通过手动去分析的话，可以点击`Histogram`,查看最大的对象占用是谁，然后再分析它的引用关系，即可确定是谁导致的内存溢出。 -![](./images/performance-tuning/java-performance7.png) +![](./images/ba07b0fee1754ffc943e546a18a3907e-1.png) 上图发现占用内存最大的对象是一个 Byte 数组，我们看看它到底被那个 GC Root 引用导致没有被回收。按照上图红框操作指引，结果如下图： -![](./images/performance-tuning/java-performance8.png) +![](./images/0605fbf554814a23b80f6351408598be-1.png) 我们发现 Byte 数组是被线程对象引用的，图中也标明，Byte 数组对像的 GC Root 是线程，所以它是不会被回收的，展开详细信息查看，我们发现最终的内存占用对象是被 ThreadLocal 对象占据了。这也和 MAT 工具自动帮我们分析的结果一致。