|
|
|
|
# profiler
|
|
|
|
|
|
|
|
|
|
[`profiler`在线教程](https://arthas.aliyun.com/doc/arthas-tutorials.html?language=cn&id=command-profiler)
|
|
|
|
|
|
|
|
|
|
::: tip
|
|
|
|
|
使用[async-profiler](https://github.com/jvm-profiling-tools/async-profiler)生成火焰图
|
|
|
|
|
:::
|
|
|
|
|
|
|
|
|
|
`profiler` 命令支持生成应用热点的火焰图。本质上是通过不断的采样,然后把收集到的采样结果生成火焰图。
|
|
|
|
|
|
|
|
|
|
`profiler` 命令基本运行结构是 `profiler action [actionArg]`
|
|
|
|
|
|
|
|
|
|
### 参数说明
|
|
|
|
|
|
|
|
|
|
| 参数名称 | 参数说明 |
|
|
|
|
|
| ----------: | :-------------------------------------------------------------- |
|
|
|
|
|
| _action_ | 要执行的操作 |
|
|
|
|
|
| _actionArg_ | 属性名模式 |
|
|
|
|
|
| [i:] | 采样间隔(单位:ns)(默认值:10'000'000,即 10 ms) |
|
|
|
|
|
| [f:] | 将输出转储到指定路径 |
|
|
|
|
|
| [d:] | 运行评测指定秒 |
|
|
|
|
|
| [e:] | 要跟踪哪个事件(cpu, alloc, lock, cache-misses 等),默认是 cpu |
|
|
|
|
|
|
|
|
|
|
### 启动 profiler
|
|
|
|
|
|
|
|
|
|
```
|
|
|
|
|
$ profiler start
|
|
|
|
|
Started [cpu] profiling
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
::: tip
|
|
|
|
|
默认情况下,生成的是 cpu 的火焰图,即 event 为`cpu`。可以用`--event`参数来指定。
|
|
|
|
|
:::
|
|
|
|
|
|
|
|
|
|
### 获取已采集的 sample 的数量
|
|
|
|
|
|
|
|
|
|
```
|
|
|
|
|
$ profiler getSamples
|
|
|
|
|
23
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
### 查看 profiler 状态
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
$ profiler status
|
|
|
|
|
[cpu] profiling is running for 4 seconds
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
可以查看当前 profiler 在采样哪种`event`和采样时间。
|
|
|
|
|
|
|
|
|
|
### 停止 profiler
|
|
|
|
|
|
|
|
|
|
#### 生成 html 格式结果
|
|
|
|
|
|
|
|
|
|
默认情况下,结果文件是`html`格式,也可以用`--format`参数指定:
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
$ profiler stop --format html
|
|
|
|
|
profiler output file: /tmp/test/arthas-output/20211207-111550.html
|
|
|
|
|
OK
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
或者在`--file`参数里用文件名指名格式。比如`--file /tmp/result.html` 。
|
|
|
|
|
|
|
|
|
|
### 通过浏览器查看 arthas-output 下面的 profiler 结果
|
|
|
|
|
|
|
|
|
|
默认情况下,arthas 使用 3658 端口,则可以打开: [http://localhost:3658/arthas-output/](http://localhost:3658/arthas-output/) 查看到`arthas-output`目录下面的 profiler 结果:
|
|
|
|
|
|
|
|
|
|

|
|
|
|
|
|
|
|
|
|
点击可以查看具体的结果:
|
|
|
|
|
|
|
|
|
|

|
|
|
|
|
|
|
|
|
|
::: tip
|
|
|
|
|
如果是 chrome 浏览器,可能需要多次刷新。
|
|
|
|
|
:::
|
|
|
|
|
|
|
|
|
|
### profiler 支持的 events
|
|
|
|
|
|
|
|
|
|
在不同的平台,不同的 OS 下面,支持的 events 各有不同。比如在 macos 下面:
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
$ profiler list
|
|
|
|
|
Basic events:
|
|
|
|
|
cpu
|
|
|
|
|
alloc
|
|
|
|
|
lock
|
|
|
|
|
wall
|
|
|
|
|
itimer
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
在 linux 下面
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
$ profiler list
|
|
|
|
|
Basic events:
|
|
|
|
|
cpu
|
|
|
|
|
alloc
|
|
|
|
|
lock
|
|
|
|
|
wall
|
|
|
|
|
itimer
|
|
|
|
|
Perf events:
|
|
|
|
|
page-faults
|
|
|
|
|
context-switches
|
|
|
|
|
cycles
|
|
|
|
|
instructions
|
|
|
|
|
cache-references
|
|
|
|
|
cache-misses
|
|
|
|
|
branches
|
|
|
|
|
branch-misses
|
|
|
|
|
bus-cycles
|
|
|
|
|
L1-dcache-load-misses
|
|
|
|
|
LLC-load-misses
|
|
|
|
|
dTLB-load-misses
|
|
|
|
|
mem:breakpoint
|
|
|
|
|
trace:tracepoint
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
如果遇到 OS 本身的权限/配置问题,然后 缺少部分 event,可以参考`async-profiler`本身文档:[async-profiler](https://github.com/jvm-profiling-tools/async-profiler)
|
|
|
|
|
|
|
|
|
|
可以用`--event`参数指定要采样的事件,比如对`alloc`事件进入采样:
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
$ profiler start --event alloc
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
### 恢复采样
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
$ profiler resume
|
|
|
|
|
Started [cpu] profiling
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
`start`和`resume`的区别是:`start`是新开始采样,`resume`会保留上次`stop`时的数据。
|
|
|
|
|
|
|
|
|
|
通过执行`profiler getSamples`可以查看 samples 的数量来验证。
|
|
|
|
|
|
|
|
|
|
### 使用`execute`来执行复杂的命令
|
|
|
|
|
|
|
|
|
|
比如开始采样:
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
profiler execute 'start,framebuf=5000000'
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
停止采样,并保存到指定文件里:
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
profiler execute 'stop,file=/tmp/result.html'
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
具体的格式参考: [arguments.cpp](https://github.com/jvm-profiling-tools/async-profiler/blob/v2.5/src/arguments.cpp#L50)
|
|
|
|
|
|
|
|
|
|
### 查看所有支持的 action
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
$ profiler actions
|
|
|
|
|
Supported Actions: [resume, dumpCollapsed, getSamples, start, list, execute, version, stop, load, dumpFlat, actions, dumpTraces, status]
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
### 查看版本
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
$ profiler version
|
|
|
|
|
Async-profiler 1.6 built on Sep 9 2019
|
|
|
|
|
Copyright 2019 Andrei Pangin
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
### 配置 framebuf 参数
|
|
|
|
|
|
|
|
|
|
> 如果遇到生成的火焰图有 `[frame_buffer_overflow]`,则需要增大 framebuf(默认值是 1'000'000),可以显式配置,比如:
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
profiler start --framebuf 5000000
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
### 配置 include/exclude 来过滤数据
|
|
|
|
|
|
|
|
|
|
如果应用比较复杂,生成的内容很多,想只关注部分数据,可以通过 include/exclude 来过滤。比如
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
profiler start --include 'java/*' --include 'demo/*' --exclude '*Unsafe.park*'
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
> include/exclude 都支持设置多个值 ,但是需要配置在命令行的最后。
|
|
|
|
|
|
|
|
|
|
### 指定执行时间
|
|
|
|
|
|
|
|
|
|
比如,希望 profiler 执行 300 秒自动结束,可以用 `-d`/`--duration` 参数指定:
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
profiler start --duration 300
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
### 生成 jfr 格式结果
|
|
|
|
|
|
|
|
|
|
> 注意,jfr 只支持在 `start`时配置。如果是在`stop`时指定,则不会生效。
|
|
|
|
|
|
|
|
|
|
```
|
|
|
|
|
profiler start --file /tmp/test.jfr
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
`file`参数支持一些变量:
|
|
|
|
|
|
|
|
|
|
- 时间戳: `--file /tmp/test-%t.jfr`
|
|
|
|
|
- 进程 ID: `--file /tmp/test-%p.jfr`
|
|
|
|
|
|
|
|
|
|
生成的结果可以用支持 jfr 格式的工具来查看。比如:
|
|
|
|
|
|
|
|
|
|
- JDK Mission Control : https://github.com/openjdk/jmc
|
|
|
|
|
- JProfiler : https://github.com/alibaba/arthas/issues/1416
|
|
|
|
|
|
|
|
|
|
### 生成的火焰图里的 unknown
|
|
|
|
|
|
|
|
|
|
- https://github.com/jvm-profiling-tools/async-profiler/discussions/409
|