A web based monitor for PM2.

Multiple hosts

With the release of 0.11 pm2 no longer uses TCP sockets for the event bus, instead using unix sockets. TCP sockets make monitoring hosts remotely possible so that functionality is currently broken.

Hopefully the pm2 team have a solution for this.

Prerequisites

For debugging to work, node-inspector must be installed and running on the same machine as pm2 (not necessarily the same as pm2-web).

To run

Install pm2-web:

$ npm install -g pm2-web

Then run:

$ pm2-web

Configuration

All configuration options are documented in the default configuration file.

pm2-web will load one of the following files if they exist (in order of preference)

  • A file specified by the --config /path/to/config.json argument
  • From the current users' home directory: ~/.config/pm2-web/config.json
  • A global configuration file: /etc/pm2-web/config.json

The default configuration file is always loaded and other config files will be applied to the default configuration.

The configuration file(s) loaded and the final configuration object will both be recorded in the logs.

Configuration files are loaded using cjson so comments are ok.

All options can be passed as command line arguments and will override any settings found in configuration files.

Authentication

To use HTTP basic auth, set www:authentication:enabled to true in your configuration file. See the default configuration file for more information.

N.b. Your password will be sent in plain text. If you enable HTTP auth, you should probably enable SSL as well.

SSL support

pm2-web can start a https server if so desired. To do so, set www:ssl:enabled to true in your configuration file and supply your certificate details. If you do not have a SSL certificate, the generate_certificate.sh script in the /certs directory will create a self-signed certificate for you.

Debugging running processes

To debug a running process, node-inspector must be installed and running on the same host as the process.

Specify the port it's running on as the inspector property of pm2:host. E.g.:

{
    "pm2": [{
        "host": "foo.baz.com",
        "inspector": 8080
    }]
}

You should then see a debug icon appear next to the stop/restart/reload icons when the process is running.

Clicking this icon will send a SIGUSR1 signal to the process to put it into debug mode and open node-inspector in a new window.

N.b. you may need to change which source file you are looking at in node-inspector to see anything useful.

Debugging multiple processes

By default node will listen for debugger connections on port 5858. If you attempt to debug multiple processes you must specify different debug ports for them:

$ pm2 start --node-args="--debug=7000" foo.js
$ pm2 start --node-args="--debug=7001" bar.js

Debugging multiple instances

This is not possible because:

$ pm2 start --node-args="--debug=7000" -i 4 foo.js

will start four separate processes all listening on port 7000.

If you are expecting to debug your process, please only start one of them.

Reload/restart processes

Restarting a process stops the current process and starts a new one, dropping connections in the process.

Reloading starts a new process in the background, killing the old process after the new one has initialised which reduces downtime.

N.b. your process must exit cleanly (e.g. close sockets, database connections) otherwise the old process will never be killed.

Hard vs soft reloading

Soft reloading (the default) will cause pm2 to send your process a shutdown message and kill it shortly afterwards. Hard reloading will kill it immediately.

To control this behaviour, specify the following in your config file:

{
    "forceHardReload": false
}

To listen for the shutdown event, add the following to your program:

process.on("message", function(message) {
    if(message == "shutdown") {
        // do tidy up here 
    }
});

Resource usage graphs

You can tweak the resource usage graph to be more or less specific. The thinking here is that lots of processes with lots of process usage data will make your browser a bit crashey.

By default it will retain 1000 resource usage measurements of a process (CPU and memory) over a maximum of five days with 40% of the measurements taken in the last 24 hours, 25% from the day before, 10% each from the two days before that and 5% from the day before that.

The update frequency of the graph is controlled by --updateFrequency as detailed above.

--graph.datapoints 1000

The number of data points that will be plotted on the graph in total. If you've got a lot of processes, you may wish to set this to a lowish number to minimise memory consumption in your browser and the pm2-web process itself.

--graph.distribution 40 --graph.distribution 25 --graph.distribution 10 --graph.distribution 10 --graph.distribution 5

The number of --graph.distribution arguments is the number of days worth of data to graph (default 5) and the value is what percentage of --graph.datapoints should be plotted on a given day (the first --graph.distribution argument is today, the second is yesterday, etc).

What this means is that any recent resource usage data will have a more accurate representation in the graph at the cost of consuming more memory and older data will be less accurate but also less likely to crash your browser.

Logs

pm2-web will display log lines emitted by monitored processes after pm2-web itself was started. In order to keep resource usage reasonable by default it will show 1000 log lines.

You can alter this behaviour by specifying --logs:max, so for example to lower this to 500 lines:

--logs:max 500

Release notes

2.0.x

  • Uses 2.x version of pm2-interface, even though it breaks monitoring multiple hosts
  • Shows an error message when attempting to monitor an old/incompatible version of pm2

1.6.x

  • Allow reloading of processes as well as restarting
  • Debug button added to use node-inspector to debug running processes
  • Batch UI updates together in an attempt to improve performance
  • Supports http basic auth
  • Supports serving over HTTPS
  • Serve websockets and UI from a single port to make proxying easier

1.5.x

  • Introduced external configuration file

1.4.x

  • Swapped d3/xCharts for HighCharts due to a memory leak

1.3.x

  • Display logs from processes emitted after pm2-web was started
  • Caches logging output between browser refreshes
  • Respects ANSI colour in logging output

1.2.x

  • Must have been something interesting here

1.1.x

  • Displays graphs of memory and cpu output

1.0.x

  • Initial release
  • Process listing
  • Restarting, stopping & starting processes

Credits

Code by achingbrain, originally based on work done by dunxrion.

Uses pm2 and pm2-interface by unitech.

Special thanks to joewalnes for reconnecting-websocket and Luegg for angularjs-scroll-glue.

pm2-web的更多相关文章

  1. pm2使用

    简单教程 首先需要安装pm2: npm install -g pm2 运行: pm2 start app.js 初次安装并运行,会有一个高大上的界面: 高大上的界面 直接我们介绍过forever,那么 ...

  2. Node.Js —— PM2介绍

    pm2 是一个带有负载均衡功能的Node应用的进程管理器.当你要把你的独立代码利用全部的服务器上的所有CPU,并保证进程永远都活着,0秒的重载, PM2是完美的.它非常适合IaaS结构,但不要把它用于 ...

  3. nodejs高大上的部署方式-PM2

    1.最常用的属nohup了,其实就是在后台执行进程,末尾加个&   [zhoujie@ops-dev ~]$ nohup node /home/zhoujie/ops/app.js & ...

  4. Node之pm2

    最近在项目中使用了Node,在程序部署的时候直接使用命令:node app.js ,这样我们的程序就可以host起来了,但是只要dos窗口关掉之后node就关闭了,这使得我们很不方便,于是乎发现了pm ...

  5. PM2的使用

    PM2 是一个带有负载均衡功能的 Node 应用的进程管理器. 安装 npm install -g pm2 启动程序:pm2 start <app_name|id|all> 列举进程:pm ...

  6. 告别node-forever,拥抱PM2

    告别node-forever,拥抱PM2 返回原文英文原文:Goodbye node-forever,hello PM2 devo.ps团队对JavaScript的迷恋已经不是什么秘密了;node.j ...

  7. PM2 管理nodejs项目

    pm2 是一个带有负载均衡功能的Node应用的进程管理器. 当你要把你的独立代码利用全部的服务器上的所有CPU,并保证进程永远都活着,0秒的重载, PM2是完美的. 它非常适合IaaS结构,但不要把它 ...

  8. pm2安装及常用命令

    安装:npm install -g pm2 启动程序:pm2 start <app_name|id|all> 列举进程:pm2 list 退出程序:pm2 stop <app_nam ...

  9. 使用pm2躺着实现负载均衡

    事实上,pm2 是一个带有负载均衡功能的Node应用的进程管理器,Node实现进程管理的库有很多,forever也是其中一个很强大但是也相对较老的进程管理器. 为什么要使用pm2 对于这个问题,先说说 ...

  10. PM2使用心得

    PM2是node进程管理工具,可以利用它来简化很多node应用管理的繁琐任务,如性能监控.自动重启.负载均衡等,而且使用非常简单. 安装 npm install -g pm2 常用命令 $ npm i ...

随机推荐

  1. json数据操作

    <!DOCTYPE HTML> <html> <head> <meta charset="utf-8"> <title> ...

  2. Python抓取微博评论(二)

    对于新浪微博评论的抓取,首篇做的时候有些考虑不周,然后现在改正了一些地方,因为有人问,抓取评论的时候“爬前50页的热评,或者最新评论里的前100页“,这样的数据看了看,好像每条微博的评论都只能抓取到前 ...

  3. asyncio的简单了解

    asyncio是Python 3.4版本引入的标准库,直接内置了对异步IO的支持. asyncio的编程模型就是一个消息循环.我们从asyncio模块中直接获取一个EventLoop的引用,然后把需要 ...

  4. MVC公开课 – 1.基础 (2013-3-15广州传智MVC公开课)

    1.MVC设计模式 Model 是指 要处理的业务代码和数据操作代码 View 视图 主要是指的 跟用户打交道 并能够展示数据 Controller 看成是 Model和View的桥梁 优点: 1.1 ...

  5. springboot 零xml集成mybatis

    maven依赖 <?xml version="1.0" encoding="UTF-8"?> <project xmlns="htt ...

  6. 查看loadrunner运行日志

    查看loadrunner运行日志   日志分两种 1.在VUGEN中运行后的日志 2.在controller中运行后的日志日志设置分两步: 1.首先,在VUGEN或controller中run-tim ...

  7. 三 Python解释器

    当我们编写Python代码时,我们得到的是一个包含Python代码的以.py为扩展名的文本文件.要运行代码,就需要Python解释器去执行.py文件. 由于整个Python语言从规范到解释器都是开源的 ...

  8. 转:gcc编译C++程序

    转:http://blog.csdn.net/liujiayu2/article/details/49864381 单个源文件生成可执行程序 下面是一个保存在文件 helloworld.cpp 中一个 ...

  9. Django+Nginx+uwsgi搭建自己的博客(三)

    (本来打算在这篇博文中介绍Users App的前端部分的,但写着写着就发现还需要铺垫很多东西才能把整个项目串的比较流畅些,因此这篇就继续介绍了后端的一些东西,前端的部分只好跳票到下一篇了-) 在上一篇 ...

  10. 深入理解javascript作用域系列第二篇

    前面的话 大多数时候,我们对作用域产生混乱的主要原因是分不清楚应该按照函数位置的嵌套顺序,还是按照函数的调用顺序进行变量查找.再加上this机制的干扰,使得变量查找极易出错.这实际上是由两种作用域工作 ...