Erlang Performance Lab

Tool for developers working with systems running on the Erlang VM (BEAM). It helps with performance analysis.

Getting started

The Erlang Performance Lab tool (erlangpl for short) can be started using escript or as a regular Erlang release.

Quick start guide

1. Download the prebuilt `erlangpl` script from here or type in the shell command:

wget https://github.com/erlanglab/erlangpl/releases/download/0.9.0/erlangpl.tar.gz

2. Untar downloaded archive using any GUI program or type in the shell command:

tar -zxvf erlangpl.tar.gz

3. Run the script

./erlangpl -n NODE_NAME -c COOKIE

To make erlangpl work a few conditions must be met:

The node to be monitored must be run in the distributed mode (using -name or -sname flag). Click here to learn more.
Both the node's and the erlangpl's cookies must match. You can set the node's cookie by starting it with -setcookie flag. Keep in mind that when you do not set a cookie for the node it will be automatically set using $HOME/.erlang.cookie. See the official Erlang documentation to learn more.
The erlangpl script must run on OTP 19.3 or higher.

Example

If the node you want to monitor is a@127.0.0.1 and has a cookie set to test (erl -name a@127.0.0.1 -setcookie test) then you should run the script as follows:

./erlangpl -n a@127.0.0.1 -c test

4. Visit http://localhost:37575

Download prebuilt script

The easiest way to get started is to download a prebuilt erlangpl script (download link).

Build it manually

Prerequisites

For building UI you need to have following dependencies installed:

Be aware that building UI can take some time. It takes around 1 minute on stock MacBook 2015 plus dependencies download for the first time. Second time dependencies will be cached.

$ git clone https://github.com/erlanglab/erlangpl.git
$ cd erlangpl
$ make release

Running erlangpl script

The erlangpl shell script is a self-contained escript, which can be started from a command line as long as you have Erlang/OTP installed.

$ ./erlangpl -h

Usage: erlangpl [-n <node>] [-c <cookie>] [-p <plugin>] [-h]
                [-v <verbose>] [-P <port>] [-V] [-s <sname>] [-l <name>]

  -n, --node     Monitored node name
  -c, --cookie   Overwrite ~/.erlang.cookie
  -p, --plugin   Path to plugins
  -h, --help     Show the program options
  -v, --verbose  Verbosity level (-v, -vv, -vvv)
  -P, --port     HTTP and WS port number
  -V, --version  Show version information
  -s, --sname    Start with a shortname
  -l, --name     Start with a longname, default erlangpl@127.0.0.1

$ ./erlangpl -n testnode@127.0.0.1 -c YOURCOOKIE

Once started, try visiting http://localhost:37575/

Examples

Connecting to an Elixir iex session

$ iex --name foo@127.0.0.1 -S mix

$ ./erlangpl --node foo@127.0.0.1

Mnesia cluster

You can generate messages between nodes by querying a distributed database Mnesia.

To setup a Mnesia cluster, start several Erlang nodes with unique names e.g. a@, b@, c@, etc. and start the database on all of them:

erl -name a@127.0.0.1
(a@127.0.0.1)1> mnesia:start().

Then create a test_table and configure it to be replicated on all nodes:

(a@127.0.0.1)2> mnesia:change_config(extra_db_nodes, ['b@127.0.0.1']).
(a@127.0.0.1)3> mnesia:change_config(extra_db_nodes, ['c@127.0.0.1']).
(a@127.0.0.1)4> mnesia:change_config(extra_db_nodes, ['d@127.0.0.1']).
(a@127.0.0.1)5> mnesia:create_table(test_table, []).
(a@127.0.0.1)6> [mnesia:add_table_copy(test_table, Node, ram_copies) || Node <- nodes()].

Here are some behaviours you can test:

[begin mnesia:transaction(fun() -> mnesia:write({test_table, Key, "value"}) end), timer:sleep(10) end || Key <- lists:seq(1,2000)].
[begin mnesia:sync_dirty(fun() -> mnesia:write({test_table, Key, "value"}) end), timer:sleep(10) end || Key <- lists:seq(1,2000)].
[begin mnesia:dirty_write({test_table, Key, "value"}), timer:sleep(10) end || Key <- lists:seq(1,2000)].

Videos from those experiments were posted on YouTube

Developing

Erlang

Running development release

You can also start the tool as a regular Erlang release and connect to its console to debug the tool itself.

$ make
$ rebar -f generate
$ ./rel/erlangpl/bin/erlangpl console node=testnode@127.0.0.1 cookie=YOURCOOKIE

User Interface

Running standalone

erlangpl-ui can be started standalone using Node with npm or yarn. We are recomending yarn for that.

yarn && yarn start

Now, application can be found at localhost:3000 and will be listening for messages from localhost:37575 where you have to have erlangpl running.

Writing Elm code

Although erlangpl-ui is written in React we belive in Elm's power. Because of that we support Elm in our build process. This is possible because of react-elm-components and elm-webpack.

You can write any separate component in Elm and then wrap it into React component which can be integrated with whole application. Elm code should be placed in ui/src/elm and every component whould have main file in this directory and all files related to this component in directory with the same name. React wrapper file should have the same name as Elm component and flow should be disabled for this file.

-- ui/src/elm/About.elm

module About exposing (..)

import Html exposing (text)

main =
    text "Hello world from Elm component"

// ui/src/about/components/About.js

import React from 'react';
import Elm from 'react-elm-components';
import { About } from '../../elm/About.elm';

import './About.css';

const AboutWrapper = () => {
  return (
    <Elm src={About} />
  );
};

export default AboutWrapper;

Have fun!

amilkr/erlangpl